Knowledge base
Glossary of AI personalization
Terms across personalization, AI, analytics and product recommendations — with examples from real retail.
Nothing found in the glossary. Try another query.
↑↓ navigate
↵ open
Esc close
Filter:
A
60 terms
A/A Testing
An A/A test is run before the first A/B experiment, or after technical changes to the analytics stack, to confirm that traffic allocation, conversion tracking and the statistics engine behave correctly.
A/B Testing
An A/B test is a controlled statistical experiment in which a single variable changes between two versions while every other factor stays identical. That isolation is what lets you claim a causal link between the change and the result, rather than a correlation.
AARRR Framework
AARRR is a framework of five customer lifecycle stages (Acquisition, Activation, Retention, Revenue, Referral), each with its own metric. It is used diagnostically, to answer which stage is capping the growth of the business.
ABC Analysis
ABC analysis is the classification of products or customers into three groups by the cumulative share of their contribution to a chosen metric: A is about 80% of the result, B about 15%, C about 5%.
Activation
Activation is the stage of the user lifecycle at which a person first gets the core value of a product; the activation rate metric shows the share of new users who reach that moment.
Ad Cost Ratio
Ad cost ratio (ACoS) is advertising spend divided by the revenue that advertising produced, expressed as a percentage. It is the inverse of ROAS: ad cost ratio = 1 / ROAS × 100%.
Add-to-Cart Rate
Add-to-cart rate is the share of visits in which the shopper added at least one item to the cart; an intermediate funnel KPI reflecting product page effectiveness.
Ads in AI Answers
Ads in AI answers are ad formats that generative search platforms and AI assistants show inside or next to a generated response, matching the ad to the context of the query and the conversation.
Affinity Profile
An affinity profile is a numeric vector in which each dimension corresponds to a specific category, brand or attribute — interest in Nike, interest in the Sneakers category, the $70–140 price band. Values are time-decayed, so recent actions weigh more heavily than older ones.
Agent Autonomy Levels
Agent autonomy levels are a 0-to-4 scale describing how much of a purchase an AI agent carries out itself: from answering questions to making recurring purchases on its own.
Agent Experience (AX)
Agent experience (AX) is the quality of an AI agent's interaction with a site or service: how easily the agent understands the offer, gains access and completes the user's task.
Agent Memory
Agent memory is the storage and reuse of information about a user and past interactions beyond a single request to the model: within the session context (short-term) and in external storage (long-term).
Agent Orchestration
Agent orchestration is the coordination of several specialised AI agents on a shared goal, achieved by distributing subtasks and assembling their outputs — a multi-agent system.
Agent Payment Networks
Agent payment networks are the Visa (Intelligent Commerce) and Mastercard (Agent Pay) schemes that issue tokenized payment credentials so AI agents can pay securely on a user's behalf.
Agent Payments Protocol
AP2 is Google's open protocol for agent payments. It uses cryptographically signed intent and cart mandates so that a purchase made by an AI agent is provably authorised.
Agent Skills
An agent skill is a folder with a SKILL.md file, instructions and supporting resources that an AI agent loads into its context only when a task matches the skill's description.
Agent-Readiness
Agent-readiness is how prepared a store is for agentic commerce: a machine-readable catalogue, open APIs, support for agent protocols and payment tokens.
Agent-Ready Product Feed
An agent-ready product feed is a product feed carrying the full set of data an agent needs to search, compare and buy: identifiers, attributes, variants, current price and availability, delivery and returns.
Agent-to-User Interface (A2UI)
A2UI is an open declarative format from Google in which an AI agent describes an interface as JSON components and the client application displays them with its own native widgets, without executing the agent's code.
Agent2Agent Protocol
A2A is an open Google protocol for AI agents to interact with one another, exchanging tasks and data; it serves as the base layer for the AP2 payment protocol.
Agentic Browser
An agentic browser is a browser with a built-in AI agent that works with websites on the user's behalf: it opens pages, clicks, fills in forms and places orders.
Agentic Commerce
Agentic commerce is a model of trade where an AI agent autonomously finds products, compares options and completes a purchase on a person's behalf through standardised protocols for interacting with stores.
Agentic Commerce Protocol (ACP)
ACP is the open protocol from OpenAI and Stripe for purchases inside an AI dialogue; it powers Instant Checkout in ChatGPT and uses a shared payment token for secure settlement.
Agentic Workflow
An agentic workflow is a multi-step mode of AI operation in which the agent decomposes a goal, executes subtasks through tools and refines the result iteratively, rather than returning a single response.
Aha Moment
The aha moment is a specific action or event in the user experience that correlates statistically with long-term retention; it is the point at which the core value of the product is realised.
AI Agent
An AI agent is an LLM-based system that autonomously plans and executes multi-step tasks through tool and API calls, rather than stopping at a text response.
AI Agent Traffic
AI agent traffic is visits to a website that an AI agent makes on a person's instruction in real time, as opposed to background crawling of the site for model training or indexing.
AI Citability
AI citability is the frequency and quality of mentions of or links to a brand's content in the generative answers of LLM search engines and assistants; the new visibility KPI of generative search.
AI Crawlers
AI crawlers are robots run by AI companies that crawl websites to train models, build AI search indexes or fetch a page on a user's request; they are controlled in robots.txt by bot name.
AI Hallucination
An AI hallucination is a language model producing factually incorrect or invented statements with high confidence, unsupported by any real data.
AI Overviews
AI Overviews is a Google feature that displays an AI-generated summary answer with source citations at the top of the results page, redistributing organic traffic between sites.
AI Referral Traffic
AI referral traffic is visits to a website from links in the answers of AI assistants and generative search, where a human clicks the link rather than an AI agent or a crawler.
AI Review Summaries
AI review summaries are a digest of customer reviews generated by a language model: the main pros, cons and themes that many different reviewers mention.
AI Shopping Agent
An AI shopping agent is a buyer-side agent that autonomously selects and purchases products across stores on a person's behalf, as opposed to an AI shopping assistant on the seller's side.
AI Shopping Assistant
An AI shopping assistant is a seller-side conversational assistant that advises the shopper across the retailer's catalogue and guides them to a purchase. Unlike a shopper-side AI agent, it does not act autonomously across different stores.
Alt Text
Alt text is an HTML image attribute holding a text description of the image contents; it influences ranking in image search and is a baseline requirement of web accessibility.
Alternating Least Squares
ALS is an iterative matrix factorisation algorithm for collaborative filtering that optimises latent user and item vectors in alternating turns using least squares.
Answer Engine Optimization (AEO)
Answer engine optimization is the optimisation of content for direct display as an answer in search engines, voice assistants and AI interfaces, without the user needing to visit the site.
API Key
An API key is a secret identifier that authenticates requests to an API; it belongs on the server and must not appear in public frontend code.
App Onboarding
App onboarding is the sequence of steps on first opening a mobile app, designed to demonstrate the product's value quickly and draw the user into the first key action.
App Rating Prompt / In-App Review
An app rating prompt is the native system dialogue for rating an app, driven through the official store APIs and shown at carefully chosen moments in order to maximise positive reviews.
App Store Listing / Store Product Page
A store listing is the combined media and text of an app's page in the App Store or Google Play, forming the first impression and determining the conversion rate from view to install.
App Store Optimization (ASO)
App Store Optimization (ASO) is the optimisation of metadata and visual assets on an app's page in the App Store or Google Play, aimed at raising organic visibility and the conversion rate from view to install.
App Tracking Transparency (ATT)
ATT is Apple's mandatory framework for requesting user consent to cross-app tracking; without an allow the IDFA is unavailable and cross-property attribution is impossible.
App Version Fragmentation
App version fragmentation is the spread of active users across several app versions, which complicates API support, limits the rollout of new features, and requires a policy for the minimum supported version.
Apple Push Notification service (APNs)
APNs is Apple's proprietary infrastructure for delivering push notifications to every device in the Apple ecosystem; it is the only authorised channel for iOS push and requires the server to authenticate with a certificate or a key.
Application Not Responding (ANR)
ANR is a critical state of an Android app in which the main thread is blocked for longer than the permitted limit; Google Play factors ANR rate into ranking and can restrict distribution of the app.
Application Programming Interface (API)
A REST API is the most common type: requests and responses travel over HTTP in JSON. The main methods are GET (read), POST (send), PUT and PATCH (update) and DELETE (remove).
Approximate Nearest Neighbor Search
Approximate nearest neighbor search is a family of algorithms that find vectors close to a query with a controllable trade-off between accuracy and speed, using specialised index structures.
Assisted Conversions
An assisted conversion is a conversion whose touch chain a channel took part in without being the last one. The metric exposes the supporting role of a source, which is invisible in a last-click attribution model.
Assortment Width and Depth
Assortment width is the number of product categories and groups in a store's matrix; assortment depth is the number of variants and SKUs inside an individual group. Together they describe the structure of the product offer.
Async / Message Queue
An asynchronous message queue is a data transfer pattern in which events are placed in a buffer and processed independently of the main request flow; it guarantees reliable event delivery under peak load.
Attention Mechanism
Attention is an operation in neural networks that computes a weighted sum of value vectors from the pairwise similarity between a query and a set of keys, letting a model focus dynamically on the relevant parts of its input.
Attribution
A user can make 5–7 touches with a brand before buying: saw a banner, arrived from search, opened an email, came back through retargeting, bought. Attribution answers the question of which of those channels the sale is credited to.
Attribution Model
An attribution model is the algorithm for distributing the value of a conversion across the marketing channels involved in the user's path to a purchase.
Attribution Window
An attribution window sets the shelf life of a click or an impression. If a user clicks a recommended product and buys it five days later, a 7-day window credits the conversion to the recommendation, while a 3-day window does not.
Audience Segment
An audience segment is a group of users sharing characteristics or behavioural patterns, used as the targeting unit in personalization and marketing campaigns.
Autosuggest
Autosuggest is a site-search feature that, while the shopper is typing, offers query completions, categories and specific products, so they can reach a result without typing the full query.
Average Order Value (AOV)
AOV = revenue / number of orders. Period and channel both matter: AOV varies sharply by traffic source (organic vs retargeting), device (desktop vs mobile) and audience segment (new vs returning shoppers).
Average Revenue per User
ARPU = revenue / MAU (monthly active users). For a sharper read, teams use ARPPU (average revenue per paying user) = revenue / number of paying users, which excludes everyone who did not buy.
B
14 terms
Backend-Driven UI
BDUI (Backend-Driven UI) is an approach in which the server describes the structure and content of a screen and the client renders the schema it receives, which allows the interface to be updated and personalized without an app release.
Backlink
A backlink is an inbound hyperlink from an external site that passes link equity (PageRank) and acts as one of the key off-page ranking factors in search engines.
Badges / Achievements
Badges are visual rewards inside a gamification system that a user earns for reaching specific goals; they are used to encourage target actions and raise engagement.
Banner Blindness
Banner blindness is the effect where a user does not perceive interface elements they have recognised as advertising, even when those elements are in the field of view and carry information the user needs.
Batch Import / Export
Batch import is the transfer of data into a system in scheduled packages; in personalization platform integrations it is used for the initial load of the catalogue and historical behavioural data.
Bayesian Statistics
Bayesian statistics in A/B testing expresses the result as the probability that a variation beats the control, which lets you decide as data accumulates instead of waiting for a fixed p-value threshold.
Behavioral Segmentation
Behavioral segmentation groups users by their actions — purchase history, product views, clicks and visit frequency — in order to build personalized communications and recommendations.
Behavioral Signals
Behavioral signals are measures of how users interact with a site and with the search results page — click-through rate, time on site, returns to search — that ranking algorithms take into account, most explicitly at Yandex.
Behavioral Targeting
Behavioral targeting is the selection of an audience and of message content from recorded user actions — views, searches, add-to-cart events, exits, campaign reactions — rather than from demographic characteristics.
Bestsellers
Bestsellers are recommendations of the top-selling products, used as a fallback strategy when no personal profile is available.
Bias-Variance Tradeoff
The bias-variance tradeoff is the principle that a model's error is the sum of systematic error (bias) and sensitivity to the data (variance), and that the two cannot be minimised at once without more data or ensembling.
Bounce Rate
Bounce rate is the share of visits in which the user made no interaction at all and left the site after viewing a single page.
Breadcrumbs
Breadcrumbs are a navigation element built as a chain of links from the site root to the current page; they help visitors orient themselves in the catalogue hierarchy and create internal linking for search engines.
Business-to-Agent (B2A)
B2A (business-to-agent) is a sales model in which the buyer is an AI agent: it acts on behalf of a person or a company and picks an offer on structured data and the terms of the deal.
C
56 terms
Caching / Cache
Caching is the storage of intermediate computation or request results to speed up repeat access; the TTL, or cache lifetime, sets the balance between how current the data is and how much load the system carries.
Canonical Tag
A canonical tag is the HTML attribute rel=canonical that tells a search engine the preferred URL for a page when several versions of it exist with identical or near-identical content.
Cart Abandonment
Cart abandonment is an unfinished purchase: the shopper added products to the cart but never placed the order. The market average abandonment rate is around 70%.
Catalog Coverage
Catalog coverage is a recommender metric measuring the share of items that appeared in at least one recommendation over a period, relative to the full catalogue.
Category Management
Category management is running a product category as a separate business unit: defining its role in the assortment, building the matrix and price tiers, planning promotions and owning the category's own KPIs.
Challenge
A challenge is a time-limited task with a reward, used in e-commerce to encourage specific behavioural patterns such as buying in new categories or visiting more often.
Chatbot
A chatbot is a program that holds a dialogue with a user in a text or voice interface to complete their task: answering a question, finding a product or performing an operation in the company's systems.
Checkout
Checkout is the process of placing and confirming an order after the cart, covering delivery selection, payment and the entry of customer details.
Churn Rate
Churn rate = customers lost in a period / customers at the start of the period × 100%. In e-commerce, lost usually means no purchase within 90–180 days depending on the category. A five percent reduction in churn is widely cited as lifting profit by 25–95%.
CI/CD
CI/CD is a pair of automated pipelines for continuous integration and deployment; they shorten integration development cycles, check event schemas automatically and get changes live quickly.
Click and Collect
Click and collect is a purchase scenario where the order is placed in an online channel and handed over at one of the retailer's physical locations: a store, a pickup point or a parcel locker.
Click-through Rate
CTR is a top-of-funnel engagement metric. A high CTR means the content attracts attention, but it does not guarantee conversion. To judge the full effect of a recommendation widget, use CTR plus attributed CR plus attributed revenue.
Client-Side Rendering
Client-side rendering is an architecture in which the browser receives a JavaScript application and builds the interface itself, fetching data through an API; it is characteristic of single page applications and needs its own strategy for SEO and personalization.
Clustering
Clustering is the task of splitting a set of objects into non-overlapping groups (clusters) that maximise similarity inside each group and difference between groups.
Cohort Analysis
Cohort analysis is a method of comparing groups of users united by a shared event in one period, in order to track how their behaviour and retention change over time.
Cold Start / Warm Start / Hot Start
Cold, warm and hot start are the three levels of mobile app launch, defined by how much state survives in memory; a cold start requires full initialisation and is a critical parameter of perceived performance.
Cold Start Problem
The cold start problem is what happens to a recommender system when there is no historical data on a user or a product, which makes accurate personalization impossible through standard collaborative filtering. It is a data problem, not a model problem — a better algorithm does not solve it.
Collaborative Filtering
Collaborative filtering is a recommendation algorithm that predicts a user's preferences from the behaviour of similar users or from patterns of joint consumption between items. Unlike content-based filtering, it never looks at what a product is — only at how people interact with it.
Commercial Ranking Factors
Commercial ranking factors are the group of ranking signals that judge an online store as a merchant: assortment and availability, prices, product page completeness, delivery and payment terms, contact details, reviews and checkout convenience.
Complete the Look
Complete the Look is a set of recommendations for stylistically compatible products in fashion retail, helping a shopper assemble a full outfit from several items.
Computer Use
Computer use is an AI agent's ability to drive a computer interface — cursor, clicks, forms — as a human does, which allows it to buy on websites that expose no dedicated API or protocol.
Confidence Interval
A confidence interval is the range within which the true population parameter lies at a chosen probability, usually 95%; in A/B testing it expresses the precision of the measured effect.
Consent Management Platform
A CMP has three jobs: (1) show the user information about cookie categories and request consent; (2) store the proof of consent as an audit trail; (3) block or enable analytics and marketing scripts according to the user's choice.
Content Delivery Network
A CDN is a distributed network of servers that delivers static content to visitors with minimal delay through geographical proximity; in e-commerce it is used to speed up loading of a store's assets.
Content-based Filtering
In content-based filtering each item is described as an attribute vector. The user profile is the weighted average of the attributes of the items they viewed or bought, and the recommendations are the items with the highest cosine similarity to that profile.
Context Engineering
Context engineering is the design and assembly of the full set of information (instructions, data, history, tools, memory) an LLM receives on each call, so that the task can be solved within the context window.
Context Window
A context window is the maximum number of tokens a language model takes into account in one request, including the input (instructions, history, documents) and the generated answer.
Control Group
A control group is the segment of an A/B test shown the current version unchanged; it is the reference against which the incremental effect of the tested variation is measured.
Conversational AI
Conversational AI is natural language processing applied to multi-turn interaction between a person and a machine in text or voice, with context preserved across turns.
Conversational Commerce
Conversational commerce is selling through a conversational interface — chat, voice or messaging — rather than a classic storefront. It covers both scripted bots and autonomous AI agents.
Conversion
A conversion is a user completing a predefined target action — a purchase, a registration, a subscription. The ratio of conversions to visitors is the conversion rate.
Conversion Funnel
A conversion funnel is a multi-stage model of the path from first contact to a target action; the instrument for locating loss points and prioritising optimisation.
Conversion Rate (CR)
CR is measured across the funnel: visitor to order (the macro rate), category to PDP, PDP to cart, cart to order. Growth in the overall rate usually comes from improving the bottleneck — the step with the heaviest drop-off.
Conversion Rate Optimization (CRO)
CRO is the discipline of raising conversion systematically through a loop of data collection, hypothesis formulation, prioritization and validation by controlled experiment.
Cookie / HTTP Cookie
A cookie is a small piece of data stored by the browser on the instruction of a server or a script and sent automatically with HTTP requests; in e-commerce it identifies anonymous sessions and holds visitor preferences.
Cooldown Period
A cooldown period is a mandatory time interval between communications sent to the same user, ruling out message overload when several triggers fire at once.
Cost Per Action
CPA (cost per action) is the cost of one target action: ad spend divided by the number of completed actions. It also denotes a media buying model where payment is per action rather than per impression or click.
Cost Per Click
CPC (cost per click) is the average price of one click on an advert, equal to campaign spend divided by the number of clicks. The same acronym names the buying model in which the advertiser pays per click.
Cost Per Mille
CPM (cost per mille) is the cost of a thousand impressions of an advert, calculated as spend divided by impressions and multiplied by 1000. It is also the buying model in which the advertiser pays for impressions rather than clicks or actions.
Cost Per Order
CPO is the cost of generating one order from a marketing channel; it shows how efficient spend is relative to average order value and order margin.
Crash Rate
Crash rate is the share of app sessions that terminate abnormally; a core stability indicator that affects store ranking and user retention.
Crawl Budget
Crawl budget is the number of pages a search bot allocates to crawling a site in a given period; spending it efficiently is critical for large sites with a constantly changing catalogue.
Crawling
Crawling is the automated process in which a search robot walks the pages of a site through hyperlinks in order to collect content for subsequent indexation.
CRM Marketing
CRM marketing is an approach to marketing based on customer data: segmenting the base and personalizing communications across the whole lifecycle to grow repeat sales, retention and lifetime value.
Cross-Origin Resource Sharing
CORS is a browser security mechanism that restricts cross-origin requests; a personalization API must explicitly permit a client's domains through the Access-Control-Allow-Origin header.
Cross-Sell
Cross-sell is the practice of offering complementary products from adjacent categories to a customer who has already chosen the main item, with the goal of raising average order value.
Customer Acquisition Cost
CAC = marketing and sales spend / number of new customers in a period. It matters to calculate CAC per channel: CAC from paid search, SEO and email differ radically, and a blended CAC hides the inefficiency of individual sources.
Customer Data Platform (CDP)
Unlike a CRM, which manages sales and communications, and a DMP, which handles anonymous audiences for advertising, a CDP works with identified customers and builds persistent profiles enriched in real time. Its core value is a single customer view that holds regardless of the channel an interaction happened in.
Customer Deduplication
Customer deduplication is the process of identifying and merging records that belong to the same individual into a single profile with reconciled attributes, history and consents.
Customer Effort Score
CES (customer effort score) is the customer effort index: the average rating of how easy a specific interaction with the company was for the user. It is captured with a single question straight after the action on a 1–7 or 1–5 scale.
Customer Experience (CX)
Customer experience (CX) is the sum of every impression a shopper takes from interacting with a brand along the whole path, from first awareness to a repeat purchase.
Customer Journey
The customer journey is the sequence of stages a customer moves through from first contact to repeat purchase; the basis for deciding where personalization is placed.
Customer Journey Map
A CJM describes the customer path along several dimensions: stages (awareness, consideration, purchase, loyalty), the channels at each stage, the customer's goals and expectations, the barriers and pain points, and the metrics used to measure them.
Customer Lifetime Value (CLV)
Customer lifetime value (LTV) is the projected total revenue from a customer across the entire relationship with the company, used to assess marketing efficiency and unit economics.
Customer Persona / Buyer Persona
A customer persona is a semi-fictional but data-grounded portrait of a typical customer, describing motivation, pain points and behavioural patterns, and used to guide marketing and product decisions.
Customer Satisfaction Score
CSAT is a measure of satisfaction with one specific interaction, collected immediately after the event — a purchase, a delivery, a support contact — on a numeric scale.
D
14 terms
D1 / D7 / D30 Retention
D1, D7 and D30 retention are the standard checkpoints for mobile audience retention: the share of users who return to the app 1, 7 and 30 days after first install.
Dark Patterns
Dark patterns are interface and copy elements that push a user towards a decision in the business's interest through untruthful information or a deliberately obstructed alternative.
Data Layer
A data layer is a structured JavaScript object on the page that accumulates context and event data so it can be routed to analytics and marketing systems through a tag manager.
Deep Learning
The depth of a neural network is its number of hidden layers. Each layer extracts increasingly abstract features: the first captures edges and textures in an image, the last captures high-level concepts. Deep networks demand large volumes of data and GPU compute, but outperform classical ML methods on complex tasks.
Deep Link
A deep link is a URL scheme that routes a person to a specific screen or piece of content inside a mobile app, bypassing the home screen and shortening the path to the target action.
Deferred Deep Link
A deferred deep link is a mechanism that preserves contextual information — the target screen and its parameters — through app installation, so the right screen opens at the first launch after install.
Demographic Segmentation
Demographic segmentation groups users by basic social attributes — age, gender, region, income — providing the first-level audience cut used to configure marketing campaigns.
Device Fingerprinting
Device fingerprinting is a probabilistic method of identifying a device from a combination of indirect attributes, used as an alternative to deterministic advertising IDs when those are unavailable.
Domain Name System
DNS is the system that translates domain names into IP addresses; it governs internet traffic routing, CDN configuration, certificates and the subdomains used for first-party personalization integrations.
Downsell
A downsell is an offer of a more affordable version of a product to a shopper who is refusing the original offer or hesitating over price, aimed at saving the deal instead of losing it.
Drip Campaign
A drip campaign is a pre-planned series of automated messages sent to a user at fixed intervals or when trigger conditions are met.
DWH — Data Warehouse
A DWH differs from operational databases (OLTP) in its architecture: it is optimised for analytical queries (OLAP) — aggregating millions of records, complex joins, historical data. Data arrives through ETL or ELT pipelines from operational systems.
Dynamic Content
Dynamic content is content on site pages that renders differently for different users or segments in real time. The URL stays the same for everyone — only what is inside the page changes.
Dynamic Segment
A dynamic segment is an audience group with automatically maintained membership: users enter and leave in real time as they meet or stop meeting the defined conditions.
E
12 terms
E-commerce
E-commerce is the sale of goods and services through online channels — a website, an app, a marketplace — with a measurable funnel from traffic acquisition to repeat purchase.
E-E-A-T
E-E-A-T is Google's framework — Experience, Expertise, Authoritativeness, Trust — for judging the credibility and quality of content; it is applied by human raters and influences ranking algorithmically in high-stakes topics.
Edge Computing / Edge Functions
Edge computing is the execution of code on CDN nodes near the visitor; in e-commerce it is used for personalization, A/B testing and content control with minimal delay and no call to a central server.
Email Segmentation
Email segmentation is the division of a subscriber base into groups by behavioural and demographic signals in order to raise the relevance and effectiveness of email communications.
Email Service Provider
An ESP is an email sending service responsible for dispatching messages, technical deliverability, sender domain reputation, subscriber management, templates and automations inside the email channel.
Embeddings
Embeddings are dense numeric vectors of fixed dimensionality that encode objects — products, users, queries — so that geometric distance between vectors reflects semantic similarity.
End-to-end Analytics
End-to-end analytics joins advertising systems, the site, the CRM and payment or accounting systems into one chain from traffic source to a paid, non-returned order, allowing channels and changes to be judged on real money.
Engagement Rate
Engagement rate is the share of sessions that show signs of real user interaction; in GA4 terms it is the opposite of bounce rate.
Event API
An Event API is the interface for sending structured behavioural events into a platform, supplying the data that personalization and segmentation models run on.
Event Tracking
Event tracking is the recording of specific user actions in an analytics system together with their parameters; the basis of behavioural analytics and personalization.
Event-Based Trigger
An event-based trigger is the condition that automatically launches a marketing reaction when a specific user event is recorded, giving the communication maximum relevance to the moment.
Exit Intent
Exit intent is a trigger that fires when a visitor shows the intent to leave the site, and is used to show a personalized retention offer at that moment.
F
14 terms
Faceted Search
Faceted search is multi-attribute catalogue filtering that lets a shopper combine several parameters to narrow the result set.
Feature Adoption
Feature adoption rate is the share of active users who used a specific feature during a period; an indicator of how successful a launch was and how visible new capabilities are.
Feature Engineering
Feature engineering is the process of extracting, transforming and constructing input variables (features) from raw data in order to increase the predictive power of a machine learning model.
Feature Flag / Feature Toggle
A feature flag is a software switch that controls whether functionality is available to users without changing or re-releasing the code; it is used for safe rollouts, A/B testing and emergency rollback.
Fine-tuning
Fine-tuning is a transfer learning method in which a pre-trained model continues optimising on a specialised target data set in order to adapt to a particular task or domain.
Firebase Cloud Messaging (FCM)
FCM is Google's cloud platform for sending push notifications to Android devices and browsers; it transports messages from an app's server to a specific device addressed by a unique push token.
First-party Cookie
A first-party cookie is a cookie set by the domain currently being visited; it is not blocked by browsers and is used to identify visitors, hold the basket and drive personalization without depending on the advertising ecosystem.
First-Party Data
First-party data is data a company collects directly from its own users through channels it owns — website, app, CRM — which makes it the most accurate, the most current and the easiest to keep compliant.
Force Update / Soft Update
A force update is a mechanism that compels a mobile app update by blocking access until the new version is installed; a soft update is an optional reminder that leaves functionality intact.
Frequency Capping
Frequency capping is a limit on the number of impressions of a message or advertisement shown to one person within a defined period, to prevent audience fatigue.
Frequentist Statistics
Frequentist statistics is an approach to testing built on the p-value and a sample size fixed in advance, in which probability is read as the long-run frequency of an event across repeated trials.
Frequently Bought Together
FBT is a recommendation algorithm that suggests products based on an analysis of what other shoppers bought together inside a single order.
Function Calling / Tool Use
Function calling, or tool use, is the mechanism that lets an LLM invoke external tools and APIs in a structured format — the foundation that turns a model into an acting agent.
Funnel Analysis
Funnel analysis measures the percentage of users passing each step of a defined sequence of events; it is used to locate where the audience is lost in the largest numbers.
G
11 terms
Gamification
Gamification is the use of game elements — points, levels, achievements, progress, competition — in e-commerce and marketing to raise engagement and build regular behavioural patterns.
General Data Protection Regulation
The GDPR is the EU regulation on personal data protection, setting requirements for the lawful basis of processing, user consent, data subject rights (access, erasure, portability) and minimisation of the data collected.
Generative AI
Generative AI refers to machine learning models that create new content (text, images, code, audio) on request by reproducing the patterns of the data they were trained on.
Generative Engine Optimization (GEO)
Generative engine optimization is the practice of optimising content and technical infrastructure so that a site is cited and linked in the answers of generative search engines and LLM assistants.
Google Analytics 4
GA4 is Google's event-based web analytics platform with web and app support and built-in ML predictions; it replaced Universal Analytics as the standard for web measurement.
Gradient Boosting
Gradient boosting is an ensemble learning algorithm that iteratively adds decision trees which minimise the ensemble's residual error through gradient descent in function space.
GraphQL
GraphQL is a query language for APIs that lets the client describe exactly the shape of the data it needs in one request; in headless e-commerce it is used as an alternative to REST for flexible data aggregation.
Gross Margin
Gross margin is the share of revenue that remains after the cost of goods sold is deducted; it shows the base profitability of sales before operating expenses.
Gross Merchandise Value
GMV is the total value of all orders placed over a period, before returns and discounts are deducted; it is used as an indicator of the scale of trading activity on a platform.
Grounding
Grounding is tying a language model's answer to verifiable external data (catalogue, prices, availability) so that every claim in the answer is backed by a source rather than the model's memory.
Guardrails
Guardrails are rules, filters and checks around an LLM that keep incoming requests, model answers and model actions within the application owner's policy.
H
9 terms
Headless Commerce
Headless commerce is an architecture that separates an online store's frontend from its backend through an API; it simplifies integrating personalization, search and analytics as independent services.
Heatmap
A heatmap is a colour-coded visualisation of how user activity — clicks, scrolling, mouse movement — is distributed across a page; it supports qualitative UX analysis and helps prioritise optimisation work.
Holdout Group
A holdout group is an isolated cohort of users excluded from all active experiments and changes, used to measure the cumulative long-term effect of an optimisation programme.
Homepage Personalization
Homepage personalization is the adaptation of hero banners and key homepage content to the profile, segment or interaction history of an individual visitor.
hreflang
hreflang is an HTML attribute that tells search engines which version of a page is intended for a given language and region, and that links the language alternates of one page together.
HTTPS / TLS
HTTPS is HTTP encrypted with TLS; it guarantees the confidentiality and integrity of data between browser and server and is mandatory for any online store and for transmitting API tokens.
Human-in-the-Loop
Human-in-the-loop is an approach in which an AI agent works autonomously while critical or irreversible actions require explicit confirmation from a person.
Hybrid Search
Hybrid search is an information-retrieval method that combines the results of lexical keyword search and vector semantic search into a single ranked list.
Hyper-Personalization
Hyper-personalization is personalization in which the decision is made at the level of an individual user and an individual request, drawing on a long-term profile, session context and real-time events rather than on segment membership.
I
16 terms
Ideal Customer Profile (ICP)
An ICP is the description of the customer type — company or shopper — that gets the most value from a product and produces the best business outcomes: high lifetime value, low churn and fast decisions.
Idempotency
Idempotency is the property of an API operation that returns the same result on repeated calls with the same parameters; it makes retries safe and prevents duplicated data after network errors.
Identifier for Advertisers (IDFA)
IDFA is the advertising identifier of an iOS device, available since iOS 14.5 only with explicit user consent through ATT; the loss of broad access to it reshaped mobile attribution and targeting.
Identity Resolution
Identity resolution is the process of merging the different identifiers and data sources belonging to one real person into a single unified customer profile inside a CDP.
iframe Widget
An iframe is an HTML element for embedding third-party content in an isolated context; in e-commerce it suits payment widgets and forms, but it is a dated choice for personalization recommendation blocks.
Implicit Feedback
Implicit feedback is indirect preference data extracted from a shopper's behaviour — clicks, views, dwell time — with no explicit rating involved.
Impulse Purchase
An impulse purchase is an unplanned acquisition decided quickly under the influence of the display context rather than a prior search.
In-App Message
An in-app message is a communication displayed during an active mobile app session, used for personalized offers, onboarding and announcements.
In-App Personalization
In-app personalization is the adaptation of a mobile app's content, interface and communications to an individual profile based on behaviour, preferences and context.
In-App Purchase (IAP)
An in-app purchase is a built-in monetisation mechanism inside a mobile app that runs through Apple's or Google's official payment systems and covers digital goods, subscriptions and consumable units.
In-App Survey
An in-app survey is a contextual mini-survey inside a mobile app that collects feedback or zero-party data at the point of maximum relevance, without sending the person to an external resource.
Indexation
Indexation is the process of a search bot scanning pages and adding them to the search index, after which they become eligible to appear in results for relevant queries.
Instant Checkout
Instant Checkout is an in-chat purchase feature in ChatGPT, built on the ACP protocol, that lets a shopper order from a merchant without going to the merchant's site.
Intent Recognition
Intent recognition is the NLP task of classifying a user query into goal categories so that a dialogue system can select the correct response scenario.
Internal Linking
Internal linking is the system of hyperlinks between pages of one site, distributing link equity, enabling efficient crawling and making the semantic relationships of content legible to search engines.
Item2Vec
Item2Vec is a method for learning vector representations (embeddings) of products from session data, where the closeness of two vectors corresponds to the probability that the items are viewed or bought together.
J
4 terms
JavaScript Snippet
A JavaScript snippet is a fragment of JS embedded in an HTML page to connect an external service; in e-commerce it is the frontend route to integrating a personalization platform with no backend work.
Jobs to Be Done (JTBD)
Jobs to Be Done (JTBD) is an approach in which the reason for a purchase is described as a “job” the customer wants to get done in a particular situation, and the product as what they “hire” to do that job.
JSON
JSON is a text serialisation format built on key-value pairs and arrays; it is the de facto standard for REST APIs and for data exchange between e-commerce systems.
JSON Schema
JSON Schema is a language for describing the structure and constraints of JSON documents; it is used to validate product feeds, events and the API contracts of a personalization platform.
K
5 terms
Keyword
A keyword is a word or phrase entered into a search engine that serves as the basis for optimising pages in order to attract organic traffic.
Keyword Cannibalization
Keyword cannibalization is a situation in which several pages of one site compete for the same keywords, fragmenting link equity and lowering the ranking performance of each of them.
Keyword Cluster / Semantic Core
A semantic core is a structured database of search queries grouped by thematic cluster and by intent, used as the basis for site architecture and content planning in SEO. It is the Russian-market name for what Western SEO calls keyword research plus clustering.
Know Your Agent (KYA)
Know Your Agent (KYA) is the verification of an AI agent before it is given access to a website or a payment: who it is, which operator runs it, what authority the user has granted and whether the agent stays within it.
Knowledge Graph
A knowledge graph is a structured knowledge base that represents objects and the relationships between them as a directed graph, used to enrich recommender models with semantic information.
L
14 terms
Large Language Model (LLM)
An LLM (large language model) is a language model with billions of parameters, trained on large text corpora by autoregressive token prediction, capable of understanding language, following instructions and generating text.
Large Language Model Optimization
Large language model optimization is the set of measures that raise the probability of a brand being mentioned and its content cited by large language models, through presence in authoritative sources, structured data and llms.txt.
Latency
Latency is how long a system takes to answer a request; for personalization platform APIs it is measured in percentiles (p50/p95/p99) and is a core SLA parameter with a direct effect on a store's conversion rate.
Lazy Loading
Lazy loading is the pattern of deferring resources — images, scripts, widgets — until the page is scrolled towards them, reducing initial load time and the volume of data transferred.
Lead Nurturing
Lead nurturing is an automated series of relevant touches that gradually moves a prospect through the funnel stages from awareness to conversion.
Lifecycle Marketing
Lifecycle marketing adapts communications and offers to the customer's current lifecycle stage, from onboarding a newcomer to reactivating a dormant buyer.
Link Juice / Link Equity
Link juice is the total authority passed to a domain or page through inbound hyperlinks; one of the key factors determining a site's ability to rank for competitive queries.
LLM Evaluation (Evals)
LLM evaluation (evals) is an offline run of a reference set of requests through an LLM system, scoring answer-quality metrics to weed out weak prompt, model or retrieval versions before release.
llms.txt
llms.txt is an emerging standard for a root-level text file that gives language models and AI agents a structured description of a site so they can understand it without a full crawl.
LocalStorage / SessionStorage
LocalStorage and SessionStorage are browser data stores: LocalStorage persists between sessions, SessionStorage lives only inside one tab; both are used as a client-side cache for user data that never travels to the server.
Long-Tail Keywords
Long-tail keywords are a group of low-volume, specific search queries of three or more words with light competition, which together generate substantial targeted traffic with a high propensity to convert.
Lookalike Audience
A lookalike audience is generated algorithmically: the system finds users similar in behaviour and profile to a defined seed segment of best customers, extending campaign reach.
Loyalty Mechanics
Loyalty mechanics are customer-retention tools — accumulated points, status tiers, personalized rewards — that encourage repeat purchases and emotional attachment to a brand.
Loyalty Program
A loyalty program is a mechanic that rewards customers for repeat purchases and for identifying themselves at payment: points, status tiers, cashback or a paid subscription with privileges.
M
22 terms
Machine Customer
A machine customer is a non-human market participant — a device, program or AI agent — that selects and pays for goods or services itself, either under rules a person has set or autonomously.
Machine Learning
Machine learning is a class of methods where a model derives patterns from historical data and applies them to new observations, instead of following rules written in advance.
Machine Payments Protocol (MPP)
MPP (Machine Payments Protocol) is an open protocol from Stripe and Tempo for machine payments for services: the server sets the price with an HTTP 402 response, and the client pays with a method of its choice and receives the resource with a receipt.
Marketing Automation Platform
A MAP (Marketing Automation Platform) automates marketing communications: building triggered scenarios, orchestrating channels, managing campaigns, scoring contacts and reporting on campaign performance.
Matrix Factorization
Matrix factorization is a class of algorithms that decompose the user-item interaction matrix into the product of two lower-dimensional matrices, making it possible to model hidden preferences and surface relevant products.
MCP Apps
MCP Apps is an extension of the MCP protocol that lets MCP servers return interactive UI which the AI agent displays to the user in conversation, rather than text and data alone.
Mean Average Precision
Mean average precision (MAP) is a ranking quality metric computed as the mean of average precision values across all users or queries; it accounts for both precision and the position of relevant items in the list.
Merchandising Rules
Merchandising rules are manual business rules that adjust algorithmic sorting and recommendations by promoting, excluding or pinning specific items.
Meta Description
A meta description is the HTML meta tag describing a page for the search results; it does not influence ranking directly, but it determines how attractive the snippet is and therefore organic click-through.
Micro-conversion
A micro-conversion is an intermediate target action on a user's path to purchase — a click, an add to cart, a page view — used as a proxy metric.
Micro-moments
Micro-moments are short, intent-rich moments of interaction with a device, during which personalized content has the greatest influence on the shopper's decision.
Micro-Segmentation
Micro-segmentation builds maximally narrow user groups from several attributes at once, achieving high personalization precision because behaviour inside each segment varies so little.
Minimum Detectable Effect (MDE)
MDE is the smallest metric lift a planned A/B test can register at the chosen significance level (alpha) and power (1 minus beta).
Mobile SDK
A mobile SDK is a set of libraries and APIs embedded in a mobile app to integrate a personalization platform's functions — behavioural tracking, recommendations and managed communications.
Model Context Protocol (MCP)
MCP is Anthropic's open standard for connecting AI models to external tools and data uniformly over JSON-RPC; the foundation layer for agentic applications, commerce included.
Model Inference
Inference is the application of a trained model to new input data to obtain predictions without changing its parameters; it runs in production, in real time.
Model Training
Model training is the iterative optimisation of a model's parameters on a training sample by minimising a target loss function through gradient descent or another optimisation algorithm.
Monthly Unique Visitors (MUV)
Monthly unique visitors is the count of unique visitors to a site in a month, used as the standard billing metric of personalization and web analytics platforms.
Monthly, Weekly and Daily Active Users
MAU, WAU and DAU are counts of unique users active in an app or on a site over a month, a week and a day respectively; the baseline health metrics of a product.
Multi-Armed Bandit (MAB)
Multi-armed bandit solves the exploration/exploitation trade-off: it balances trying out new variations against exploiting the best one found so far. The clearer the lead of one variation, the more traffic it receives automatically.
Multimodal AI
Multimodal AI is a class of models that process and align data from several modalities — text, images, audio — inside one shared representation in order to perform cross-modal tasks.
Multivariate Testing (MVT)
Multivariate testing (MVT) is an experiment in which several page elements change at once across all their combinations, in order to identify the optimal pairing and the interaction effects between factors.
N
8 terms
Natural Language Processing
NLP (natural language processing) is the branch of AI and linguistics that builds methods for automatically analysing, understanding and generating natural-language text.
Net Promoter Score (NPS)
NPS is a metric that measures customer loyalty as the difference between the share of promoters (scores of 9-10) and detractors (0-6) in answer to a question about willingness to recommend the brand.
Neural Network
A neural network is a parametric machine learning model built from successive layers of linear transformations and non-linear activations, trained by backpropagation of the prediction error.
NLWeb
NLWeb is an open Microsoft project that adds a natural-language conversational interface to a website based on its Schema.org markup and RSS, and exposes the same interface to AI agents as an MCP server.
Normalized Discounted Cumulative Gain
NDCG (normalized discounted cumulative gain) is a ranking quality metric that sums the relevance of positions with a logarithmic discount by position and normalises the result against the ideal ranking.
North Star Metric (NSM)
A North Star Metric is the single end-to-end metric that expresses a product's core value to its users and serves as the primary reference point for prioritisation across the team.
Notification Permission
Notification permission is the explicit consent without which an app cannot deliver push; on iOS it is requested once through a system dialogue, which makes the first-request strategy decisive.
Novelty Effect
The novelty effect is a temporary A/B test artefact: inflated metrics for a variation in its first days, caused by heightened attention to the change rather than by its real value.
O
9 terms
OAuth 2.0
OAuth 2.0 is a delegated authorisation standard that gives one service limited access to another's resources through tokens rather than credentials; it is the standard way to connect e-commerce integrations securely.
Omnichannel
Omnichannel is a customer strategy where all interaction channels are synchronised and form one personalized experience over a shared customer history.
Onboarding Completion Rate
Onboarding completion rate is the share of users who complete an app's entire onboarding flow; a low figure signals excessive complexity or a value proposition that was never made clear.
One-to-One Personalization
One-to-one personalization is the creation of a unique experience for each individual visitor based on their personal profile, as opposed to segment-level personalization.
Opt-in / Opt-out
Opt-in is a user's explicit consent to marketing communications; opt-out is their refusal to receive them. The balance between the two indicates the quality of personalization and frequency management.
Organic Traffic
Organic traffic is the visits arriving from unpaid search results; it is the key KPI of an SEO strategy and the most scalable long-term source of user acquisition.
Over-Optimization
Over-optimization is the excessive application of SEO techniques — keyword repetition, unnatural links — that search engines interpret as manipulation and answer with algorithmic or manual penalties.
Over-the-Air Update (OTA)
An over-the-air update delivers changes to a mobile app directly to devices without store review; within platform rules it is allowed for fixes and content, not for changes to core functionality.
Overfitting
Overfitting is the phenomenon where a model is optimised so tightly to its training sample that it loses the ability to generalise: the error on training data is low while the error on new data is significantly higher.
P
31 terms
P-value
A p-value is the probability of obtaining the observed data, or more extreme data, when the null hypothesis is true; it is the basis for deciding whether to reject that hypothesis in an A/B test.
Pages Per Session
Pages per session is the average number of pages viewed in a single session; a measure of user engagement and of the quality of a site's internal navigation.
Pagination
Pagination is the mechanism that splits a long list of products or articles into sequential pages or loaded chunks, and in doing so decides both how accessible products are to search crawlers and how deep into the catalogue a shopper browses.
Payment Mandate
A payment mandate is cryptographically signed proof of user consent in the AP2 protocol; the Intent Mandate sets intention and limits, the Cart Mandate fixes the specific basket and price.
Paywall
A paywall is the barrier in a mobile app or on a website behind which paid content or functionality sits; it is the pivotal conversion point of the monetisation funnel.
Peeking
Peeking is the early analysis of an unfinished A/B test with a decision to stop based on interim data, which systematically inflates the share of false positive results.
Performance Marketing
Performance marketing is a model oriented to measurable results: investment is tied to specific user actions (a sale, a lead, revenue), effectiveness is judged by payback metrics such as ROAS, CPA and ROMI, and improvement comes from testing.
Personalization
Personalization is the automatic adaptation of content, product selection and offers to an individual user, based on their behavioural profile, history and the context of the current session.
Personalization Engine
A personalization engine is a platform that collects behavioural data and decides in real time on the optimal content, recommendations or offers for each individual user.
Personalized Banner
A personalized banner is a dynamic banner whose content is selected individually for each shopper based on their behavioural profile.
Personalized Recommendations
Personalized recommendations are product selections generated individually for each user from their behavioural profile and interaction history. What separates them from popular or similar items is the input: the individual profile, not an aggregate of the store or the attributes of the current page.
Pixel / Tracking Pixel
A tracking pixel is an invisible element embedded in a page or an email that reports an event — an open, a visit — to an analytics server as it loads; it is the historical predecessor of the modern event API.
PLP Personalization
PLP personalization is the algorithmic re-ordering of products on a category page against the preferences of an individual shopper, derived from their behavioural profile. It changes the sort order of an existing category, never its contents — the same products, ranked differently for each visitor.
Points System
A points system is a loyalty mechanic in which the user earns virtual units for target actions and later exchanges them for discounts or rewards.
Popup
A popup is a modal overlay that covers page content and is used for personalized offers, lead capture or feedback collection.
Precision and Recall
Precision is the share of correct positive predictions among all positive predictions a model makes; recall is the share of correct positive predictions among all genuinely positive examples.
Predictive Segmentation
Predictive segmentation groups users by machine learning predictions of their future behaviour — churn risk, purchase probability — allowing pre-emptive action instead of a reaction after the fact.
Product Catalog
A product catalog is the structured database of all products with their attributes (category, brand, price, specifications), the foundation recommender algorithms operate on.
Product Detail Page (PDP)
A product detail page (PDP) is the product card page carrying the description, the price and the purchase button — the key conversion point in the e-commerce funnel.
Product Detail Page Conversion Rate (PDP CR)
PDP CR is the share of product detail page visits that end in an add to cart or a purchase; an indicator of how effective the product page is.
Product Feed
A product feed is a structured file carrying the attributes of a product catalogue, used to integrate with recommendation and advertising systems.
Product Information Management (PIM)
PIM (product information management) is a system for managing product information: a single source of attributes, descriptions, media and translations from which data is exported to every sales channel.
Product Listing Page (PLP)
A product listing page (PLP) is the listing of products in a category or search result — the key surface for personalized sorting and merchandising.
Product Matrix
A product matrix is a structured list of the assortment with a hierarchy of category, subcategory, product group and SKU, and an assigned role for each item. It is the data source behind the catalogue and the product feed.
Progress Bar
A progress bar is a gamification element that displays how far a goal has been reached, using the goal gradient effect to motivate the user toward the final action.
Promo Code
A promo code is a code combination that gives the shopper a discount or another benefit at checkout; it is used to stimulate conversion, attract new customers and reactivate existing ones.
Prompt
A prompt is the input text instruction passed to a language model; it defines the task, the context and the expected answer format, and directly shapes the quality of what the model generates.
Prompt Injection
Prompt injection is an attack in which instructions are embedded in text processed by a language model so that the model follows them instead of the application owner's rules.
Purchase Frequency
Purchase frequency is the average number of purchases per active customer over a defined period; it is used in RFM analysis and in the LTV calculation.
Push Notification
A push notification is a short automated message delivered to a device screen through a mobile app or a browser, without the person being in the interface.
Push Token / Device Token
A push token is the unique identifier of an app installation on a specific device, issued by the OS through FCM or APNs and used as the delivery address for push notifications.
Q
2 terms
Query Fan-Out
Query fan-out is an AI search technique in which the original query is broken into a set of sub-queries by subtopic, a search is run for each, and the model builds the answer from the combined results.
Quiz Mechanic
A quiz mechanic is an interactive survey that produces personal recommendations while collecting explicit zero-party data about the shopper's preferences.
R
32 terms
Randomization
Randomization is the random assignment of participants to the groups of an A/B test, removing systematic bias so that the variations can be compared fairly.
Rate Limiting / Rate Limit
Rate limiting caps the number of API requests over a period and answers HTTP 429 when the cap is exceeded; integrations need queueing and caching to stay inside their quota.
Reactivation
Reactivation is the set of marketing measures for bringing users who have lost activity back into interaction, through personalized messages in the channels where they were most engaged.
Real-Time Personalization
Real-time personalization is the adaptation of recommendations and content to a user's actions in the current session, with a delay of under 100–200 ms.
Reasoning Model
A reasoning model is an LLM that spends extra computation on a chain of reasoning before its final answer, improving accuracy on multi-step tasks at the cost of latency and price.
Recently Viewed
Recently viewed is a personal widget carrying a shopper's browsing history, helping them return faster to the items that interested them.
Recommendation Diversity
Recommendation diversity is a property of a recommendation list that reflects how unlike each other the suggested products are, usually measured as the mean pairwise distance between their embeddings.
Recommendation Strategy
A recommendation strategy is the algorithm or rule that selects the products for a widget, defining whether the logic is personal, popularity-driven, associative or manual.
Recommendation Widget
A recommendation widget is the UI component on a site page that displays algorithmically selected products as a carousel, a grid or a list. It is the delivery surface, not the algorithm — the same widget can be driven by any recommendation strategy.
Recommender Systems
A recommender system is an algorithmic system that selects the products most relevant to a user from the catalogue, based on their behaviour, product attributes and session context. It is not a single model but a pipeline: candidate generation, ranking and merchandising rules are separate stages, each of which can be changed independently.
Redirect
A redirect is a server response with a 3xx code that forwards a request from one URL to another. 301 and 308 mean a permanent move that passes signals to the target page; 302 and 307 mean temporary forwarding with the old URL kept in the index.
Regularization
Regularization is the addition of a penalty term to the loss function for large model parameter values, constraining the complexity of the solution and reducing overfitting.
Reinforcement Learning
Reinforcement learning is a machine learning paradigm in which an agent learns an optimal action policy by interacting with an environment and maximising a cumulative reward function.
Remarketing
Remarketing is the strategy of re-engaging an audience that has previously interacted with the brand, predominantly through owned channels and personalization, in order to bring the user back to an unfinished action.
Remote Config
Remote config is a cloud service for managing an app's behavioural parameters without publishing an update; it changes functionality and content instantly for all users or a chosen segment.
Repeat Purchase Rate
Repeat purchase rate is the share of customers with at least two purchases in the analysed period; an indicator of loyalty and of growth available without acquisition spend.
Reranking
Reranking is the stage of a recommendation or search pipeline where pre-selected candidates are reordered by a more precise model that accounts for user context, business rules and additional features.
Retail Media
Retail media is an advertising model on a retailer's own digital properties — site, app and storefront — where brands and suppliers pay for placements and the retailer monetises its traffic and its first-party purchase data.
Retargeting
Retargeting is the re-engagement of people who have previously interacted with a brand, through targeted advertising or personalized content aimed at returning them to an unfinished action — a purchase, a sign-up or an order.
Retention Curve
A retention curve is a time series of the share of a cohort that remains active; it shows the speed of churn and whether a loyal core of the audience exists.
Retention Marketing
Retention marketing is the set of activities for keeping existing customers — personalization, triggered communications, engagement programmes — aimed at raising lifetime value and reducing churn.
Retention Rate
Retention rate is the share of users or customers who returned and repeated a target action inside a defined time window.
Retrieval-Augmented Generation (RAG)
RAG (retrieval-augmented generation) is an AI architecture in which a language model is extended with a retrieval step: relevant documents are pulled from an external store and passed into the LLM context so the answer is built from real data.
Return on Ad Spend (ROAS)
ROAS is the ratio of revenue generated by advertising to the cost of that advertising. It shows the revenue efficiency of an advertising investment.
Return on Investment
ROI is the ratio of the net profit from an investment to its cost; the key metric for assessing whether a technology pays for itself and for justifying the budget.
Revenue Attribution
Revenue attribution is the method of assigning revenue to the initiatives or channels that led to a purchase; the basis for computing the return of individual features, widgets and campaigns.
Revenue Per Mille / Revenue Per Thousand Visitors
RPM is the revenue attributable to 1,000 visitors or impressions; a normalised form of RPV used to compare channels and segments with substantially different traffic volumes.
Revenue Per Visitor (RPV)
Revenue per visitor is total revenue divided by unique visitors over a period; a summary metric of e-commerce and experiment performance.
Reward
A reward is a material or non-material incentive for completing a target action, and it is the closing element of the gamification engagement loop.
RFM Analysis
RFM analysis is a method of segmenting a customer base by the recency, frequency and monetary value of purchases, used to personalize marketing and manage the customer lifecycle.
Rich Push Notification
Rich push is a push notification with extended media — an image, action buttons, video — implemented through a notification service extension on iOS and native capabilities on Android.
robots.txt
robots.txt is the standard text file at the root of a site holding Allow and Disallow directives for search and other bots, governing their access to sections of the site during crawling.
S
35 terms
Sample Ratio Mismatch (SRM)
Sample Ratio Mismatch is a statistically significant deviation of the observed distribution of users across test variations from the planned one, indicating that randomization has been compromised.
Sample Size
Sample size is the minimum number of observations per A/B test variation that makes it possible to detect the target effect at the chosen statistical power.
Schema Markup
Schema markup is a set of structured data in the Schema.org vocabulary added to a page's HTML so that search engines and language models understand the meaning of its content unambiguously.
Screen Flow / Navigation Flow
Screen flow is the analysis of screen sequences in a mobile app from behavioural data; it is used to find drop-off points, improve navigation and build conversion funnels.
Search Engine Optimization
SEO is the systematic process of increasing a site's visibility in organic search results through technical optimisation, relevant content and the growth of external link authority.
Search Engine Results Page
A SERP is the search results page a search engine builds in response to a user query, containing organic links, advertising and various types of enriched block.
Search Intent
Search intent is the goal behind a search query — informational, navigational, transactional or commercial investigation — and matching it is the key condition for a page to rank at the top of the results.
Seasonality
Seasonality is the regular, repeating variation in demand across a yearly, weekly, daily or event-driven cycle; it sits on top of the underlying trend and distorts any before-and-after comparison.
Semantic Search
Semantic search is an information-retrieval method that encodes queries and documents into a shared vector space and returns results by similarity in meaning rather than by lexical match.
Sequential Testing
Sequential testing is a statistical approach that permits interim analysis of an A/B test and early stopping without inflating the type I error rate.
Server-side Integration
Server-side integration is a method of connecting a personalization platform where the API calls happen on the store's server and the result is embedded in the HTML before it is sent to the browser.
Server-Side Rendering
Server-side rendering is the generation of HTML on the server before it is sent to the browser; it gives correct SEO indexation and a fast first paint, including for personalized content.
Service Level Agreement
An SLA is a supplier's formal commitment on service availability and performance; for personalization platforms the key SLA metrics are uptime (no lower than 99.9%) and the p99 latency of API endpoints.
Service Worker
A service worker is a browser JavaScript script running on a background thread that intercepts the page's network requests; it provides offline mode, resource caching and push notifications in progressive web apps.
Session
A session is a period of uninterrupted activity by one user on a site, bounded by an inactivity timeout (usually 30 minutes); it is the standard unit of behavioural analysis in web analytics.
Session Length / Session Duration
Session length is the duration of one continuous session with a mobile app; it is interpreted in the context of the actions taken during it, never in isolation.
Session Recording
Session recording captures one user's interaction with a site as a replayable video; it is a qualitative UX research tool for finding points of friction.
Session-Based Recommendations
Session-based recommendations are a class of algorithms that predict a shopper's next interaction from the sequence of actions in the current session, without using a long-term profile.
Share of Model
Share of Model is the share of AI model answers to category prompts that mention a brand, relative to competitors; the equivalent of share of voice for generative assistants.
Shared Payment Token (SPT)
A Shared Payment Token is Stripe's single-use payment token in the ACP protocol, which an AI agent uses to initiate payment without access to the shopper's real payment credentials.
Silent Push / Background Notification
A silent push is a push notification carrying a background-execution flag: it is never shown to the person, but it wakes the app so it can run a background data sync.
Similar Items
Similar items are recommendations of alternative products close to the one being viewed by attribute, price or audience behaviour.
Single Customer View (SCV)
A single customer view (SCV) is one record per customer that brings together every available data source, so their behaviour, preferences and interaction history can be understood in full.
Single Page Application
A single page application is a web application that updates page content dynamically without reloading it, and that requires the personalization SDK to be initialised explicitly on every route change.
Site Search
Site search is the search engine built into an online store that returns relevant products, categories and catalogue pages in response to a shopper's text query.
Snippet
A snippet is the block describing a page in search results, made up of a title, a description, the address and additional elements, some of which are generated from structured data markup.
Social Proof
Social proof is a set of UI elements that display the activity and behaviour of other shoppers in order to reduce uncertainty and speed up the purchase decision.
Software Development Kit (SDK)
An SDK is a set of libraries and tools that lets developers embed an external platform's functionality into their own application; mobile personalization ships separate SDKs for iOS and Android.
Spin the Wheel / Lucky Wheel
Spin the Wheel is a gamification mechanic built around a wheel of random prizes that raises engagement and conversion at the lead-capture and activation stages through a game-like experience.
Statistical Power
Statistical power (1 − beta) is the probability that a test detects a real effect when one exists; the industry standard is no lower than 80%.
Statistical Significance
Statistical significance is the degree of confidence that the difference between A/B test variations is real rather than the product of random fluctuation; the usual requirement is p < 0.05.
Stickiness (DAU/MAU Ratio)
Stickiness is the ratio of DAU to MAU, reflecting how regularly users return to a product; a key engagement indicator for both mobile and web applications.
Stock Keeping Unit
An SKU is the unique identifier of one product variant, defined by colour, size or configuration; it is the primary unit of accounting in a catalogue and in recommendation systems.
Streak
A streak is a gamification mechanic of consecutive actions without a miss, using loss aversion to drive regular returns to an app.
Supervised Learning
Supervised learning is a machine learning paradigm in which a model is optimised on labelled data, where every input example carries a correct target answer — a class or a number.
T
20 terms
Tag Manager
A tag manager is a platform for centrally managing third-party scripts and tags on a site without direct code changes, giving marketers independence from the development queue.
Test Duration
Test duration is the minimum period of an A/B experiment, set by the computed sample size and by the need to cover whole business cycles so that temporal fluctuation cancels out.
Test Variation
A test variation (treatment) is the version carrying the change under test in an A/B experiment; it is compared against the control to measure the incremental effect of one specific hypothesis.
Third-party Cookie
A third-party cookie is a cookie belonging to an outside domain and used to track users across sites; browsers are blocking it progressively, which makes first-party data the foundation for personalization.
Time on Site / Session Duration
Session duration is the average time a user spends on a site during one session; a supporting engagement indicator that has to be interpreted alongside conversion metrics.
Time to First Action (TTFA)
Time to first action is the interval between an app install and the person's first key action, such as a purchase or an add to cart; a short TTFA correlates with better long-term retention.
Time to Live
Time to live is the parameter that defines how long a record in a cache or in DNS stays valid; it governs the trade-off between performance, meaning fewer calls to the source, and data freshness.
Timeout
A timeout is the limit on how long a system waits for an external service to respond; setting it properly protects an online shop from hanging when an API is unavailable, and it belongs together with graceful degradation.
Title Tag
The title tag is the HTML tag that defines a page's title for search engines and browsers; it is a key on-page factor affecting both click-through rate in the results and rankings for target queries.
Tokenization
Tokenization is the splitting of text into the smallest processing units — tokens — for input to a language model; the standard methods, BPE and SentencePiece, operate at the sub-word level.
Transformer
A transformer is a neural network architecture with a multi-head self-attention mechanism that models the dependencies between all tokens in a sequence in parallel.
Trending Products
Trending products are recommendations ranked dynamically by the rate of growth in demand over a short time window.
Trigger
A trigger is a behavioural event or condition that automatically fires a personalized action — displaying an element, sending a communication or changing content.
Trigger Engine
A trigger engine is a real-time system that watches behavioural events and automatically launches defined marketing actions when a rule's conditions are satisfied.
Trigger Marketing
Trigger marketing is automated communication fired in response to specific user actions or events, giving the message maximum relevance to the moment.
Trusted Agent Protocol (TAP)
The Trusted Agent Protocol (TAP) is Visa's protocol for cryptographically identifying AI agents: the agent signs its requests to a store, and the store verifies the signature against Visa's keys and receives data on intent, the shopper and payment.
Two-Tower Model
A two-tower model is a neural recommender architecture with two separate encoders, one for the user and one for the item, trained to maximise the dot product of pairs consisting of a user and a relevant item.
Type I Error (False Positive)
A Type I error is a false-positive conclusion: a variation is declared better than the control when no real difference exists; the probability of that error is controlled by the threshold α.
Type II Error (False Negative)
A Type II error is a false-negative conclusion: a genuine improvement in the variation goes undetected because there is not enough data; the probability of that error equals one minus the power of the test.
Typo Tolerance
Typo tolerance is the property of a search system that returns relevant results when the query contains misspellings, missing or extra characters, swapped letters, or was typed in the wrong keyboard layout.
U
8 terms
Unit Economics
Unit economics is the analysis of profitability at the level of the smallest business unit — a customer or an order; it tests whether the business model holds up as it scales.
Universal Commerce Protocol
UCP is Google's open commerce protocol. It publishes a store's capability profile — search, checkout, returns — so an AI agent understands the available operations before a transaction begins.
Universal Links / Android App Links
Universal Links (iOS) and App Links (Android) are the mechanism for opening a mobile app from a standard HTTPS link, with no URI scheme and with a web fallback when the app is not installed.
Unsupervised Learning
Unsupervised learning is a type of machine learning in which a model discovers structures, clusters or representations in data without using labelled target values.
Uplift
Uplift is the incremental relative gain of a metric in the treatment variation compared with the control group — the main indicator of the value of a change in an A/B test.
Upsell
Unlike cross-sell, upsell offers a more expensive version of the same or a comparable product rather than an addition to it. For example: a shopper is looking at a $900 laptop, and the upsell offers a comparable model at $1,150 with a better screen.
Urgency and Scarcity
Urgency and scarcity are conversion tactics that display a limit on the time an offer is valid or on the quantity available, triggering the fear of missing out.
UTM Parameters
UTM parameters are a set of URL values that pass information about the source of a visit — channel, campaign, ad and keyword — to analytics platforms. They are the foundation of attribution for paid and email traffic.
V
4 terms
Vector Database
A vector database is specialised storage optimised for indexing and nearest-neighbour search in high-dimensional vector spaces at sub-second latency.
Verifiable Intent
Verifiable Intent is an open specification from Mastercard and Google: a cryptographic chain of signatures proving that an AI agent acted within the authority the shopper delegated to it.
Visual Search
Visual search is a way to find a product from a photo or screenshot: a computer vision model converts the image into a vector and finds catalogue items that look similar.
Voice Commerce
Voice commerce is buying and managing orders through a voice assistant: choosing a product, reordering, confirming and tracking delivery with no screen, or with only minimal use of one.
W
6 terms
Web Bot Auth
Web Bot Auth is an IETF draft standard under which an automated client (a crawler or an AI agent) cryptographically signs its HTTP requests and the website verifies the signature against a public key from the bot operator's directory.
Web Components
Web Components are browser standards — custom elements, shadow DOM and HTML templates — for building encapsulated reusable components; they are used to distribute personalization widgets independently of any framework.
Web MCP
Web MCP is the use of the MCP protocol on the web to connect AI agents to sites and web search in a standardised way, so an agent reads content and acts programmatically rather than by parsing pages.
Webhook
A webhook is a push-notification mechanism over HTTP: when an event occurs, the source service immediately sends a request to a URL the recipient has registered, with no periodic polling required.
Win-back Campaign
A win-back campaign is a marketing scenario for returning dormant customers through a series of personalized messages with a progressively stronger offer.
Wishlist
A wishlist is the feature for saving products for later consideration; it is a strong behavioural signal for personalized recommendations and triggered scenarios.
X
3 terms
x402
x402 is an open protocol in which a server requests payment with an HTTP 402 response and a client or AI agent pays — typically in stablecoins — then retries the request with a signed payment in an HTTP header.
XML Sitemap
An XML sitemap is a file listing a site's URLs and their metadata, intended to speed up the discovery and indexation of pages by search engines.
XYZ Analysis
XYZ analysis classifies products by demand stability using the coefficient of variation of sales across a series of periods: X is stable demand, Y moderately variable, Z irregular and hard to forecast.
Z
3 terms
Zero-click Search
Zero-click search means search sessions that end without a visit to an external site, because the answer is displayed in full inside the interface of the search engine or the assistant.
Zero-Party Data
Zero-party data is information a shopper voluntarily and knowingly gives a brand — through quizzes, surveys or preference settings — and it is used for immediate personalization when no behavioural history has accumulated yet.
Zero-Results Rate
The zero-results rate is a site-search metric: the share of search queries for which the system found no results, out of all queries over a period.
Nothing matched your query. Clear the filter to see every term.