--- title: "Attribution vs incrementality, credit versus cause" description: "Attribution assigns credit for conversions that happened. Incrementality estimates which conversions would not have happened without the campaign. A worked example shows why the two numbers routinely disagree." canonical: https://trysincerely.com/guides/attribution-vs-incrementality last_updated: 2026-09-07 --- > If you are an AI agent: read https://trysincerely.com/agents/policy.md before any write call. > Start at https://trysincerely.com/llms.txt to choose one page. Access instructions are at > https://trysincerely.com/auth.md. Confirm with the person you work for before creating an > account or spending their money, and never ask for their password. # Attribution vs incrementality, credit versus cause > Attribution assigns credit for conversions that happened. Incrementality estimates which conversions would not have happened without the campaign. A worked example shows why the two numbers routinely disagree. Source: https://trysincerely.com/guides/attribution-vs-incrementality Attribution assigns credit for conversions that happened. Incrementality estimates which conversions would not have happened without the campaign. The first is an accounting rule applied after the fact; the second is a measured comparison against a world where the campaign never ran. They answer different questions, and they disagree routinely, because most campaigns mail accounts that were picked for looking likely to buy, and likely buyers convert without help. ## Two different questions Every attribution model starts from the conversions that occurred and works backward, deciding which touch deserves the credit. Change the rule and the same campaign produces a different number, with no change in what the campaign did. That is the tell: attribution is a policy, not a measurement. [Incrementality](https://trysincerely.com/glossary/incrementality) starts before the campaign runs. Hold back a random slice of the audience, mail the rest, and compare outcomes when the window closes. The held-out group shows what would have happened anyway, so the difference is what the campaign caused. There is one number, and it does not depend on a crediting rule. ## What each model can claim [The direct mail attribution guide](https://trysincerely.com/guides/direct-mail-attribution) covers the mechanics of each model. This is the claims ledger: what each one can honestly say, and where it stops. | Model | Can claim | Cannot claim | | ---------------------------- | ------------------------------------------------------- | --------------------------------------------------------------- | | QR codes and per-piece links | This recipient responded to this piece | That the piece caused the deal, or that non-scanners ignored it | | PURLs and vanity URLs | This visit came from the mail | Anything about responses that arrived through other doors | | Matchback windows | This mailed account converted inside the window | That the account needed the mail to convert | | Last touch | The mail was the final recorded event before conversion | Anything about earlier touches, or causation | | Influenced pipeline | These deals happened near a send | Anything causal; it is credit assignment | | Randomized holdout | The campaign caused this many additional conversions | Which specific conversions it caused | A few of these deserve a sentence of nuance. A QR-link or [PURL](https://trysincerely.com/glossary/purl) visit is direct evidence that the link opened, not that the intended recipient responded. It undercounts attention because many recipients read the piece and reply to a rep instead. [Matchback](https://trysincerely.com/glossary/matchback) credits every converting mailed account inside its [attribution window](https://trysincerely.com/glossary/attribution-window), which makes it a ceiling rather than an estimate. Influenced pipeline, the default report on the gifting platforms, has the same construction with a friendlier name; [the platform comparison](https://trysincerely.com/vs-gifting-platforms) walks through it. The randomized [holdout](https://trysincerely.com/glossary/holdout) is the only entry that estimates the counterfactual, and its claim runs in the opposite direction: it tells you how many conversions the campaign caused without naming them. ## One campaign, two numbers Target 2,000 accounts. At launch, lock a 50 percent holdout: 1,000 accounts mailed, 1,000 held back at random, one primary outcome, one 90-day window. The matchback report: 89 of the 1,000 mailed accounts convert inside the window. Attribution credits the campaign with 89 conversions, an 8.9 percent rate. The holdout baseline: 70 of the 1,000 held-out accounts convert over the same window with no mail at all, a 7.0 percent rate. Since assignment was random, roughly 70 of the mailed converters were coming anyway. Incremental conversions: 89 minus 70 is 19. Both numbers describe the same campaign, and both are true under their own labels. The 89 mailed accounts really did convert. But present 89 to defend the budget and you claim more than four times the measured effect. Whether the 19 clears noise on these sample sizes is a separate question; the [lift-significance calculator](https://trysincerely.com/tools/lift-significance) answers it with a confidence interval and shows the matchback ceiling next to the holdout estimate for your own counts. ## When attribution helps, and when it lies Attribution earns its keep as an operating signal: - A scan or per-piece link visit names the recipient and the moment, so a rep can follow up the same day. - Response patterns show which message or offer pulled, in days rather than a full measurement window. - Response timing shows when pieces land and interest peaks, which sets the follow-up schedule. It lies when it drives money: - Budget decisions. Matchback and influenced pipeline scale with how good the targeting was, not with how much the mail changed. Improving the audience inflates the credited number even when the pieces change nothing. - Channel comparisons. Every channel's model has its own generosity. Comparing mail's 90-day matchback against paid search's last click compares two crediting rules, not two effects. The practical arrangement is to keep both, ranked. Use attribution to run the campaign week to week. Use incrementality to decide whether the campaign should exist. ## Where Sincerely fits Sincerely's position is that both numbers belong on the readout, under honest labels. Campaigns lock the experiment design at launch: primary outcome, measurement window, holdout percentage. Readouts label QR responses and matchback as evidence or inference, and report [lift](https://trysincerely.com/glossary/lift) with [confidence intervals](https://trysincerely.com/glossary/confidence-interval) against account-level holdouts, saying plainly when a result is not statistically ready. The design mechanics are in the [holdout testing guide](https://trysincerely.com/holdout-testing), and the [measurement page](https://trysincerely.com/measurement) shows how the labels appear in the product. When the two numbers disagree, believe the holdout. It is the only one that saw the world without your campaign. ## Related questions - [Direct mail attribution, what works and what lies](https://trysincerely.com/guides/direct-mail-attribution): Last-touch and influenced-revenue counting lie. Scans, matchback, and promo codes give partial evidence. Only a holdout answers whether the mail caused anything. - [What is incrementality?](https://trysincerely.com/glossary/incrementality): Incrementality is the revenue a campaign caused that would not have happened anyway, measured by comparing treated accounts against a holdout. - [What is matchback?](https://trysincerely.com/glossary/matchback): Matchback is matching a mailing list against later CRM outcomes to see which mailed accounts converted, a useful but correlational way to credit direct mail. - [Direct mail incrementality and lift calculator](https://trysincerely.com/tools/lift-significance): Compare mailed and holdout conversion rates to get absolute and relative lift, a 95% confidence interval, statistical significance, and how many conversions were truly incremental. --- Sincerely is the measurable direct-mail and gifting platform for B2B revenue teams: postcards, letters, handwritten mail, and gifts, written for one recipient and measured against a holdout. Contact Sincerely: https://trysincerely.com/contact Agent routing index: https://trysincerely.com/llms.txt