--- title: "What is marketing attribution? | daily.dev Ads glossary" url: https://business.daily.dev/glossary/marketing-attribution/ description: "Marketing attribution is deciding which touches get credit for a result. Last-click gives it all to the final visit, multi-touch spreads it, view-through credits an ad seen but not clicked, and incrementality measures what the campaign caused." lastUpdated: "2026-10-20" llmsTxt: https://business.daily.dev/llms.txt --- ## What is marketing attribution? Marketing attribution is deciding which touches get credit for a result. Last-click gives it all to the final visit, multi-touch spreads it, view-through credits an ad seen but not clicked, and incrementality measures what the campaign caused. Also called: Attribution, View-through conversion, Incrementality. The problem is that buying is a path, not a click. A developer sees a post in a feed, reads the docs a week later from a search, asks a colleague, and signs up on another device by typing the address. Last-click attribution hands that sign-up to "direct" and reports that the campaign did nothing. The fixes are partial and complementary. Multi-touch models spread credit across the path they can see. View-through conversions credit ads that were seen before a conversion even without a click. First-party identifiers (a hashed email, your own user ID) connect the visit on one device to the sign-up on another. Incrementality tests compare exposed and unexposed groups and measure the difference, which is the only method that answers "what did the campaign cause". For developer marketing, assume the path is long and mostly invisible, instrument what you can (pixel, UTMs, first-party IDs), and judge channels on lift and on cost per activated developer over weeks, not on last-click reports over days. ## What are the attribution models, and what does each get wrong? - **Last click:** all credit to the final visit. Simple, and it hands most developer sign-ups to "direct". - **First click:** all credit to the first touch. Flatters top-of-funnel channels and ignores what closed. - **Linear and position-based:** credit spread across the touches that were seen. Fairer, still blind to touches that left no trace. - **Data-driven:** a model weights touches by their observed contribution. Needs volume developer campaigns rarely have. - **View-through:** credits an ad that was seen before a conversion without a click. - **Incrementality:** compares exposed and held-out groups. The only one that measures cause. Every model that works from observed touches misses the ones that were not observed, which for developers is most of the path. ## What does a real developer buying path look like? A backend engineer sees a promoted post about tracing in their feed on Tuesday and reads it. The following week a colleague mentions the same tool. They search for it on Friday, read the docs, and sign up two weeks later from a laptop at home by typing the address. Last-click attribution files the sign-up under direct. The first touch, the post, left a cookie on a different device and a memory on the same person; only one of those is in the report. Multiply that by a team and a quarter, and a campaign that started a hundred evaluations reports a handful of conversions. ## How do you instrument attribution for developer marketing? - **A [conversion pixel](https://business.daily.dev/pixel.md)** that fires on sign-up and activation, with a weeks-long window. - **First-party identifiers** (a hashed email or your user ID) passed with the event, so a sign-up on another device connects to the ad. - **UTM parameters** on every link you control, in one scheme. - **View-through conversions** reported separately from click-through. - **A "how did you hear about us" question** at sign-up, with the answers read monthly. - **An incrementality test** once a year on the biggest channel: hold out a region or a segment and compare. None is complete. Together they turn "the campaign did nothing" into a floor with a known shape. ## How should attribution shape channel decisions? Judge channels on cost per activated developer with a long window, on lift where you can measure it, and on the self-reported source, rather than on last-click reports over days. Expect search and direct to be over-credited and feeds, newsletters and communities to be under-credited, and correct in that direction when the reports and the self-reported answers disagree. The channel that keeps showing up in "how did you hear about us" and never in last-click is usually the one doing the work. **Common mistakes** - Cutting a channel because last-click reports it as zero. - A seven-day window on a path that takes a month. - Ignoring the self-reported source because it is not "data". - Never running an incrementality test on the biggest spend. **How developers experience it** - They saw the post, forgot where, and signed up when the need arrived. - They answer "how did you hear about us" honestly when the options are real. - They evaluate across devices and weeks, invisibly to any single tool. ## On daily.dev The daily.dev conversion pixel reports clicks and what happened next, view-through conversions for developers who saw the ad and converted later, and matches sign-ups to the campaign with your own hashed IDs, so a sign-up days later on another device is still credited. ## Attribution models for developer campaigns | | Last click | Multi-touch | View-through | Incrementality | | --- | --- | --- | --- | --- | | **Credits** | The final visit | Every observed touch | Ads seen before a conversion | What the campaign caused | | **Misses** | Everything before the last visit | Touches that left no trace | Nothing it did not serve | Nothing, by design | | **Effort** | None | Analytics setup | A pixel | A holdout test | | **For developers** | Hands sign-ups to "direct" | Better, still partial | Catches the read-now-sign-up-later path | The only honest answer | | **Use it** | Never alone | For directional reporting | Always, reported separately | Yearly, on the biggest channel | ## Go deeper - [Marketing attribution for developer products](https://business.daily.dev/resources/marketing-attribution-for-developer-products-what-works.md) - [Proving ads work without attribution: incrementality](https://business.daily.dev/resources/incrementality-dev-campaigns-proving-ads-work-without-attribution.md) - [The conversion pixel on daily.dev](https://business.daily.dev/pixel.md) Related terms: [Conversion pixel](https://business.daily.dev/glossary/conversion-pixel.md) | [UTM parameters](https://business.daily.dev/glossary/utm-parameters.md) | [Developer acquisition](https://business.daily.dev/glossary/developer-acquisition.md) ### Frequently asked questions **Why does last-click attribution undercount developer campaigns?** Because developers evaluate before they commit: they see the ad, read docs later, and sign up by typing the address or from a search. Last-click credits the final visit and reports the campaign as having no effect. **What is a view-through conversion?** A conversion by someone who saw an ad and later converted without clicking it. It is reported separately from click-through conversions, as extra lift on top of the clicks you plan around. **What is incrementality?** The share of results that would not have happened without the campaign, measured by comparing a group that saw it with a matched group that did not. It is the most honest attribution method and the hardest to run. **What attribution window should a developer campaign use?** Weeks, not days. Thirty days is a reasonable default for sign-up and longer for activation, because developers read now and act when the need arrives. **How does daily.dev handle attribution?** The conversion pixel reports click-through and view-through conversions and matches sign-ups to the campaign with your own hashed identifiers, so a sign-up days later on another device is still credited. [All terms](https://business.daily.dev/glossary.md) | [Explore your options](https://business.daily.dev/schedule/?ref=glossary-marketing-attribution) | [How pricing works](https://business.daily.dev/pricing.md)