When (and When Not) to Boost Bestsellers in Your Recommendations
Bestsellers are popular for a reason, so it feels obvious to put them everywhere. But boosting bestsellers in recommendations is one of those tactics that helps in small doses and quietly hurts in large ones. Lean on your proven winners too hard and you start recommending products people were already going to buy, starving new arrivals of the exposure they need, and ignoring what is actually in the cart in front of you.
This article walks through the real trade-off between recommending proven sellers and giving discovery products a chance, and gives you concrete guidance on when to boost popularity strongly, mildly, or not at all.
Why "always show bestsellers" backfires
A bestsellers-only recommendation strip looks productive. It is full of products that convert. The problem is what it costs you, and those costs are easy to miss because the strip still "works".
- You recommend what people would buy anyway. Your top sellers already have strong organic demand. Surfacing them again in a recommendation slot often takes credit for a sale that was going to happen regardless. The recommendation did not add incremental revenue, it just sat in front of an existing intent.
- You cannibalize new products. Fresh arrivals and long-tail items never get the impressions they need to prove themselves. If the only products that get shown are the ones already selling, nothing new can ever climb. You lock in last quarter's winners and freeze your catalog in place.
- You ignore cart context. A shopper buying a yoga mat does not need to see your single best-selling product overall - they need a strap, a block, or a cleaning spray. Popularity is a store-wide signal. Relevance is a per-cart signal. When popularity dominates, the recommendation stops answering the question "what goes with this?"
None of this means bestsellers are bad. It means popularity should be one input among several, not the whole decision.
Soft nudge vs hard override
The most important distinction here is between a recommendation that prefers bestsellers and one that forces them.
A hard rule removes products. "Only show items from this collection" or "never recommend anything over $200" are hard rules - they filter the candidate set down before anything is ranked. They are absolute.
A soft rule re-ranks without removing. A popularity preference says "all else being equal, lean toward the proven sellers", but it never deletes the relevant, lower-volume product that genuinely complements the cart. That is the key idea: a popularity boost should be a nudge, not an override.
This matters because your recommendation engine already learned what actually sells together from your order history. If you force bestsellers to the top, you are overwriting that hard-won relevance with a blunt store-wide average. If you nudge instead, you keep the relevance and only tilt the close calls toward your winners. The engine still respects the patterns in your orders, and popularity just breaks the ties.
When to boost popularity strongly, mildly, or not at all
Here is the practical part. The right amount of popularity boost depends on the surface, the catalog, and what you are trying to do.
Boost strongly when
- You have a long tail of weak or untested products and want to keep shoppers on safe, high-converting ground. A strong lean toward bestsellers is a reasonable default when most of your catalog is unproven.
- The slot is high-stakes and generic, like a homepage or a "you may also like" strip with no cart context to draw on. With little relevance signal available, popularity is one of the better fallbacks you have.
- You are clearing decision paralysis. A shopper staring at hundreds of options often just wants to know what other people picked. In that moment, "most popular" is a genuinely helpful answer.
Boost mildly when
- You have decent relevance signal but want a gentle quality floor. A mild boost keeps obviously-good complements in the running while quietly favoring the versions of them that sell best.
- Your catalog is healthy but uneven, with a mix of strong and developing products. A light nudge supports your winners without slamming the door on everything else.
- You are running cross-category complements, where the goal is a relevant pairing first and popularity is only a tiebreaker between similar options.
Do not boost (or dampen) when
- The cart context is strong and specific. If a shopper has a clear product in their cart, a tightly relevant accessory beats a generic bestseller almost every time. Let relevance lead.
- You are deliberately running a discovery slot to surface new arrivals or move long-tail stock. Here a popularity boost works against your own goal, and dampening it can actively help new products earn their first impressions.
- Your bestsellers are already over-exposed elsewhere on the site. If the same five products dominate your homepage, collections, and search, adding them to recommendations too is redundant and crowds out everything else.
How EliteCart handles the balance
EliteCart's Fine-tune EliteAI™ Ultra is built around exactly this soft-versus-hard distinction. Inside a fine-tuned version you shape recommendations with two kinds of rules: Filters, which are hard and remove products, and Boosts, which are soft and re-rank without removing anything.
One of the boost types is Popularity - "prefer your bestsellers" - and you set its strength on a five-step scale: Strongly dampen, Dampen, Neutral, Boost, Strongly boost. Because the boost is soft, it nudges ranking instead of forcing it. The engine trains on your own store's order history, so it already knows what sells and what sells together; the popularity boost just decides how heavily to lean on that store-wide signal versus the per-cart relevance it learned. Alongside Popularity you can boost by tag, vendor, product type, collection, or price direction, so you can combine a light popularity lean with, say, a same-vendor preference.
You can keep up to three live fine-tuned versions and assign a different one to each surface. That is what makes the guidance above actionable: a strong popularity lean on a generic "you may also like" strip, and a neutral or dampened version on a cart slot or a discovery section where relevance and new-product exposure matter more. You can also pick the base each version builds on - frequently-bought-together pairings, or cross-category complements - so popularity layers on top of the right starting point.
For the full mechanics, see the Filters and Boosts reference and the fine-tuning setup guide.
Start with a mild popularity boost and adjust per surface. Treat your bestsellers as a tiebreaker, not a takeover. Keep relevance in the driver's seat where the cart gives you context, lean harder on popularity where it does not, and leave room for new products to earn their place. For the wider picture, read our merchandising rules for product recommendations, see how the same soft-rule thinking applies to price-aware product recommendations, brush up on the difference between cross-sell and upsell recommendation strategy, and explore what is new in EliteAI fine-tuning.