If you run UA for a mobile game on TikTok, you have probably noticed that the platform’s most mature optimization engine was not built for you. It was built for e-commerce: people who see an ad, land on a website and buy something. TikTok has had years of purchase data from those flows, and its web-based Sales campaigns are very good at finding people who pay.
App campaigns are a different story. You point the ad at the store, you optimize for installs or a handful of in-app events, and on iOS you work inside SKAN’s delays and aggregation. The algorithm learns, but it learns slowly and from blurry signals.
The classic workaround is web-to-app. Run a web campaign, send people to a landing page, then push them on to the store. It works, but you are now maintaining a landing page whose only job is to exist, and every extra tap costs you users. In this article we explain how web-to-app works on TikTok when the landing page disappears, what you actually unlock, which signals you should feed it, and how to test it without betting the whole budget.
TL;DR
TikTok’s web-based Sales campaigns can optimize toward purchases and value, but they normally require a web destination. With Audiencelab, a mobile game can unlock that campaign type and send users straight to the App Store or Google Play, without building or hosting a landing page or bolting on a separate web2app tool. The campaign type is only half the story, though. It performs when you feed TikTok an engineered value signal (for example IAP and ad revenue blended) through the Events API. Start with a controlled pilot against your current app campaign baseline and judge it on cohort value, not CPI.
What web-to-app on TikTok actually means
The industry uses “web2app” for a few different things, so let’s be precise.
TikTok web-to-app lets you run mobile game acquisition campaigns through TikTok’s web campaign setup. For example, you can use Sales with web conversion optimization instead of standard App Promotion. Users still install your game through the app store. The goal is to access web-side optimization and value signals throughout the install funnel.
Most advertisers use a web funnel between the ad and the app store. This usually involves a landing page, quiz, or web paywall. Subscription apps use this approach frequently, often collecting payments outside the app store.
Mobile games rarely need this extra step. Nobody wants to read about your merge game before downloading it. The landing page exists because the campaign setup requires a website. Every additional step creates another opportunity for potential players to drop off.
Audiencelab takes a different approach to web-to-app for mobile games. You access TikTok’s web campaign setup without building a web funnel. Audiencelab provides the destination and redirect infrastructure needed to run the campaign. Players go straight from your TikTok ad to the App Store or Google Play listing. This also supports Custom Product Pages and Custom Store Listings where applicable. You don’t need to build or host a landing page. You also don’t need an external web-to-app service to connect the flow.
In short, the three paths look like this:
- App Promotion campaign: Ad → store → install. A short path, but optimization runs on app events and, on iOS, SKAN postbacks.
- Classic web2app: Ad → landing page → store → install. Web optimization, plus an extra step that leaks users and needs constant maintenance.
- Web2app with Audiencelab: Ad → store → install. A short path with web-side optimization, fed by in-app events that Audiencelab sends back to TikTok.
The Audiencelab path is as short as App Promotion but runs on web-side optimization.
Why the campaign type matters more than the redirect
Removing the landing page is nice, and your users will thank you by not bouncing. However, the real reason to do this is what the campaign type gives you access to.
You get an optimization engine trained on buyers
TikTok’s web Sales setup is designed to find people who purchase. It learns from purchase events across a huge volume of advertisers. When your game uses this setup, TikTok optimizes toward your chosen conversion event and value. It doesn’t just target people likely to tap “Get”. This matters when your app campaigns attract cheap installs that never open the shop.
You decide what “conversion” means
On the web path, TikTok receives conversion signals through its Events API. Audiencelab controls which events and values go into those signals. You can send signals that reflect what makes a player valuable to your game. These might include a first purchase, revenue value, retention milestone, or blended IAP and ad revenue.
You are not waiting on SKAN to learn
The web-first path does not depend on SKAdNetwork postbacks for its optimization signal. Your MMP may still use Apple’s attribution elsewhere, and ATT obligations still apply to your app. That said, the campaign itself learns from events you send through the Events API. This means faster and more granular feedback than delayed, aggregated postbacks.
The catch: a better engine still needs better fuel
Here is where many web2app tests quietly fail. Teams switch campaign type, keep sending the same weak event, and conclude that “web2app didn’t work on TikTok”. If the only thing you report is an install or a tutorial start, you have given a purchase-hunting algorithm nothing to hunt.
A useful conversion signal balances four things: how strongly it correlates with long-term value, how often it fires, how quickly it fires after install, and how reliably it is measured. Further downstream is not automatically better. A whale purchase on day 14 is valuable, but it is too rare and too late to train anything.
For example, a hybrid-casual puzzle game might find that very few players purchase in the first 48 hours, but many watch rewarded ads. Optimizing for first purchase alone would starve the campaign of events. A blended value signal (IAP revenue plus ad revenue within the first days) gives TikTok more events and a more honest picture of who is worth paying for. A narrative game with strong early IAP, on the other hand, might do fine with purchase value alone.
The goal is to find the one that tells TikTok who your valuable players are, early and often enough for it to act on.
What Nanobit’s TikTok results tell us about signals
The best evidence that signal quality changes TikTok outcomes comes from a case TikTok published itself. Nanobit, the studio behind Unfolded: Webtoon Stories, worked with Audiencelab on TikTok campaigns aimed at higher ROAS, IAP and ad revenue.
To be clear, this was not a landing-page-free web2app campaign. It ran on TikTok’s Purchase Optimization campaigns with the TikTok Events API, using Audiencelab’s signal infrastructure. It is relevant here because it runs on the same foundation the web2app setup depends on: an engineered value signal delivered through the Events API.
Two things drove it, according to TikTok’s case study:
- Signal engineering: instead of optimizing for installs, the signal blended in-app purchase and ad revenue data, so TikTok learned from total player value.
- Creative diversity: Nanobit used TikTok One creators with structured direction and creative freedom, which gave the campaign a wide pool of hooks and angles to test.
TikTok reported a total ROAS of 2.08, a Day-7 ROAS of 1.68 and a CPI of $1.13 for the campaign. These are one studio’s results in one setup, not a benchmark you should expect to copy. The lesson is more transferable than the numbers: when TikTok can see value, it goes looking for players who generate value, and creative variety gives it more ways to find them.
The web2app campaign type gives that same signal access to TikTok’s purchase-focused web optimization, without the landing page tax. Whether that beats your current app campaign is something you test, which brings us to the practical part.
How to set up and test web-to-app on TikTok
Treat this as an experiment with a baseline, not a migration. In practice, a sensible first run looks like this:
- Declare your baseline. Pick the current TikTok app campaign you want to beat, and write down the metric you will judge it on before you launch. D7 ROAS or cohort revenue per install beats CPI here. If you decide the metric afterwards, you will pick whichever one makes your favourite result look good. We all do it.
- Map your funnel and pick the signal. Look at your own data. Which early event or value correlates with D7 or D30 value, and how many times a day does it fire at your planned budget? If the answer is “a handful”, go one step earlier or blend values until TikTok has enough events to learn from.
- Connect the data. Audiencelab gets in-app events through its SDKs (iOS, Android, Unity) or server-to-server. You can keep your MMP for governance and broad reporting. Audiencelab sits beside it and handles the web2app path and the Events API delivery for TikTok.
- Choose the destination. Send users to your default store page, or to a Custom Product Page or Custom Store Listing that matches the creative. The ad creates the expectation, and the store page needs to confirm it. A direct path to a mismatched page just helps people bounce faster.
- Launch a contained pilot. Run it on a slice of budget in one or two geos, with enough spend to exit learning. Keep creatives comparable to the baseline so you are testing the campaign type and signal, not a new hook.
- Judge mature cohorts. Wait until the cohorts have matured to your chosen window before calling it. Early CPI on a value-optimized campaign is often higher, and that is fine if the players are worth more.
What to measure
| Metric | Why it matters | What to watch for |
| D7 / D30 ROAS by cohort | The real verdict on player value | Compare against the declared baseline, same window |
| Optimization event volume per day | Whether TikTok has enough to learn from | Campaign stuck in learning = signal too rare |
| Signal latency | How fast TikTok sees value | Values arriving days late weaken optimization |
| Store page CVR | Whether the promise survives the click | Drop after switching to direct redirect = page mismatch |
| Creative-level value | Which hooks bring payers, not just clickers | High CTR, low value creatives to cut |
| CPI | Context, not the verdict | Higher CPI can still win on ROAS |
Higher CPI can still win on ROAS
One more practical note: keep the MMP view and the Audiencelab view side by side, and agree upfront which source decides the test. Numbers will differ because windows and methods differ. That is normal, and arguing about it after launch is how a two-week test becomes a two-month debate.
Conclusion: stop paying the landing page tax
For mobile games, the landing page in a web2app flow has mostly been a toll booth. You paid it in drop-off and maintenance so you could get into TikTok’s better optimization lane. Removing it is a nice win for the user experience, but it is not the main event.
The main event is getting a purchase-hunting algorithm and a value signal that actually describes your best players into the same room. Nanobit’s TikTok results show what happens when TikTok can see value. The web2app setup gives that value access to a campaign type built around it.
So test it properly: declare a baseline, engineer a signal with enough volume, send people straight to a store page that matches the ad, and judge the result on mature cohort value. If it wins, scale it. If it doesn’t, you have learned something specific about your signal or your funnel, which is still more than “TikTok is weird this month”.
If you want to see how the TikTok web2app setup would work for your game, let’s talk!
