Bad Reviews and Disagreeing Reviews Send New Entrants in Opposite Directions

Based on the research ofAmy Zhao-Ding and Vibha Gaba, "Positioning in Digital Markets: A Demand-Side View," Organization Science, 2026

A developer building a new photo-editing app never has to guess what the market thinks of the apps already in the store. It can read the reviews. Before a single line of code ships, the App Store has already published, star by star, exactly how satisfied, and how divided, customers are with every existing competitor. Amy Zhao-Ding and Vibha Gaba's research on Photo & Video apps in the Apple App Store shows that new entrants use this pre-existing feedback to decide not just whether to enter, but how to position once they do, and that two different features of the same review data pull entrants toward opposite strategies. High customer dissatisfaction with rivals, meaning a low average rating, is associated with entrants sharpening their core function and differentiating harder on peripheral features. High heterogeneity in those same ratings, meaning customers disagreeing sharply about quality, is associated with entrants doing the opposite: dialing back core focus and imitating rivals' peripheral functions instead.

Reading Reviews Before You Have Any of Your Own

New entrants face what Zhao-Ding and Gaba call demand uncertainty: they do not yet know what combination of core and peripheral functions will match what customers actually want, and they have no purchase history of their own to learn from. What they do have is the existing market's report card. Apple aggregates every app's ratings into a public 1-to-5-star average, tracked and displayed per storefront territory, refreshed continuously, and weighted the same regardless of a rating's age or the app version it was left on. That average is a single visible number, but it collapses two distinct pieces of information into one: how high or low the number sits, and how much the individual ratings that produced it agree with each other. A 3.5-star average built from a tight cluster of mostly three- and four-star reviews describes a different market than a 3.5-star average built from a pile of five-star raves sitting next to a pile of one-star pans. Zhao-Ding and Gaba's contribution is showing that entrants respond to these two components differently, and that the response shows up in a measurable, structural choice: what fraction of the new app's functions concentrate on the core job, and how much its peripheral functions echo or depart from what rivals already offer. This is also standard practice on the supply side of app marketing: teams that study competitor reviews before a launch are explicitly hunting for this same split, what is broken versus what users cannot agree on, because the two point to different product bets.

When the Complaint Is Clear: Sharpen and Differentiate

Photo-editing apps built around Instagram's launch-era filters gave every entrant after 2011 an easy read: generic, similar-looking filter tools were common, and customers wanting anything more specific were left dissatisfied. Snapseed, built by Nik Software, launched on the iPad in June 2011 and answered that gap by going well past simple filters. It let people touch-edit specific regions of a photo, import RAW files, and make precise tonal corrections, the kind of control a one-tap filter could not offer, and it won Apple's iPad App of the Year within months. Google acquired Nik Software the following year specifically to bring that capability in-house. VSCO, launched in 2012 by Joel Flory and Greg Lutze, took the same dissatisfaction, that filter apps looked cheap and interchangeable, and answered it by building a distinct film-emulation aesthetic and stripping out likes and follower counts entirely, an unmistakably narrow bet on being the one place with a specific, recognizable look rather than the one place with the most features. Facetune followed the identical logic in a different corner of the category: launched by Lightricks in March 2013, it ignored general photo editing altogether and concentrated on one job, retouching selfies and portraits with teeth whitening, blemish removal, skin smoothing, and lighting correction, becoming one of Apple's most downloaded apps within a few years and passing 200 million downloads worldwide. None of these entrants tried to be a complete editing suite on day one. Each read a market where the standing complaint was legible and specific, and answered it by narrowing, not broadening.

When Users Cannot Agree on What's Wrong: Broaden and Borrow

Contrast that with the entrants who read a market of disagreement rather than a market of a single clear complaint. PicsArt, launched in 2011 out of Armenia, walked into a photo app market where user needs were fragmenting fast: some wanted filters, some wanted collage and layering, some wanted sticker and text overlays, some wanted community and remixing. Rather than pick one job, it built toward all of them, combining a layer-based editor, a template library, and eventually generative-AI tools into a single workspace that now serves roughly 150 million monthly active users. Snapseed's own trajectory tells the same story a second time, on a longer clock: the sharply focused touch-editing tool of 2011 has, through years of updates, absorbed batch editing, AI-assisted object removal, and an expanding tool palette, moving from specialist origin toward generalist breadth as the category's needs diversified. Adobe brought a similar full-suite logic to mobile with Lightroom, landing on iPad in April 2014 and iPhone that June, pairing organization with editing and tying the mobile app to the same cloud library and tools professionals already used on desktop rather than isolating one function. Where Facetune and VSCO bet everything on one recognizable job, PicsArt, latter-day Snapseed, and Lightroom bet on covering the waterfront: the generalist, imitate-broadly move the paper associates with heterogeneous rival feedback, where no single unmet need dominates the picture clearly enough to justify narrowing.

Why the Same Data Point in Opposite Directions

The logic is not arbitrary once you separate the two statistics. A low average rating with low dispersion is close to consensus: customers broadly agree the existing options fall short on some identifiable dimension, so an entrant that reads it has a specific hole to fill, and sharpening the core function while differentiating peripheral choices from the pack is close to a safe bet. A rating distribution with high dispersion carries the opposite information: customers are not agreeing on what is wrong, which means preferences themselves are heterogeneous, and no single specialist position reliably serves the whole addressable market. Broadening the core and imitating already-validated peripheral functions from multiple rivals is the lower-risk response to that kind of ambiguity. You are not betting on any one reading of what customers want; you are hedging across several readings at once. Zhao-Ding and Gaba also find this dependence on external, rival-generated feedback is strongest at entry and fades as the entrant accumulates its own experiential feedback loop. Once a new app has its own reviews to learn from, it leans on those instead of on what customers said about someone else's product.

Two entrants can read the identical 3.5-star category average and make opposite bets, because the number that matters is not the average. It is what produced it.

The App Store did not design its rating system to teach entrants how to position against each other, but that is functionally what the aggregate star display does at category scale. It broadcasts, continuously and for free, both how satisfied the existing market is and how much that satisfaction disagrees with itself, and new entrants are reading both numbers before they write a line of code.