DeconstructAlphaDECONSTRUCTALPHA中文
← All articles

Product & business

What a voluntary ad click cannot tell you

A rewarded ad can offer a worthwhile exchange or charge time to remove an obstacle the product created. The difference becomes clearer when we examine what remains after someone declines.

Two explanations behind one click

Imagine two puzzle games offering an ad in exchange for an undo after a failed round. In the first, restarting is immediate and free, and the rules are clear. An undo preserves the player’s progress. In the second, a poorly placed control causes frequent mistakes; the ad is the easiest way to recover from them. Both games could record high voluntary viewing rates.

This hypothetical comparison exposes a measurement problem. A click shows a choice among the options available. It does not explain why the alternatives became unattractive. A product designs those alternatives and then records behavior within them as evidence of demand. That can produce a self-reinforcing conclusion.

Our testable proposition is that, when rewarded ads mainly relieve friction created by the product, repairing the free path may reduce viewing while improving task completion and subsequent return. When ads mainly provide additional value, demand for the exchange may persist after that repair. Both mechanisms can operate together. A click alone cannot distinguish them.

The exchange includes more than viewing time

A person may be weighing three costs: time spent watching, loss of current progress, and the effort of doing the task again. Even with an identical ad duration, the exchange changes with the point of failure, accumulated effort and difficulty of restarting.

Increasing the reward from one undo to three is therefore not the only way to increase viewing. Making restarting more cumbersome could produce the same short-term result. A system optimized solely for viewing rate could reward both changes similarly, although they create very different experiences.

Challenge complicates the distinction. Failure having a cost does not by itself indicate bad game design. The relevant question is whether the difficulty belongs to the task the player wants to practice. Reading a board and planning a route belong to the game. Finding an obscure dismiss button generally does not. In a utility task such as exporting a file, additional waiting also needs an explanation in actual service costs or functional value.

Explicit choice answers part of the question

Google’s rewarded-ad policy requires clear advance disclosure of the required action and reward. Ordinary rewarded ads require an affirmative choice; rewarded interstitials use a different opt-out mechanism. Skipping or declining must not interfere with normal use, and the promised reward must be delivered after the required action.[1]

Those requirements address understandable terms, effective refusal and delivery of the promise. The causal question requires additional evidence: with this underlying experience, is the user seeking an extra benefit or relief from avoidable friction? Satisfying an interface rule does not automatically answer that question.

A reward can be described perfectly while the free route has recently acquired two unnecessary steps. Those steps still belong in the evaluation. Knowing what one is accepting cannot establish that every cost imposed by the product is worthwhile.

Why high retention among viewers can mislead

Suppose 100 people reach a reward offer, 20 watch and 16 of those return the following week. Viewers have an 80% return rate. That number can be accurate without establishing that the ad caused the return. People willing to preserve the round may already be more engaged and more likely to come back.

These numbers are invented. The problem is the comparison: watching is a decision made by the user, so viewers and non-viewers may differ beforehand. Comparing the two groups mixes prior engagement with the effect of the ad.

The denominator matters as well. People who leave because of friction before reaching the offer are absent from a report restricted to offer viewers. A less usable version could leave a more tolerant audience behind, increasing viewing rate among survivors. Evaluation should therefore include all eligible users assigned to the version, including departures before the reward entry point.

An experiment that can separate mechanisms

First describe the current free experience: whether restarting is immediate, what progress persists, and what happens when an ad cannot load. Choose a friction that can be repaired independently, such as a redundant confirmation. Do not manufacture a worse free path for the experiment.

With enough traffic, and only where temporarily omitting the offer would not remove an existing entitlement, cross two changes: current versus improved free path, and reward offer present versus absent. This creates four groups. Randomize users before they encounter the offer and keep their assignment stable during observation. You can then compare the effect of adding the offer within the same free experience and whether improving that experience changes the effect.

Specify primary outcomes and windows in advance. Examples include completed tasks per assigned user, return within seven days and cumulative revenue over twenty-eight days. Check failed ad loads, missing rewards and complaints as possible side effects. These measures will not automatically agree. State acceptable experience losses before seeing revenue, rather than relaxing standards after it rises.

If repairing the free route reduces viewing but increases completion and return, that supports the explanation that some ad demand came from friction. If the repaired route still gains revenue from the offer while completion and return stay within prespecified bounds, that better supports independent value. Each result concerns this audience, reward and observation window. With small samples, report uncertainty instead of declaring a winner from a few clicks.

When the exchange is worth keeping

Rewarded advertising can be a reasonable exchange. Someone unwilling to pay money may willingly spend a little time for an extra opportunity, while the developer earns revenue to sustain the service. A free route need not include every reward, and every alternative need not be costless.

The opposing boundary is concrete too. If declining prevents someone from retrieving work already completed, or rewards mainly repair interaction errors created by the product, high viewing deserves scrutiny. Enlarging the reward may conceal deterioration in the underlying task.

Evaluation should record what people accomplish after declining, how they continue when rewards are unavailable and whether they return later, alongside acceptance. That makes it possible to assess revenue in relation to the task people originally came to complete.

Sources & scope

Scenarios and numbers are hypothetical. Mechanisms and the experiment design are the author’s analysis. The Google policy supports the summary of presentation, refusal and reward-delivery requirements; it supplies no experimental finding about retention or revenue. Checked November 11, 2026.

  1. Google AdMob — Policies for ad units that offer rewards ↗
DeconstructAlpha (析元) is a brand of Beijing Pengcheng Technology Co., Ltd., which operates this website.
京ICP备2026070261号-1 京公网安备11010802050788号