← Blog

Stop Fixing Your Dashboard. Give Your Data to Someone Who Cannot Lie.

The credibility of your analytics falls as your profit from bending them rises. The fix is to hand the raw numbers to a party with no stake in the answer.

Based on the research ofYi Liu and Fei Long, "Data and Algorithms: Strategic Disclosure of Competitiveness on Platforms Through Marketplace Analytics," Marketing Science 45(3), 2026

Platform Performance · Marketplace analytics and trust · August 2026

In their 2026 Marketing Science article, Yi Liu and Fei Long ask why a platform would ever open its proprietary data to a neutral third-party analytics provider rather than keep it for itself. Their answer inverts the usual data-moat logic. A platform earns a commission on sales, so it is tempted to design analytics that make the market look less competitive than it is, nudging sellers to price higher. Sellers anticipate this, discount the first-party tool, and under-adopt it, so they price poorly and the platform's commission base shrinks. A disinterested third party with no claim on the seller's margin supplies a credible signal. Opening data works as a commitment device that recovers adoption, pricing, and commissions. It is optimal, the authors show, when platform-seller competition is weak to moderate.

The tell is the size of the cut: the more a platform pockets per transaction, the more its scoreboard is worth bending, and the harder the sophisticated side discounts it (see the exhibit).

Amazon: when you compete with the sellers you measure

Amazon runs the clearest version of the scorekeeper's conflict, and the stakes are enormous. Third-party sellers generate roughly 60 percent of Amazon's retail sales, and on each of those sales Amazon takes a referral commission of about 15 percent, rising past 45 percent once fulfillment and advertising fees are stacked on, per the Federal Trade Commission's 2023 complaint. The scale is not abstract: Amazon booked about $156 billion in 2024 from third-party seller services, and 62 percent of units sold in the fourth quarter of 2024 came from its 1.9 million active third-party sellers. Amazon also controls the search ranking that decides who sells, and it operates private-label lines. In April 2020 the Wall Street Journal reported that Amazon employees used individual third-party seller data to build private-label products such as AmazonBasics. That conflict is now the center of two enforcement actions: the FTC and 18 states sued Amazon in September 2023 in a monopolization case whose bench trial is now set for 2027, and in December 2022 the European Commission accepted binding commitments barring Amazon from using nonpublic marketplace seller data for its own retail and private-label business.

A seller reading Amazon's own demand and competition signal must price in the possibility that the party reporting the number also wants to launch a rival and can bury them in search results. So sellers pay for neutral tools. Jungle Scout and Helium 10 build competitive and keyword analytics on top of the sales and catalog data Amazon exposes through its Selling Partner API. Adoption of that neutral layer now dwarfs many of the brands it measures: Helium 10 reports over 1,000,000 users of its sales estimates against roughly 400,000 for Jungle Scout. The disinterested tool, with no private-label line and no search dial to turn, becomes the trusted read.

Wherever you compete with the sellers you also measure, your own analytics inherit a discount no dashboard redesign can remove. A neutral data layer functions as a commitment device that a first-party tool cannot replicate. Treat API openness to certified independent tools as trust infrastructure, not developer convenience.

Wherever you compete with the sellers you also measure, your own analytics inherit a discount no redesign can remove.

Airbnb: when your fee scales with volume, your pricing tool has a visible bias

Airbnb's Smart Pricing is widely distrusted by hosts, who document that it pushes rates toward their set minimum and optimizes for filling the calendar rather than maximizing per-night revenue. That bias is not a bug to the platform. As of a 2025 rollout that became universal in 2026, Airbnb charges software-connected hosts a single host-only service fee of 15.5 percent of the booking subtotal, rising to about 16 percent in some markets. When your revenue is a fixed cut of volume, a tool tuned for occupancy serves your economics, and sophisticated hosts detect it and route around it to neutral, host-side, revenue-maximizing engines.

Then Airbnb did what the paper predicts. PriceLabs, a dynamic-pricing engine that charges hosts about $19.99 per listing per month and takes no share of their nightly rate, became an official Airbnb Software Partner with authorized API access to push prices directly onto listings. The pricing engine hosts trust is precisely the one with no claim on their revenue, and PriceLabs now prices more than 600,000 listings across more than 150 countries. The real host decision is now Smart Pricing versus neutral tools like PriceLabs, Beyond, and Wheelhouse that optimize host earnings rather than platform bookings. The mechanism is exactly the one Liu and Long model: because the neutral engine cannot profit from shading the number down, hosts believe it, adopt it, and price with it, and the higher-quality pricing feeds back into the booking value Airbnb's fee rides on.

The implication for any volume-fee platform: your first-party pricing tool carries a visible downward bias, and your most valuable operators will find it and defect. Blessing a neutral partner recovers those hosts and their inventory, at the cost of accepting somewhat higher prices you no longer set.

App stores: the seller of the ads cannot be the scorekeeper

Advertisers refuse to let the party selling the ads also count the results. Apple's App Store Connect analytics are structurally distrusted: they reflect only opt-in users, roughly 20 to 30 percent of a base, and cannot tie an install to a specific campaign. So the industry standardized on neutral, cross-platform measurement partners, AppsFlyer and Adjust, as the agreed counting house. Their independence has real market weight: AppsFlyer alone counts installs for more than 15,000 brands and, with Adjust, books close to 45 percent of global measurement revenue, and investors valued it at $2.7 billion in a June 2026 round. Apple's own economics are not neutral either: it keeps 30 percent of most digital sales, or 15 percent for developers under $1 million. Persistent discrepancies between store-reported and partner figures are exactly why the independent layer exists. Apple's SKAdNetwork and AdAttributionKit supply a privacy-safe signal the partners aggregate across stores, and a March 2026 App Store Connect update still left cross-channel individual attribution outside first-party analytics.

The implication for any ad-selling platform: a neutral measurement layer is the precondition for advertisers spending with confidence. Feeding clean signal to certified third-party measurers grows ad and in-app-purchase budgets more than forcing everyone onto your own numbers. One caveat worth naming to your team: measurement partners are paid by developers, so neutral here means not-the-ad-seller, not disinterested in the strict sense.

YouTube: opening a read API turns optimization into a growth flywheel

YouTube's recommendation and monetization logic is opaque, and creators suspect it is tuned to watch time and ad load rather than their channel economics. Those economics are concrete: YouTube pays creators 55 percent of ad revenue on long-form videos and keeps 45 percent, an unchanged split that defines exactly the earnings a creator fears the algorithm does not optimize for. YouTube's response again tracks the paper. It opened a read Analytics API through OAuth so a neutral ecosystem, VidIQ, TubeBuddy, and Social Blade, can surface the competitive benchmarks YouTube Studio withholds. Only tools authorized through the official OAuth API see a creator's true RPM; scraper-based estimators can be off by tenfold or more. The credibility of a third-party number depends on whether the platform actually granted it clean access.

The implication: opening a read API to a neutral optimization ecosystem raises creator investment in titles, cadence, and thumbnails, which raises the watch time and the 55-percent ad share you both live on. But data you open gets monetized by others, and bad estimators create confusion you get blamed for. Accuracy governance matters as much as access.

The rule any platform can act on

The credibility of your analytics is inversely related to how much you profit from bending them. If you take a commission, sell the ads you also measure, or compete with your own sellers, you hold a scorekeeper's conflict, and the sophisticated side of your market prices it in by discounting your tool and under-adopting it. They then price, spend, and invest worse, which shrinks the very commission base you skewed the numbers to protect.

The escape is not a better dashboard. It is delegation. Open the raw data to a party with no claim on the seller's margin: Amazon's SP-API to Jungle Scout and Helium 10, Airbnb's API to PriceLabs, the app stores' postbacks to AppsFlyer, YouTube's Analytics API to VidIQ. Because that party cannot cash in on distortion, its number is believed, adoption returns, and you recapture commission you were leaving on the table. This wins when platform-seller competition is weak to moderately strong. When your competitive stake in shading is large, a big private-label or first-party ad business, the temptation to hoard grows, and the commitment value of opening data is exactly what becomes hardest to give up.

Do not try to earn trust by redesigning a tool whose owner benefits from lying. Earn it by handing the data to someone who does not.