Skip to content

Blog · 2026-11-04

Last-click vs multi-touch attribution: how they differ and when to use each

Last-click attribution gives 100% of the credit to the final touchpoint. Multi-touch shares it. See the trade-offs and steps to move from one to the other.

Marketer arranging cards showing a TV, phone and shopping bag in a row on a desk

Last click attribution gives 100% of the credit for a conversion to the final touchpoint before it. Multi-touch attribution spreads that credit across the earlier ads, emails and visits as well. Last-click is simpler to set up and suits lower-funnel channels like search and retargeting. Multi-touch shows what display, video and CTV contributed, but it needs more data to build.

What last-click attribution is

Last-click attribution is a rule that gives a conversion entirely to the last thing a customer interacted with before buying or signing up. Every earlier touchpoint gets zero credit. You'll also see it called last-touch or last-interaction attribution. They all mean the same thing.

A common variant is last non-direct click. If someone types your URL straight into a browser, there's no campaign to credit, so this version skips the direct visit and credits the last trackable channel before it, such as a paid ad or an email. That keeps credit on things you can measure. It still doesn't explain what made the person type your URL.

Here is a simple path. A shopper sees a sneaker ad on social media, sees a second ad from the same brand a few days later, browses the website, gets an abandoned-cart email, and finally clicks a shopping search ad and buys. Under last-click, the search ad gets all the credit. The two social ads, the site visit and the email get nothing.

Last-click was the default model for years, including in Universal Analytics, Google's older analytics product. Because it was the standard for so long, many analysts still use it as the baseline when they compare other models.

What multi-touch attribution is

Multi-touch attribution (MTA) gives each tracked touchpoint a fraction of the credit for a conversion. In the sneaker example, the social ads, the email and the search ad would all get a share. How large each share is depends on the model:

  • Linear splits credit evenly across every touchpoint.
  • Time decay gives more credit to touches closer to the conversion.
  • U-shaped (position-based) weights the first and last touches most heavily. A common split is 40/40/20: 40% to the first touch, 40% to the last, and 20% shared across the middle, as described in a 2026 guide to attribution models.
  • W-shaped adds a third heavy weight at a mid-funnel event, such as lead creation, which makes it useful in B2B.
  • Algorithmic and data-driven models calculate the weights from your own conversion data instead of a fixed rule.

People often mix up two terms. Multi-channel describes scope: the customer journey crosses several media. Multi-touch describes method: credit is shared across interactions. Most multi-touch models are multi-channel, but a report that covers many channels can still be last-click.

How we compared the two approaches

We make Waveband DSP, a self-serve advertising platform that includes its own measurement reporting, so weigh our view with that in mind. This section compares two measurement methods, not two vendors. Any analytics tool can run either one.

We judged both approaches on the same five points:

  • how much effort setup takes
  • what data and identity they need
  • which channels they tend to over-credit or under-credit
  • how much privacy exposure they carry
  • which businesses they fit best

Last-click vs multi-touch attribution side by side

CriterionLast-clickMulti-touch
Setup effortLow. A spreadsheet and a few months of data can work.Higher. Needs journey stitching and model configuration.
Data and identityCan run on campaign-level UTM data without persistent identifiers.Needs user-level journey data across touchpoints.
Channel biasOver-credits closers such as branded search and retargeting. Under-credits display, video and CTV.U-shaped weightings are an assumption, not a finding, and time decay structurally undervalues awareness.
Privacy exposureLower, because it can work with anonymized campaign data.Higher, because it tracks individuals across sessions.
Cross-device journeysCredits only the last device or session.Better, but still has gaps when tracking loses a person between screens.
Best fitSmall teams, short cycles, lower-funnel tactics.Several channels, longer cycles, real awareness spend.

Neither approach is complete on its own. Last-click tells an easy story that leaves things out. Multi-touch tells a fuller story, but it's harder to build and explain, and its answers depend on the weights you choose.

When last-click attribution is the right call

Last-click is a good choice in these situations:

  • You're a small team starting out. It's the easiest model to understand and set up, and it beats having no attribution at all.
  • Your purchase cycle is short. When most people buy within a visit or two, the last touch usually is the main reason they bought.
  • You're judging lower-funnel tactics. Retargeting, branded search and affiliate campaigns exist to close sales, and last-click measures exactly that. If you run retargeting, start by building retargeting audiences from your site tag, then read those campaigns on a last-click basis.
  • You need a report that finance and sales will read. Revenue leaders like last-click because it matches the moment they saw a deal come in. Finance teams accept it for quick snapshots.

The view is clear but narrow. Use last-click to measure your closers, not to decide how much to spend on awareness.

Where last-click misleads programmatic buyers

The biggest problem for programmatic buyers is that people see a lot of ads they never click. Someone watches a 30-second spot in a streaming app on their TV (connected TV, or CTV), sees a few display ads during the week, then searches your brand name on their phone and buys. Last-click gives the whole sale to branded search. That search often captures interest the TV ad and display ads created.

The result is predictable. Display, online video and CTV look weak in reports, the channels that close the sale look strong, and budget moves toward the closers. When teams cut upper-funnel spend, fewer people come into the funnel later, and the closing channels then have less demand to capture.

Device switching makes this worse. People often start on one screen and finish on another. Last-click only sees the final device or session, so the TV or desktop touch that started the journey disappears.

Long B2B cycles add another problem. When a deal takes months to close, last-click can credit one email or one demo-request click and ignore the months of content, ads and visits that educated the buyer.

What changed in Google Analytics 4

Google Analytics 4 (GA4) uses data-driven attribution as its default model. Google's help page lists three attribution models: data-driven, paid and organic last click, and Google paid channels last click. GA4 removed first-click, linear, time-decay and position-based models in November 2023. To use one of those now, you have to calculate it yourself outside GA4, for example in BigQuery.

GA4's data-driven model looks at up to the last 50 interactions on a path. By default it uses a 90-day lookback window, which you can shorten to 30 or 60 days. Universal Analytics' data-driven model, by comparison, only considered the previous 4 interactions.

If GA4 and your ad platform report different conversion counts for the same campaign, the cause is often attribution, not broken tracking. The two tools read different inputs. An ad platform focuses on its own ad interactions and bidding. GA4 looks at the wider journey across organic, paid, direct and social. Some conversions are also modeled when signals are missing. Analytics Mates gives the example of 30 conversions in the ad platform and 22 in GA4 for the same campaign and period.

How to move from last-click to multi-touch

  1. Keep last-click as your baseline. Don't switch it off. You'll compare every new model against it.
  2. Add a linear view as a sanity check. If a channel looks far better under linear than under last-click, last-click is probably under-crediting it.
  3. Try time decay or position-based models for longer cycles. In GA4 that means calculating them yourself, because they're no longer built in.
  4. Match the attribution window to your sales cycle. A 90-day window is often too short for mid-market B2B deals, where 120 to 180 days may fit better.
  5. Segment by deal size and buyer type. A self-serve trial and a large contract follow different paths, so one model across both produces averages that don't describe either.
  6. Compare results across models. Channels that score much higher under multi-touch than under last-click are your candidates for more budget.
  7. Audit your tracking on a schedule. Broken tags and missing UTMs distort every model, so check that conversion events and campaign tags still fire correctly.
  8. Test before you move real budget. An attribution model shows correlation. An incrementality test (a holdout group that doesn't see the ads) or media mix modeling shows whether a channel actually caused sales.

Measuring display, video and CTV in Waveband

Waveband Signal is the measurement layer built into Waveband DSP. It includes a first-party site tag, impression and click tracking, conversion events, click-through and view-through attribution, and multi-touch attribution reporting. Reports export to CSV and can be shared with a link.

View-through attribution matters most for channels people rarely click. A CTV spot plays on a TV with no link to click, so last-click will never credit it. View-through attribution records that a person saw the ad and later converted on your site, which lets you compare CTV, online video and display against your search and retargeting numbers.

Attribution reports still show correlation, not proof, so run the incrementality checks from step 8 before you make large budget changes. For the case for owning this data instead of relying on each platform's own numbers, read why first-party measurement matters.

See how it works in Waveband

By Waveband Team