Fresh Stock and Proven Sellers Fight for the Same Ten Slots
Rank purely on sales history and new inventory never sells. Boost newness blindly and the best products sink. The ranking compromise that gives new stock a fair trial without taxing the shopper.
Every ranking system runs a quiet rigged election. Products with sales history accumulate reviews, conversion data, and click evidence, so they rank. Products without history rank nowhere, so they never acquire history. Left alone, a catalog's top slots fossilize around last season's winners while new stock, which the buyers paid for and the marketing team is promoting, sits on page four acquiring nothing. The usual correction, a newness boost, creates the opposite scandal when it is crude: a large launch week floods the first page with untested products and conversion craters. The working compromise is a trial system, not a boost. New items receive a bounded number of impressions in relevant results, in middle slots rather than the crown positions, and they graduate or fade on the evidence those impressions produce. The shopper funds a fair trial with a small amount of attention. The catalog stops eating its own future.
Trials, quotas, and the middle slots
Give a new item its trial where a trial is cheap: slots four through ten of the results it genuinely matches, the second row of the category grid, the tail of a recommendations rail. Those positions get real clicks from real shoppers with real intent, and a few hundred impressions at those slots will tell you whether the item converts at, above, or below its category's rate. Crown slots, positions one to three, are where the category's credibility lives. Putting an untested item there is not fairness, it is gambling with the page's conversion to subsidize a product that has not earned it yet.
Bound the trial in advance and say the number in the ranking doc: every eligible new item gets its impressions within its first two weeks, spread across days and dayparts so a launch into a quiet Tuesday is not judged against a Saturday baseline. When the trial ends, the item ranks on the same evidence as everything else. No permanent newness dividend, because a permanent dividend is how a six month old product with a refreshed feed date ends up outranking the item shoppers actually buy.
Related reading on this site: RelevanceRail and Search Relevance Tuning: Fixing Zero Result Searches cover the neighbouring problems in more depth.
Feed dates lie, so police them
Every trial system creates an incentive to fake freshness, and feeds oblige: a supplier re sends the catalog with new timestamps, a variant gets a new SKU and the parent product is reborn as new, a price change refreshes the record and the newness clock with it. If your ranking reads freshness from the feed, your ranking is for sale to whoever edits the feed. Anchor novelty to first seen date in your own index, keyed to a stable product identity that survives SKU churn, and audit the new items list weekly for familiar faces wearing new paperwork. The audit takes twenty minutes and it is the difference between a trial system and a loophole.
Seasonal returns are the legitimate edge case. A product that sells every winter and vanishes every spring should re enter with its history intact, not as a newborn. Keep the evidence, decay it over the off season, and let the product re earn its slot quickly on last year's record plus a short re trial. Treating proven seasonal stock as new wastes the trial budget on items that do not need it.
Report the fossilization, not just the clicks
Add one number to the ranking review: the age profile of the top twenty results in your top categories, median days since first seen, tracked monthly. When that median climbs quarter after quarter, the catalog is fossilizing no matter what the click through rate says, because click through measures how shoppers respond to what they were shown, and they cannot click the new stock they never saw. The fossilization number is the early warning. The conversion number is the late one, and by the time it moves, the buyers have a season of dead inventory to explain.
Frequently asked questions
Why does new inventory struggle to rank?
Ranking systems reward products with accumulated clicks, reviews, and sales. New products have none of that evidence, so they rank low, get few impressions, and never gather the evidence needed to rise. It is a self reinforcing loop.
How should new products be given a chance in search results?
Through bounded trials: a set number of impressions in middle slots of relevant results over the first weeks, judged against category conversion rates. Items that perform graduate to normal ranking; items that do not fade without having damaged top slots.
What is catalog fossilization?
It is the gradual aging of top ranked results as historical winners hold their positions and new stock never breaks in. Tracking the median age of top results in key categories makes it visible before conversion suffers.
How do sellers fake product freshness?
By resending feeds with new timestamps, issuing new SKUs for existing products, or triggering record updates that reset newness clocks. Ranking systems should anchor novelty to the first seen date in their own index, not to feed supplied dates.