CTV for Puzzle Games: How to Run a Test You Can Actually Read

Most CTV tests for puzzle games fail in the reporting, not in the market. The campaign runs, the CTV line in the dashboard shows a number that looks small next to what was spent, and somebody kills it in week three. The channel may have been working the whole time.

That happens because CTV is credited differently from every other line in the same report, and most test plans never account for it. Here is how to build one that survives contact with the numbers.

How a TV impression gets connected to a phone

Start here, because every measurement decision downstream depends on it.

A television and a phone share no advertising identifier. No click carries a device ID, and nothing in the ad hands the phone a token to redeem later. The household IP address connects the two, matched within a short window after the impression.

Three consequences follow, and all three change how you read a test:

The match is to a home, not a person. Multiple installs from the same household can be credited to one impression. The player whose behavior you are about to measure may not be the person who watched the ad.

The window does more to your numbers than your creative does. Cross-device view-through windows are typically configured somewhere between one and three days, and the default is often around 24 hours. Move it and your reported CTV volume moves with it, with nothing on screen explaining why.

The evidence ranks low. Attribution waterfalls order engagements by how reliable they are, with clicks ahead of impressions and verified device matches ahead of probabilistic ones. A probabilistic CTV impression sits near the bottom, one rung above organic. When it loses, the install does not vanish. It gets credited somewhere else, often to organic.

That last point is why the dashboard reads low. Our post on why CTV installs arrive late and land in organic walks through the decay in detail.

On iOS, it is stricter than that. Neither SKAdNetwork nor AdAttributionKit provides a direct way to attribute a CTV impression to an app install. Apple's frameworks were built for app-to-app and web-to-app paths. Your iOS CTV read is a modeled or blended one, and any platform describing it otherwise is selling you modeled output as measurement.

Define the game more precisely than its category

Genre is where a media plan starts, not where the creative brief ends. Write down the puzzle mechanic, the progression system and the monetization model before anyone opens an edit timeline.

A match-based board, a word game and a logic challenge give a creative team three different things to show. A board game can demonstrate a satisfying cascade in four seconds. A word game usually cannot, and forcing it into the same creative template produces an ad that describes the game instead of showing it.

Build an audience hypothesis out of your own player data rather than a category stereotype. "Puzzle players watch cooking shows" is a hypothesis with a cost attached. Test it against a broad buy before you let it narrow your inventory.

Show a puzzle the player can actually play

Pick one recognizable decision from the game and show what happens after it.

Then watch the ad and the first session back-to-back. Does a new player reach that kind of puzzle in the first two minutes, or does the ad promise a board they meet on level 40? A mismatch here shows up as a retention problem that looks like a media problem, and no amount of targeting work will fix it.

Keep the store listing consistent with what the ad showed. This matters more on CTV than anywhere else, because the viewer cannot tap. They pick up a phone and search, and they search for what they remember seeing. If the icon, the name, and the first screenshot do not match the ad they just watched, that intent leaks to whatever the store surfaces instead.

Worth testing: two openings cut from the same footage, one posing the hard move and one paying it off. That comparison isolates something real, because the gameplay is held constant and only the emotional beat changes.

Choose player-quality measures for your business model

Installs are a top-funnel event and carry no information about whether the player liked the game. Decide up front which early outcome means something to your business, and instrument it before the first dollar goes out.

For an ad-funded game, measure ad revenue per user. Impressions per user will rise whenever you show more ads, which tells you about your placement density and nothing about the cohort.

For a purchase-led game, define payer conversion and fix the revenue window in advance.

For a game that does both, report the two separately. The double-counting risk people warn about comes from running a mediation platform alongside direct network integrations and ingesting both, not from hybrid monetization itself. The error that actually distorts hybrid reads is comparing ad revenue reported net against in-app purchase revenue reported gross, which makes purchases look more profitable than they are.

One CTV-specific caution on all of this. Because the match is to a household, a CTV-attributed cohort is a slightly different population from a click-attributed one. Compare it against a matched cohort from another channel rather than against your blended average, or the household effect will read as a quality difference.

Use comparable cohort ages. A cohort with incomplete retention or revenue data is immature, and immature cohorts shouldn't be compared to mature ones.

Separate the credit question from the growth question

These are two different questions, and only one of them is answerable from a dashboard.

Which touchpoint gets credit is an attribution question. Report the window and the method alongside the result, because a number produced under a 72-hour window and a number produced under a six-hour window are not the same measurement.

Did the advertising produce additional players is an incrementality question, and attributed installs cannot answer it. A geo holdout can: suppress CTV in matched markets, run it in others, and compare total installs with organic included. It never asks which touchpoint deserves credit, so the household-match problem and the organic-spillover problem both stop mattering. Our post on the three things to look at when measuring CTV covers the holdout design, the time-to-install curve, and how CTV moves blended CPI.

Here is the position worth stating plainly: if your decision rule reads the CTV line in the dashboard, you have already decided to turn CTV off. Write the rule against total installs and organic movement instead.

Write the decision rule before you spend

Record what evidence would make you continue, revise, or stop, and record it before the campaign starts.

Budget and duration come from a power calculation based on your own history, not a number a vendor quotes you. What you can fix in advance is the read date, and on iOS the arithmetic is unforgiving. AdAttributionKit's third conversion window closes 35 days after a user's first launch, and its postback arrives up to 144 hours after that. The measurement tail runs roughly six weeks past your last install, on top of the flight itself. A test read at day 14 is reading your reward curve, not your players.

Then log every other change during the window: onboarding, difficulty balance, store creative, promotions, and spend in other channels. If three things move at once, a before-and-after comparison cannot tell you which one did it, and you will have bought an expensive anecdote.

Puzzle-game pilot checklist

• Document the puzzle mechanic and confirm the advertised experience matches the first session.
• Write the audience hypothesis down, with the first-party evidence behind it.
• Clear gameplay footage, music, and artwork rights before the cut is locked.
• Check the store listing against the ad for icon, name, and first screenshot.
• Define the retention, player-quality, and revenue events, and instrument them.
• Record the view-through window and the attribution method you are running under.
• Set the read date from the slowest platform in the test, not the fastest.
• Agree on the decision thresholds and who owns the call.
• Log concurrent product and marketing changes as they happen.

FAQ

How is a CTV ad connected to a mobile install?

Through the household IP address, matched within a configurable window after the impression. There is no shared advertising identifier between a television and a phone, so the match is probabilistic and lands at household level rather than device level. Multiple installs in one home can be credited to a single impression.

Why does CTV look weak in my dashboard?

Because probabilistic impression matches sit near the bottom of the attribution waterfall. When the match fails or the window closes first, the install is still there, credited to another channel or logged as organic. Read total installs and organic movement, not the CTV row.

Can you attribute CTV on iOS?

Not directly. Neither SKAdNetwork nor AdAttributionKit supports CTV impressions as an attribution source, so any iOS CTV number is modeled. Plan the iOS read around a geo holdout rather than a per-campaign attribution figure.

Does every puzzle subgenre need its own strategy?

Only where the mechanic changes the creative or the success metric. A word game and a match-based board demonstrate differently in four seconds, which is a real difference. A separate campaign that differs only by category label will not teach you anything.

How long should a CTV test run?

Longer than the flight. On iOS, the third conversion window closes 35 days after first launch, and the postback follows up to 144 hours later, so the measurement tail runs about six weeks past your last install. Size the spend from a power calculation on your own conversion volume.

Explore Edge226's gaming solutions and CTV offering, or book a demo to talk through a test design for your game.