Price sensitivity in recommendations: when showing the expensive option helps and when it kills the click

October 23, 2026

Short answer: Every shopper carries a working price frame built from the product they are viewing, their basket, and their history. Recommendations inside or near that frame feel helpful. Recommendations far above it need a clear reason the shopper can see, better material, a set, a flagship feature, or they read as the store ignoring the budget in front of it. The click data punishes that feeling quickly and consistently.

Band recommendations to the anchor

The practical starting point is a band around the anchor product's price, commonly tested somewhere between a modest step down and a step up of roughly half to double, tuned per category. Within the band, rank by relevance and margin freely. Outside it, require a justification the interface can express: a bundle that completes a set, the pro version where the difference is the story, clearly labelled as a trade up rather than slipped into a row of peers.

Bands should flex by context. A shopper viewing a flagship product has declared a higher frame than the store average, and recommendations should honour that declaration rather than dragging them back to the median. A shopper with a basket of sale items is telling a different story. Static storewide bands ignore both signals, which is why the first improvement is usually making the band relative to the anchor and the session, not picking a cleverer fixed number.

The trade up, done honestly

A deliberate trade up slot can work beautifully: one clearly better option, visibly labelled, with the difference explained in the terms the shopper cares about, durability, capacity, warranty, performance. It works because it respects intelligence. The shopper is offered a choice with a reason, not ambushed by a row where every alternative quietly costs twice the product they were considering, which reads as contempt for their budget.

Watch what trade up clicks do next. A healthy trade up click converts or returns to the original product and buys it. An unhealthy one exits. If trade up slots raise exit rates on the page, the store is spending its best recommendation real estate on a margin wish. Demote the slot, narrow its band, or restrict it to categories where the quality ladder is real and demonstrable rather than a badge and a hope.

New visitors have no frame but the page

For a first time visitor, the viewed product is the entire price signal, so band discipline matters most exactly where data is thinnest. Resist filling new visitor recommendations with the store's highest margin lines on the theory that there is no history to contradict it. The page itself is the history. A recommendation row priced as if the visitor were someone else teaches them in ten seconds that the store's suggestions are ads, and that lesson survives into every later visit.

Sale contexts need their own rule. A shopper viewing a discounted anchor is anchored on the paid price, not the reference price, whatever the merchandising team wishes. Recommendations banded to the reference price look inflated by exactly the discount, and click rates show it. Band to the transaction reality the shopper experienced. The frame lives in the shopper's head, not in the product file, and it is built from the number they would actually pay.

Measure the band, then set it

Run the band as an experiment, not a belief: test widths by category, read conversion and basket outcomes rather than click rate alone, and segment by new and returning visitors. Most stores find their profitable band is narrower than the initial guess and different across categories in ways nobody predicted, which is the entire argument for measuring instead of debating it in a meeting room.

Revisit the band as prices move. Inflation, a range restructure, or a new premium line shifts every anchor in the store, and a band set two years ago in absolute terms may now exclude the core range or include the absurd. Relative bands age better than absolute ones for this reason. The principle outlives any setting: recommendations should feel like they were chosen for the shopper in front of you, at the price they are actually considering. The moment they feel chosen for the store's margin, the click, and eventually the customer, goes elsewhere.

Keep reading: Replenishment recommendations: reminding repeat buyers at the moment they actually run out, Context aware recommendations: using season, weather, and place without being creepy, and Recommendations on zero results pages: recovering the lost search cover the nearby cases in this series.