> ## Documentation Index
> Fetch the complete documentation index at: https://docs.fleack.io/llms.txt
> Use this file to discover all available pages before exploring further.

# A/B test interstitial ad frequency without SDK changes

> Test three interstitial ad frequencies on your mobile game and find the cadence that balances D7 revenue and retention in about 30 minutes.

Interstitial ad frequency is the single most-requested test from mobile gaming teams — and for good reason. Show ads too often and you earn more per session but drive players away; show them too rarely and you leave revenue on the table. There is no theoretical right answer. Fleack lets you run a controlled experiment on the actual backend config your app already fetches, with no SDK changes, no binary update, and no risk of breaking existing sessions.

## What you'll measure

| Variant       | Interstitial frequency                  | Hypothesis                                                                     |
| ------------- | --------------------------------------- | ------------------------------------------------------------------------------ |
| **Control**   | Whatever your backend currently returns | Baseline.                                                                      |
| **Variant A** | +30% (more frequent)                    | Higher ad revenue per session; may hurt D7 retention.                          |
| **Variant B** | −30% (less frequent)                    | Lower revenue per session; may improve D7 retention and lift long-term ARPDAU. |

**Primary metric:** D7 retention (binary)

**Secondary metric:** ARPDAU delta at D7 (continuous scalar)

## Pre-flight check

Before creating any levers, confirm that your backend exposes interstitial frequency as a discrete field in a `config-candidate` endpoint. Open the Fleack backoffice, navigate to **Endpoints**, and look for your monetization config endpoint. Click into its body sample and scan for paths like:

* `data.ads.interstitial_freq`
* `data.monetization.interstitial.frequency`
* `ads.interstitial.cooldown_seconds`

<Warning>
  If the frequency value is computed client-side from a level number or hardcoded in the binary, you cannot test it with Fleack yet. Move the value to a backend config endpoint first, then return here.
</Warning>

The endpoint's classification must read **`config-candidate`** — the same response goes to every user, which is what makes a lever testable. If it reads `user-data` or `mixed`, check whether you're looking at the right endpoint.

## Main workflow

<Steps>
  <Step title="Declare or confirm the lever">
    Open the **Levers** page. If Fleack's AI enrichment already detected an interstitial frequency lever, it will appear under a label like "Frequency" or "Interstitial cadence". Click it, verify the path and current value match your reality, and edit the label if needed.

    If no lever exists yet, click **+ New lever**:

    1. Pick the endpoint that returns your monetization config.
    2. Use the path picker to search for `interstitial` and click the matching path.
    3. Fill in the lever details:
       * **Label**: `Interstitial frequency`
       * **Type**: `frequency`
       * **Description**: e.g. `Number of rounds between interstitial ads.`
       * **Test suggestions**: values reasonable for your genre, such as `3, 4, 5, 6, 8`

    The lever now appears in the catalog with the live backend value shown alongside it.
  </Step>

  <Step title="Set up the test">
    From the lever detail page, click **Test**.

    Configure the test as follows:

    * **Variant A value**: current value × 1.3, rounded to the nearest integer.
    * **Variant B value**: current value × 0.7, rounded to the nearest integer.
    * **Allocation**: 33% / 33% / 34% (control gets the remainder automatically).
    * **Segment**: leave at **All users** for your first run, or scope to Tier-1 countries if you want a cleaner monetization read.
    * **Primary metric**: **Retention day 7** — select the endpoint your app calls on session start (e.g. `GET /api/get_appsettings` or `GET /api/session/start`).
    * **Secondary metric**: **Scalar delta on `arpu`**, observation window 7 days. This requires profile attributes from your `user-data` endpoints — see [Segments](/core-concepts/segments) for how Fleack builds those.

    Click **Launch**. The test goes live immediately on every shipped app version.
  </Step>

  <Step title="Watch the results">
    The test detail page updates in real time. For a mobile game with 50K+ DAU you can expect:

    * First exposure rows within minutes of launch
    * 1,000+ exposures per variant within an hour
    * An early D2–D3 retention signal within 2–3 days
    * A statistically meaningful D7 read between days 7 and 14

    Per-variant panels show exposure counts, the running D7 retention rate (marked **not yet eligible** for users exposed fewer than 7 days ago), and Bayesian win probability vs control once each variant accumulates 30+ exposures.

    <Info>
      Early D2 reads are useful for spotting a disaster (e.g. Variant A tanking retention by 15+ points) but are not grounds for promotion. Wait for the D7 window.
    </Info>
  </Step>

  <Step title="Make the call">
    Use these thresholds to decide when to act:

    | Verdict     | Condition                                                  | Action                 |
    | ----------- | ---------------------------------------------------------- | ---------------------- |
    | **Promote** | Win probability vs control ≥ 90% **AND** ≥ 14 days of data | Click **Promote**      |
    | **Stop**    | Win probability falls below 10%                            | Click **Stop variant** |
    | **Wait**    | No clear difference yet                                    | Give it more exposures |

    <Note>
      D7 retention requires the full 7-day window per user, so the earliest you can have a clean read is day 8 of the test — and you need enough cohorts to stabilise variance, hence the 14-day minimum.
    </Note>

    A common outcome: Variant A (more ads) wins on per-session revenue but loses on D7 retention; Variant B (fewer ads) is flat on retention but up on long-term ARPDAU. Which variant is the "real winner" depends on your LTV model — that's a business decision, not a statistics question.
  </Step>

  <Step title="Promote the winner">
    From the test detail page, click **Promote** on the winning variant. Fleack immediately routes 100% of traffic to that value, moves the test to `completed`, and updates the lever's effective value in the catalog.

    If you prefer a staged rollout — say 50% first, then 100% — pause the test and adjust the allocation manually instead of using one-click promote.
  </Step>
</Steps>

## Common pitfalls

* **Don't test before D2 retention is healthy.** If your game has a 30% D2 drop-off baseline, ad frequency tests will produce noisy, uninterpretable swings. Fix the funnel first; test the cadence second.
* **Watch for genre confound.** Hypercasual players tolerate more interstitials than mid-core players. A cadence that wins on a hypercasual title does not automatically win on a mid-core one — run the test on each title separately.
* **Account for ad fill rate, not just frequency.** Increasing the number of interstitial slots doesn't automatically increase revenue if your mediation partner's fill rate at higher slot counts drops below 70%. Check the fill report before declaring victory.

***

<CardGroup cols={2}>
  <Card title="A/B test in-app pricing" icon="tag" href="/recipes/ab-test-pricing">
    Test bundle composition and bonus content for a fixed-price IAP without touching the SKU tier.
  </Card>

  <Card title="A/B test an onboarding flow" icon="graduation-cap" href="/recipes/ab-test-onboarding-flow">
    Reorder and adjust onboarding steps to lift D1 retention on new installs.
  </Card>
</CardGroup>