Attribution tells you who touched a conversion. Incrementality tells you whether touching it mattered. Trcker runs live holdout tests per partner with full statistical rigor — so you stop paying for conversions that would have happened anyway.
Randomized holdouts are what Google, Meta, and every data-driven brand use internally. Most tracking platforms don't ship it. Trcker does — one click per partner, and it's running.
From Reports / Incrementality, click 'Start a holdout.' Pick a partner, optionally scope to a specific offer, set the percentage (10% is the default). Trcker starts diverting that slice of traffic immediately.
A deterministic hash of each click decides whether it lands in the treated group or the held-out control. The held-out clicks still redirect to your offer normally — the visitor's experience is unchanged. Only the attribution flag differs.
Every night at 07:00 UTC, Trcker computes treated CVR vs holdout CVR per partner, runs a two-proportion z-test, and bounds the lift with a Wald 95% confidence interval. Results appear on the Incrementality dashboard the next morning.
Three summary cards show total incremental revenue across significant partners, revenue at risk from partners with near-zero lift, and total measured treatment revenue. Per-partner horizontal bars show what you paid vs what the partner actually drove.
Every partner with an active holdout gets a significance badge: Incremental (green), Not incremental (red), Not significant (grey), or Pending. The badge plus the dollar-impact chart tells you exactly where to cut, where to double down, and where to wait.
Attribution assigns credit for conversions that happened. Incrementality asks whether the credit is deserved. A partner can touch 20% of conversions (attribution) but only cause 2% of them (incrementality). The gap is where overpayment lives.
Yes — that's the measurement cost. A 10% holdout on a $100K/month partner means about $10K of potential commissions flow into the control group. In exchange, you find out whether the other $90K is real. If the answer is 'mostly happens anyway,' you save the full spend going forward. It's cheap insurance.
Depends on your baseline CVR. At 3% baseline CVR, you need about 45,000 clicks to detect a 10% lift at 95% confidence. Below that threshold, lift numbers are directional but not statistically proven. Trcker shows 'pending' for low-volume partners rather than faking a significant result.
No. Held-out clicks redirect to your offer normally. The partner sees the same click count they always did. Only the attribution and conversion-rate calculation differ. This is what makes the measurement valid — if partners could see the holdout, they'd route their best traffic away from it.
Pause, renegotiate, or restructure. A partner driving zero real incremental conversions at your payout rate is the same as paying for organic traffic. Trcker's Autopilot reference agent can pause partners with statistically significant non-positive lift automatically — or you can review manually each month.
Yes. The get_incrementality_lift MCP tool returns per-partner treated/holdout CVRs, lift percentage, Wald 95% CI, and p-value. Filter to significantOnly=true to get just the rows worth acting on.
Want a deeper dive? Read the Incrementality guide in the docs, or start with what incrementality is and how a holdout test works.
Incrementality holdouts are included on every plan. Start one in 60 seconds from the dashboard.
Request Access →