Your Priority Lane Is Not a Price Tier. It Is a Supply Promise Your Drivers Can Break.

Based on the research ofZhang, Zhang and Raghunathan, "Service Differentiation through Priority Matching by Ride-Sharing Platforms," MIS Quarterly, 2026

Zhang, Zhang, and Raghunathan's model of priority matching in ride-sharing, forthcoming in MIS Quarterly, starts from a familiar-looking setup: riders differ in how much they hate waiting, and a platform that cannot tell who is who offers two service levels so riders self-select. That is textbook quality-based price discrimination, the kind Mussa and Rosen formalized for conventional monopolists back in 1978. But the ride-sharing version breaks the textbook in two specific ways, and both are about supply, not demand. One-to-one matching means the platform cannot manufacture "quality" out of thin air the way a software company versions a product with a feature flag. And differentiating driver pay to fund the premium tier creates a second population of strategic actors, drivers, who react to the wage structure in ways that can undo the very differentiation the platform built.

Why the Quality Ladder Breaks in a Matching Market

In a conventional product line, quality is elastic. A software company can offer a "pro" tier with more features at close to zero marginal cost, and the classic prescription, going back to Mussa and Rosen's 1978 paper "Monopoly and Product Quality" in the Journal of Economic Theory, is to distort quality downward for every tier except the top one: degrade the cheap option just enough to protect the premium option's margin, while leaving the top tier undistorted.

A ride-sharing platform cannot do that cleanly, because match quality is not a dial the platform turns independently for each tier. It is an allocation of a shared, finite resource: the pool of available drivers at a given place and time. Assign your best-rated, closest driver to a High-tier rider and that driver is no longer available to serve a Low-tier rider nearby. One-to-one matching means every unit of quality handed to one segment is a unit of quality taken from the other. Zhang, Zhang, and Raghunathan's central finding follows directly from this constraint: the platform does not just degrade the low segment's match quality to protect the high segment, the way a conventional firm would. It enhances the high segment's match quality and degrades the low segment's match quality simultaneously, squeezing the menu from both ends, because the two tiers are drawing on the same pool rather than sitting on independent cost curves.

You can see this constraint operating in the products platforms actually ship. Uber Comfort, live in more than 40 cities, is explicitly built from "newer vehicles with extra legroom and highly rated, experienced drivers," a description of drivers pulled out of the general pool, not a software toggle. Uber Black goes further, restricting eligibility to a defined fleet of late-model luxury sedans and SUVs from brands like Audi, BMW, and Mercedes-Benz, with up to 15 minutes of guaranteed wait time versus a much shorter grace period on standard rides. Every one of these promises is a claim on a specific, limited subset of the driver pool. The "menu" riders see in the app is really a claim on a supply schedule the platform does not fully control.

A Premium Tier Is Only as Credible as the Supply Behind It

This is where Uber Reserve is the most honest product in the lineup, because its own terms admit the constraint the model formalizes. Reserve lets riders lock in a price with no surge and book up to 90 days ahead, explicitly trading flexibility for certainty. Yet Uber's own fine print concedes what the theory predicts: "Uber doesn't guarantee that a driver will accept your ride request... your ride is confirmed once you receive your driver details." The platform can promise a price. It cannot unilaterally promise a driver, because that promise ultimately depends on whether enough drivers are willing to be in the right place at the right time under whatever wage structure is attached to that trip.

Lyft's own product history makes the same point from a different angle. The tier once marketed as "Lyft Lux" and "Lux SUV" no longer exists under that name; Lyft's help pages for those legacy products now redirect to "Black and Black SUV rides," a premium tier promising high-end vehicles and "top-rated drivers" with a defined wait-time allowance. Consolidating three premium labels (Lux, Lux Black, Lux Black XL) into two (Black, Black SUV) is not just a branding refresh. It is a sign that maintaining several finely graded quality rungs, each requiring its own slice of a scarce, qualified driver pool, is harder to sustain than the rider-facing menu implies. The number of credible tiers a platform can offer is bounded by how finely it can actually segment its driver supply, not by how many SKUs a product manager can dream up.

Paying Drivers More for the Top Tier Can Make Them Stop Working for You

The paper's sharpest and least intuitive result concerns what happens once the platform tries to fund service differentiation by differentiating driver wages, paying more for trips assigned to the premium tier. If drivers are naive, this works the way you would expect: higher pay for premium trips, better drivers self-select into serving them, quality improves as designed. But real drivers are not naive. If a driver anticipates that holding out, declining the next low-paying assignment, might land them a higher-paying premium trip instead, they will wait. Zhang, Zhang, and Raghunathan call this forward-looking behavior, and they show it can force the platform's hand: if drivers wait to obtain the higher wage, the platform may find it optimal to offer a uniform wage across tiers even while it keeps offering differentiated service to riders. Supply-side cannibalization, drivers effectively cannibalizing the low-wage assignment pool by refusing to serve it, mirrors the well-documented demand-side cannibalization where price-sensitive riders trade down or price-insensitive riders trade up across a product line.

This is not a hypothetical risk invented for a theory paper. Ride-hailing platforms have already had to redesign real pay mechanisms because drivers behave exactly this strategically around differentiated pricing. Garg and Nazerzadeh's research on driver surge pricing, the theoretical work that informed Uber's shift to an additive surge mechanism, shows that the historically standard multiplicative surge is not incentive-compatible once drivers are modeled dynamically: drivers time their online status and trip acceptance around anticipated pay spikes rather than responding honestly to the platform's real-time supply needs. Uber had to redesign the entire surge payment structure because treating driver pay signals as fixed inputs to a static optimization problem, ignoring the fact that drivers watch those signals and plan around them, produced worse outcomes than a mechanism built to be robust to strategic waiting. Wage differentiation across service tiers is the same problem wearing a different costume: any time you pay drivers differently based on which rider segment they serve, you are creating a signal that forward-looking drivers will learn to game.

The Two Cannibalization Loops Do Not Move Independently

The paper's final and most operationally important result is that these two loops are coupled, not separate problems you can solve one at a time. Demand-side and supply-side cannibalization have qualitatively similar same-side effects (each pushes toward flattening the differentiation that created it) but asymmetric cross-side effects: a lever that fixes rider self-selection does not fix driver holdout the same way in reverse, and can make it worse. Widen the gap between Low and High match quality to stop riders from trading down, and you may need an even larger wage gap to keep enough drivers willing to serve the low tier at all, which is precisely the wage gap that triggers forward-looking holdout in the first place. Flatten the wage gap to stop supply-side cannibalization, and the platform still has to find a non-wage way to keep enough good drivers routed to the premium tier, since the match-quality differentiation on the rider side has to come from somewhere.

A premium tier is a supply commitment wearing a pricing costume.

Design the tiers as if they were shelf space in a warehouse, not features on a spec sheet. Every unit of "quality" you promise the high segment is a driver you are not promising to the low segment and a wage signal your driver pool will eventually learn to anticipate. The platforms that get this right, Uber restricting Reserve's guarantee with explicit acceptance caveats, Lyft consolidating three premium labels into two it can actually staff, are quietly admitting that the credible menu is smaller than the one a product roadmap would naively generate. Before you launch the next service tier, ask not "what quality can we describe in the app" but "what quality can our current driver pool deliver without drivers gaming the pay structure we just built to deliver it." The two questions have different answers, and the gap between them is where cannibalization, on both sides of the market, actually lives.