Skip to content
trackrift
Product overviewAttribution models

Attribution math you can audit.

Seven models, all documented, all auditable. We publish the formulas, the assumptions, and the failure modes. No "Scientific" black box.

The seven models we offer

Every trackrift account can run any of these models simultaneously and compare. There is no "default" we hide from you — you choose.

  • Last-click — credit to the last touch before conversion
  • First-click — credit to the first ad click in the journey
  • Linear — equal credit across all touches
  • Time-decay — exponential decay toward the conversion
  • Position-based (U-shaped) — 40/40/20 first/last/middle
  • Markov chain — counterfactual removal of each channel
  • Shapley value — game-theoretic fair credit across channels

Model cards (the anti-black-box bit)

Every model ships with a model card in our documentation center (/docs): the mathematical formulation, the assumptions, the input data shape, the known failure modes, and a worked example with synthetic data. If we cannot explain it to a CMO, we will not ship it.

Lookback window controls

Set the lookback per model. Default is 30 days. Configurable from 1 day to 180 days. Per-channel overrides supported (e.g., Meta 28 days, Google 14 days, organic 90 days).

Why we publish the math

Hyros calls its model "Scientific" and won't define it. Cometly's docs are sparse. We think you deserve better. If you're paying for attribution math that drives bidding decisions, you should be able to read the formula, replicate it in a spreadsheet, and decide whether you trust it. That's the whole point of attribution: it's a story we tell about ad performance. We owe it to you to make the story checkable.

Stop scaling blind. Start scaling what closes.

From $249/mo. Live in 30 minutes. 30-day money-back guarantee.