first_touchFirst touch
All credit goes to the earliest eligible touch.
X-Ray methodology
Attribution you can inspect.
Models change how credit is allocated. They do not rewrite the journey, upgrade weak evidence, or turn a platform total into a person-level fact.
The seven observed models
Run each model over the same eligible evidence and date/window configuration. A model chooses a credit rule; it does not add touches that were not observed.
last_touchLast touch
All credit goes to the final eligible touch.
last_non_directLast non-direct
Credit goes to the latest eligible touch that was not direct.
linearLinear
Credit is shared equally across all eligible touches.
time_decayTime decay
Later touches receive more credit according to the selected half-life.
position_basedPosition based
40% goes to the first touch, 40% to the last, and 20% is shared across the middle.
custom_weightedCustom weighted
The operator supplies explicit channel weights; X-Ray normalizes only within the governed model contract.
Windows and evidence
Click and view windows are independent. Time decay adds a half-life inside the eligible window. The report discloses the chosen settings so comparisons use the same evidence.
Confirmed
Touches joined through consent-eligible, deterministic Site evidence can receive observed attribution credit.
Candidate
Probabilistic or incomplete associations remain assists. They do not establish identity or qualify downstream activation.
Same-evidence comparison
Change the model while retaining the same journey set, windows, outcome, and filters to isolate the effect of the allocation rule.
Observed plus reported influence
This optional projection combines two visible ledgers. It is not an eighth observed model and never inserts a survey answer into the touch table.
Observed share
The selected observed model is scaled proportionally by an explicit nonnegative integer weight.
Reported share
The explicit reported weight is split equally across the response's distinct selected options using deterministic largest remainder.
The weights must be explicit
Observed and reported weights must total exactly 1,000,000 micros. There is no hidden default. An unresponded conversion remains 100% observed. Credit and minor-unit value are conserved per conversion and per currency; X-Ray does not fabricate cross-currency totals.
View-through stays unavailable without a receipt.
A platform's aggregate impression or attributed-conversion count cannot prove that a particular Site subject saw an ad. X-Ray enables a view touch only when an event-level provider receipt supplies an eligible subject join.
| Provider | Current state | What remains usable |
|---|---|---|
| Meta | Unavailable | Aggregate provider comparison only; no person-level view touch. |
| TikTok | Unavailable | Aggregate provider comparison only; no person-level view touch. |
| Unavailable | Aggregate provider comparison only; no person-level view touch. | |
| AppLovin | Unavailable | Aggregate provider comparison only; no person-level view touch. |
| Unavailable | Aggregate provider comparison only; no person-level view touch. |
Provider totals are comparisons, not corrections.
Ad platforms use their own identities, windows, models, and reporting delays. X-Ray can place those totals beside its observed report, but it does not silently make either ledger match the other.