Sharing Your Market Data Doesn't Help Sellers Sell More. It Helps Them Build Better Products.

Based on the research ofYubo Chen, Xuebin Cui, Aishen Li, Banggang Wu, and Liu Yang, "The Role of Digital Platforms in Data Markets: How Platform-Shared Market Data Through Advanced Analytics Empower Firm Innovation," Management Science, 2026

The reflex answer to "what does market data do for a seller" is that it helps them price better or target ads better. A new paper anchored in Alibaba's Taobao Marketplace shows the answer is neither. Platform-shared analytics push retailers to build a better catalog, and that alone is what moves sales.

Yubo Chen, Xuebin Cui, Aishen Li, Banggang Wu, and Liu Yang study a natural experiment inside Taobao in which the platform shared advanced data analytics, built on top of market-wide sales and search data, with a subset of retailers. Retailers who got the advanced layer, not just descriptive dashboards of their own transactions, sold 31.8 percent more than retailers left with descriptive analytics alone. That is a large effect for a change that involved no discount, no coupon, and no new placement in the app. The platform did not hand out demand. It handed out a sharper picture of demand that already existed, and let the retailer act on it.

The paper's most useful move for anyone running a platform is what it does next: it rules out the obvious explanations. Retailers with access to advanced analytics did not change their prices differently than the control group. They did not change advertising spend differently either. Both are the first places a platform operator would expect the effect to show up, since price and ad intensity are the levers most seller tools are built to inform. Neither moved. What did move was the product line itself: retailers who received platform-level market analytics expanded their category scope by 6.8 percent and introduced 10.5 percent more new SKUs than the control group. The sales gain traces back to a wider, better-targeted catalog, not to sharper bidding or cheaper prices.

Why information, not data, is the thing platforms can actually share

The regulatory backdrop matters here. Data-sharing across firms usually runs straight into privacy law: a platform cannot casually hand one seller another seller's transaction records. Chen and coauthors' framing is that a platform with enough scale and analytics capability can sidestep this constraint by sharing processed market information rather than raw data. The retailer never sees a competitor's order history. They see a derived signal, what's trending in the category, where demand is shifting, which subcategories are underserved, that the platform computed across its whole marketplace. This is the same logic that shows up, in a much lighter-weight form, in the seller-facing analytics tools that Amazon, eBay, Etsy, and Walmart already run.

Amazon's Brand Analytics is the clearest existing example of platform-level information substituting for raw data. It is available to Brand Registry members with a Professional selling account, and its dashboards, Search Query Performance, Top Search Terms, and Market Basket Analysis among them, are all built from aggregated buyer behavior across the marketplace, not from any single seller's own order data. The Market Basket Analysis view, in particular, tells a brand which other sellers' products get bought alongside its own and at what combination rate, which is platform-level assortment intelligence in exactly the shape the Taobao study describes: information a retailer could never construct from its own sales log alone.

eBay's Terapeak, acquired by eBay in 2017 and made free to all Seller Hub users in an April 2021 rollout, works the same way at a category level. Terapeak Product Research gives sellers up to three years of sales history across millions of listings, sales trends, sold-price ranges, and sell-through rates, aggregated marketplace data a seller uses to decide what to list and how to price it, without ever seeing another seller's individual books. Terapeak Sourcing Insights goes a layer further, showing category-wide demand patterns to inform what to source next, which is closer to the assortment-planning use case the anchor paper isolates as the actual driver of the sales gain.

Etsy's Marketplace Insights tool, and Walmart's redesigned Seller Analytics dashboard unveiled at its Let's Grow! Summit in April 2026, both point the same direction. Etsy's tool surfaces keyword search volume and trend data drawn from marketplace-wide buyer search behavior, capped at 15 keyword lookups per week on the free tier, explicitly to help sellers spot what to make or stock next rather than how to price an existing listing. Walmart's rebuilt dashboard added category-level benchmarking, letting a seller compare their own conversion rate and content score against the marketplace average for their category, plus a roadmap toward predictive demand forecasting later in 2026. None of these tools were built primarily as pricing calculators or ad-optimization consoles. They were built as assortment-planning aids, and the Taobao paper is the first causal evidence, that a growing family of academics has been pointing to, that this framing is right: the sales lift shows up through the catalog, not the price tag.

Assortment moves are slower and stickier than price moves, and that is the point

There is a structural reason the effect surfaces in assortment rather than price. A price change is reversible within a day; a new SKU or an expanded category requires sourcing, listing, and inventory commitment that takes weeks and cannot be casually undone. That the paper finds retailers making this slower, costlier kind of move in response to better information says something about what the information was actually worth to them. A retailer does not commit to ten new SKUs on a whim. They commit when the data convincingly shows unmet demand in an adjacent category, the kind of pattern that is only visible at platform scale, across thousands of other sellers' search and purchase behavior, not inside one shop's own transaction log.

This also explains why the gain is more persistent when the analytics draw on platform-level data rather than firm-level data alone. A price cut's effect decays the moment a competitor matches it. A well-chosen new product category keeps generating sales as long as the underlying demand persists, which is longer than most competitive price wars last. Platforms optimizing purely for short-run GMV might be tempted to build seller tools around pricing and promotion, since those show up fast in a dashboard. This paper is evidence that the more durable lever, for the platform's own long-run marketplace health as much as for the seller, is helping sellers figure out what to sell in the first place.

The platform's real product isn't traffic or ad tools. It's the market intelligence that tells a seller what to build next.

What this means for platforms weighing a data-sharing feature

The instinct to withhold market-level data, treating it as a scarce competitive asset the platform alone should exploit, misreads where the value actually sits. Shared as processed information rather than raw data, it does not commoditize the platform's own analytics business, and it does not obviously help sellers undercut each other on price, since price barely moves. It mostly does something a platform should want more of: it makes the marketplace's catalog richer, with more SKUs and wider category coverage supplied by sellers who now know where the gaps are. A denser, better-targeted catalog is a demand-side asset for the platform itself, not just a favor to the sellers who receive the analytics.

The caution is proportional design. Not every seller needs, or can act on, the same depth of market intelligence, and building predictive-analytics infrastructure at platform scale, the kind Taobao ran and Walmart is now rolling toward, is a real cost center that has to be justified by the assortment and retention gains it produces. But the mechanism this paper isolates, that market information, shared carefully and in aggregate, moves sellers to build rather than to just compete harder on the same shelf, is a strong argument for building that infrastructure rather than defaulting to descriptive-only dashboards and calling the job done.