Your ROAS is solved. Your user quality isn't.

Two campaigns can hit the and send .
Traven tells you which , before the budget's gone.

Show me my wasted spendSee the proof
Traven · Campaign Quality
CampaignROAS d1–3Quality d1–3Forecast @ d30CQIVerdict

On ROAS alone, all five look scalable.Same early ROAS, opposite calls — quality splits them by day 3. Click a row to see the forecast.

If you buy users, Traven scores them.

We help game studiosfind players who actually spend.

Trusted by

0M+
users served across the platforms we power
~0%
of client UA spend flagged as underperforming
0 days
earlier than attribution settles

Hundreds of campaigns. One call each. Every day.

Every morning your team opens a board of live campaigns and makes the same decision on every one: cut it, hold it, or give it more budget, and exactly how much. Each campaign is graded on two scorecards at once. Your own ROAS, and the user-quality metrics your partners hold you to. Those two numbers pull against each other constantly.

ROAS will not settle the call, because ROAS converges. Run it out to day 30 and most campaigns land at roughly the same multiple no matter who they brought in. The thing that actually separates a good campaign from a bad one is user quality, whether these people play, deposit, and stay, and that signal stays hidden until attribution catches up weeks later. Long after the budget is gone.

So the call gets made on instinct, hundreds of times a day, on the one number that cannot tell two campaigns apart.

Campaign Board
0live campaigns
  • MA_Gamers_IAP_US2.3x
  • GG_Search_Brand_US2.2x
  • TT_Lookalike_Players_CA2.4x
  • AL_Rewarded_Players_Global2.1x
  • TT_Broad_UGC_Cashout_US2.3x
  • SN_Story_Cashback_UK2.2x
  • MA_Reels_Conversion_DE2.2x
  • GG_Display_RT_AU2.3x
  • AL_Interstitial_RPG_JP2.1x
  • TT_Spark_Boost_MX2.4x
  • SN_AR_Lens_Players_BR2.2x
  • GG_YouTube_Action_IN2.3x
  • MA_Adv_Plus_Shopping_FR2.2x
  • AL_Rewarded_Survey_PH2.3x

Same ROAS on every row. Quality all over the place. Now pick which ones to cut.

Same ROAS. Different users.

Two campaigns, same channel economics, opposite outcomes. Traven separates them by day 3.

TikTokTT_Broad_UGC_Cashout_US
Audience
Broad 18-34, US, lookalike off recent installers
Creative
High-energy UGC, "Earn $100 playing games" hook, 9:16
Spend
$42K/wk, CPI $0.91

Looks like a winner on cost. It is not.

Cumulative ROAS
0x1x2xd0d7d14d21d301.0x payback
User quality, first-deposit %
0%5%10%15%d0d7d14d21d30Traven predicts d30 here

Game-Studio ARPU @ d30, $0.41

MetaMA_Gamers_Interest_IAP_US
Audience
Interest-targeted mobile gamers, US, narrower
Creative
Static + carousel, real game screenshots, "Play & earn"
Spend
$38K/wk, CPI $2.40

Costs more per install. Pays back on the users that matter.

Cumulative ROAS
0x1x2xd0d7d14d21d301.0x payback
User quality, first-deposit %
0%5%10%15%d0d7d14d21d30Traven predicts d30 here

Game-Studio ARPU @ d30, $1.85

Traven verdict

TT_Broad_UGC_Cashout_US
ROAS ✓ Good
Quality ✗ Poor
Cut spend
MA_Gamers_Interest_IAP_US
ROAS ✓ Good
Quality ✓ Strong
Scale spend

ROAS said both were fine. Quality said otherwise, and Traven made the call by day 3, before another $120K went into the bleeder.

Your history in. One score out.

Training data

Your cohorts, your history.

Forecast accuracy0 cohorts ingested
60%75%90%MarAprMayJunJul

Sharper with every cohort.

Fit to your economics
Payback target
21 days
Margin
68%
Good user
$4.50 LTV
Partner quality metric
D7 first-deposit %

  • Trained on your own history.

    Traven fits a model to your own historical cohorts, the campaigns and users you have already run. Not a generic benchmark, not someone else's data. It learns how your users behave, who plays, who deposits, who churns.

  • Tuned to your economics.

    The model is built around your numbers. Your payback target, your margins, how your partners grade you, what a good user is actually worth to your business. Two companies on the same channels get two different models.

  • Sharper every cohort.

    Every new cohort makes your model more accurate. And as more teams run on Traven, anonymized cross-client patterns lift the floor for everyone. The longer you run, the better the calls.

Hundreds of campaigns. One number each.

Instead of weighing ROAS against partner quality across a few hundred live campaigns by gut, your team opens one view. The Campaign Quality Index is a single 0 to 100 score per campaign that folds ROAS and user quality (first deposits, in-app spend, engagement) into one number, color-coded for the call. The model decides how much ROAS should count against quality, so you get a ranked, decision-ready list instead of a spreadsheet to argue over.

Campaign Quality Indexlive, 6 of 248
  1. 0
    MA_Gamers_IAP_US
    Scale
  2. 0
    GG_Search_Brand_US
    Scale
  3. 0
    TT_Lookalike_Players_CA
    Hold
  4. 0
    AL_Rewarded_Players_Global
    Hold
  5. 0
    TT_Broad_UGC_Cashout_US
    Cut
  6. 0
    SN_Story_Cashback_UK
    Cut

Works with the stack you already run.

Traven reads from where your data already lives.

ConnectionsRead-only access
  • SnowflakeWarehouseRead-only
  • AppsFlyerAttribution / MMPRead-only
  • Meta AdsAd platformRead-only
  • AmplitudeAnalyticsRead-only
OutputCampaign Quality Index, one score per campaign

No SDK to ship. No new tracking. Traven reads from the warehouse and MMP you already run, and writes nothing into your app.

Acquisition
Snap
Reddit
Pinterest
Attribution / MMP
Warehouse
Analytics / CDP
GA4
BI