A category as a small business
Category management grew out of retail and rests on a simple idea: a category is not a line in a
report but a separate business unit with its own P&L, its own strategy and its own owner. The
category manager is accountable not for a specific product or a specific supplier, but for the
economics of the entire category.
In practice that means five areas of responsibility:
| Area | The question it answers | The main artefact |
|---|---|---|
| Category role | Why the assortment needs it | The category strategy |
| Assortment | Which products and how many | The product matrix |
| Prices | How the price tiers are distributed | The price ladder, positioning rules |
| Promotions | When and with what demand is stimulated | The promo calendar |
| The shelf | In what order the shopper sees it | The planogram — online, the listing output rules |
The last row is where category management meets personalization instruments directly. Everything a
manager does offline with a planogram becomes, online, the order of cards on the
listing page and the rules that build it.
Category roles
Different categories solve different problems, and they cannot be measured with one ruler.
| Role | The job | The main KPI | How it is laid out on the site |
|---|---|---|---|
| Traffic-driving | Bring the shopper in, set price perception | Traffic, market share, order count | Prominent entry points, strong price offers at the top |
| Margin | Earn | Gross margin in absolute terms | Priority for margin items inside relevant blocks |
| Image | Support positioning and expertise | Share of premium brands, assortment perception | Brand showcases, editorial selections |
| Seasonal | Deliver a peak in a defined period | In-season revenue, clearing residual stock | Seasonal blocks, rotation by calendar |
| Companion | Grow on add-on sales | Share of basket, share of orders with two or more categories | Cross-sell blocks in the cart and on the product page |
A mistake that recurs regularly: a traffic-driving category starts being judged on margin and gets
improved by raising prices. The traffic leaves, and with it the companion sales that justified the
category in the first place.
Category KPIs
A single metric always lies. The minimum set worth reading together:
Category revenue - trend year on year and versus plan
Gross margin (absolute and %) - what the category actually earns
Stock turnover - how much money is frozen in inventory
Share of GMV - the category's weight in turnover
Share of SKUs with no sales - the quality of the matrix
Listing conversion - how many category visitors reached the cart
Share of zero-result searches - gaps in assortment and attributes
Online adds two things to the classic set that do not exist offline. The first is listing
conversion: in a shop you cannot count how many people walked past the shelf and took nothing, while
on a site it is a basic metric. The second is in-category search queries with zero results: a direct
list of what the shopper expected and did not find. Both are material for matrix decisions, not only
for a report.
Before changing a category’s assortment, read the internal search report for the quarter. Queries
with zero results, and queries with high frequency and low conversion, give a more accurate picture
of the gaps than a sales analysis — sales only show what was already on the shelf.
How this translates online
A category manager’s work in an online store breaks down into four instruments.
Listing sorting. The default order is the category’s main decision. Sort by popularity in
practice means by what already sold well, and it cements the past: new products get no impressions
and never become popular. This is where PLP personalization helps
— it adjusts the order to the individual user rather than to an average one.
Merchandising rules. Pinned slots, exclusions, limits by brand and availability. This is the
digital planogram — covered in a separate article on
merchandising rules.
Cross-category recommendations. Blocks linking categories together: accessories for electronics,
care products for footwear, companion items for a large purchase. Here the category manager defines
the logic of the links and the algorithm picks the specific items.
Search and filter analytics. Which filters get used, which stand empty, which in-category
queries return nothing. This is an audit of both the matrix and the quality of product attributes.
The conflict of manual rules versus the algorithm
The most common argument in teams rolling out listing personalization: the category manager wants to
control the layout, the algorithm wants to optimise it for the user. Both sides are right, and the
resolution is a split of zones rather than a fight over who outranks whom.
| What the rules set | What the algorithm optimises |
|---|---|
| Hard bans: out of stock, discontinued, not for this region | The order among the permitted products |
| Pinned slots: a new collection, a promotion, a priority supplier | Everything below the pinned slots |
| Diversity limits: no more than N items of one brand in a row | The choice of specific products inside those quotas |
| Priority for margin items, all else being equal | The judgement of which products are equal for this user |
| Seasonal and calendar windows | Ranking inside the window |
The working formula: rules set the frame, the algorithm optimises inside it. Rules describe
business constraints the algorithm cannot know: supplier agreements, brand strategy, plans for
clearing a collection. The algorithm answers the question a human cannot resolve by hand — which of
a thousand permitted products is most relevant to this particular visitor.
The sign that the balance has broken: the rules pin so many slots that personalization only affects
the tail of the output. If the first 12 cards are always fixed manually and exactly 12 fit on the
first screen, the algorithm is switched off in practice.
The balance is checked by measurement rather than discussion: an A/B test
with different shares of pinned slots gives a direct answer on how much manual control the category
can carry without losing conversion.
A checklist for running a category online
- The category role is written down, and the KPIs match the role.
- The matrix is reviewed regularly, and there is a procedure for retiring products with no sales.
- The share of bestsellers in the output is capped — otherwise the tail of
the assortment gets no impressions and dies. - Products without images or with empty key attributes do not reach the top of the listing.
- Out-of-stock items are excluded from the output by a rule, not by hand.
- The promo calendar is synchronised with the pins on the site: when a promotion ends, the pin is
removed automatically. - Every change to the default sorting goes through an A/B test rather than shipping to everyone at
once. - The zero-result search report is reviewed monthly.