Choosing what to put in the recommendation slots on a storefront
Ranks the catalogue for a given shopper from what comparable shoppers went on to buy and what this one has already viewed, filling the slots on the product page, the basket and the follow-up mail, with the ranking refitted on the interactions the storefront logged rather than on a fixed hand-written rule about which items go together.
- Effort
- Weeks of work
- Skill level
- Some technical skill
- Organisation size
- Small business
- Value
- Revenue, Quality
Tools named for this
- An implicit-feedback recommender fitted on the storefront's own interaction log
- A properly tuned nearest-neighbour or popularity method kept running as the control
- An online test harness, because offline ranking metrics do not settle this on their own
What to check before you ship it in India
- A behavioural profile is built out of personal data. Consent under section 6(1) is limited to such personal data as is necessary for the specified purpose, so a recommender whose feature set quietly widens to whatever the analytics stack happens to collect has moved past the notice the shopper agreed to.
- The model is trained on the log of what it previously chose to show. Items it never surfaced generate no interactions, so a slow tail of the catalogue reads as unwanted when it was only unseen, and each refit hardens the mistake.
Sources
Every claim on this page traces to one of these, on the date it was read.
- The Digital Personal Data Protection Act, 2023 (No. 22 of 2023) — most obligations commence 13 May 2027 under the DPDP Rules 2025 — s.6(1) · Ministry of Electronics and Information Technology · a rule · read 2026-09-01
- Are We Really Making Much Progress? A Worrying Analysis of Recent Neural Recommendation Approaches · arXiv (Ferrari Dacrema, Cremonesi, Jannach); ACM RecSys 2019 · how it is done · read 2026-09-01