The role of bestsellers in a recommender system
Bestsellers are the simplest and most universal recommendation strategy. They need no personal
profile and work from the first visit. In a mature recommender they serve two functions.
A cold-start fallback. A new visitor with no view or purchase history cannot receive personal
recommendations, so the system shows the top sellers instead. That beats a random selection or an
empty block.
A standalone widget. A Bestsellers block on the homepage or in a category is a pattern shoppers
trust: if other people are buying it, it is probably good.
Popularity as a quality signal
A high sales count is an aggregated satisfaction signal. Unlike a rating (which can be gamed) or
views (which do not imply a purchase), the number of transactions is a hard metric.
| Strategy | Signal | Precision | Needs a profile |
|---|---|---|---|
| Bestsellers | Sales volume | Low (not personal) | No |
| Trending | Sales growth | Medium | No |
| Personalized | One shopper’s behaviour | High | Yes |
Tip: treat bestsellers as a complement to personalization, not a substitute. When a personal
profile exists, serve personal recommendations; when it does not, category bestsellers beat a
global top list.
Merchandising rules for bestsellers
An algorithmic sales ranking rarely matches commercial priorities exactly. Merchandising rules
adjust the result:
- Boost — lift an item above its algorithmic position (a high-margin promotional product)
- Pin — fix a position (sponsored placement, an exclusive)
- Exclude — remove an item (discontinued, or unwanted by policy)
- Stock filter — automatically drop items with zero inventory