Quick Notes Are AI Product Recommendations a Black Box? October 16, 2026 by Wade Most are. Typical AI recommendation widgets cannot tell you why a specific product appeared in a specific slot, which means you cannot debug an embarrassing suggestion or steer the logic toward business goals like margin. The alternative is explainable ranking: recommendation logic written as rules you can open, read, and edit. The Problem with Black […] Read more »
Quick Notes Handling Products With Many Variants in Shopify Collections October 15, 2026 / June 12, 2026 by Wade Products with many variants create two collection problems on Shopify: the 99 variant limit forces some merchants to split one product into several listings, and those split listings then appear as near-duplicates in collections. The fix is combined products: keep separate products behind the scenes, present one tile in the collection, and deduplicate so shoppers […] Read more »
Quick Notes How to Document Your Shopify Merchandising Strategy October 13, 2026 / June 12, 2026 by Wade Document your merchandising strategy as a set of rules, not a slide deck. For each collection type, write down the goal, the signals that should drive ranking (and their relative weight), the exceptions (pinned heroes, excluded products), and how success is measured. If you use a sorting tool, the recipes themselves become the documentation: they […] Read more »
Quick Notes Product Recommendations for a New Store with Little Data October 11, 2026 by Wade A new store can run useful recommendations from day one by leaning on catalog signals it already has: product recency, price position, stock, margin, and collection relationships. Behavior signals like conversion rate join the ranking automatically as traffic accumulates, so the rows improve gradually instead of guessing from nothing. Start with Catalog Signals Recommendation logic […] Read more »
Quick Notes How Long Should You Run a Collection Sort A/B Test? October 10, 2026 / June 12, 2026 by Wade A collection A/B test should run until each variant has accumulated enough conversions to separate a real difference from random noise, and for at least one or two full weekly cycles so weekday and weekend shoppers are both represented. For most collections that means weeks, not days. Set the end condition before you start, and […] Read more »
Quick Notes Should Low-Stock Products Rank Higher or Lower in Collections? October 8, 2026 / June 12, 2026 by Wade In most stores, low-stock products should rank lower, not higher. The last few units are usually broken sizes and leftover colors, so promoting them sends shoppers into products they cannot actually buy in their variant. The exception is deliberate scarcity (limited drops, one-of-a-kind items), where low stock is the story. For everyone else, treat nearly-gone […] Read more »
Quick Notes How to Stop One Brand from Dominating Your Recommendation Rows October 6, 2026 by Wade On multi-vendor catalogs, recommendation rows tend to collapse toward the most popular brand: its products win every demand-based ranking, so it quietly fills every slot. The fix is a per-brand cap, a rule that limits how many products one vendor can hold in a row, forcing variety without abandoning data-driven ranking. Why It Happens Popularity […] Read more »
Quick Notes The Best Sort Order for a Best Sellers Collection October 5, 2026 / June 12, 2026 by Wade The best sort order for a best sellers collection is sales velocity over a recent window, typically the last 30 days, combined with inventory awareness so sold-out products fall to the bottom. Shopify’s built-in best selling option works, but it leans on longer sales history, so it goes stale and keeps yesterday’s winners on top […] Read more »
Quick Notes The Cold-Start Problem: Why Best-Selling Ignores New Products October 3, 2026 / June 12, 2026 by Wade Shopify’s best selling sort ranks products by past sales, so a new product with zero sales history starts at the bottom, gets little visibility, and therefore makes few sales. That feedback loop is the cold-start problem. The fix is a blended sort that weights product newness alongside sales, giving new items a temporary visibility window […] Read more »
Quick Notes How Many Recommended Products Should You Show? October 1, 2026 by Wade Fewer than most stores show. Post-purchase, three or four products read as a confident suggestion. On product and cart pages, one clean row sized to the device works better than a grid. And every row needs a minimum count, at least two, below which it hides itself entirely rather than render sparse. Fewer, Better A […] Read more »