Recommendation Engines: How 'You May Also Like' Works
Those 'customers also bought' and 'recommended for you' panels are powered by recommendation engines. They can lift sales noticeably, and the basic ideas are easy to grasp.
This article explains the common approaches in plain terms.
Two Main Approaches
- Content-based: suggest items similar to what someone already likes.
- Collaborative: suggest what similar customers liked.
- Hybrid: blend both for better results.
What They Need
Recommendations improve with data — product details and a history of customer behaviour. New stores face a cold start, where there is too little data to personalise well at first.
Measuring Success
Recommendations are easy to add but easy to fool yourself about, so measure them against a sensible baseline rather than assuming they help.
- Agree a metric such as click-through or extra revenue.
- Compare against a simple baseline like best-sellers.
- Keep what genuinely beats the baseline.
Frequently Asked Questions
Do recommendations work for small catalogues?
They can, but with few products the gains are smaller and simple best-seller lists often perform nearly as well.
How long before they improve?
They get better as customer behaviour accumulates, so expect modest results at first and steady improvement over time.
If you need a hand with any of this, your Progressive Robot delivery team is ready to help. Raise a ticket from the Support area of your client portal or speak to your account manager and we will guide you through the next steps.