Most customers see you more than once before they call. Multi-touch marketing attribution is the set of rules for dividing the credit among those moments. Here are the seven common models, run on one real-looking customer, and an honest take on how much of this a small business actually needs.
A touch (or touchpoint) is any recorded moment where a person interacted with your marketing: an ad click, a visit from a search result, an email click, a call. Attribution is deciding which of those moments earns credit for the lead or the sale. Single-touch models hand all the credit to one moment, usually the first or the last. Multi-touch attribution spreads it across several, using a fixed rule or a statistical model.
The appeal is obvious. A homeowner who first saw your remodel video on Instagram, searched for you a week later and finally called from a branded search ad was influenced by all three. Giving the branded search ad 100% of the credit tells you to cut the video that started it. Multi-touch models try to correct that.
The catch is that every model, however clever, can only divide credit among the touches it can see. That point decides most of what follows, so keep it in mind as we go through the models.
An illustrative example, not a real client: a homeowner touches your ads five times over 24 days, fills in a quote form on day 14, then books an estimate by phone on day 24. The job closes at $10,000. Here is how much of that each model credits to each touch.
| Touch | First | Last | Linear | Time-decay | U-shaped | W-shaped | Data-driven |
|---|---|---|---|---|---|---|---|
| 1. Meta video adDay 1 · first touch | $10,000 | — | $2,000 | $420 | $4,000 | $3,000 | Learned |
| 2. Google search adDay 9 | — | — | $2,000 | $926 | $667 | $500 | Learned |
| 3. Branded Google search adDay 14 · quote form (lead) | — | — | $2,000 | $1,520 | $667 | $3,000 | Learned |
| 4. Meta retargeting adDay 21 | — | — | $2,000 | $3,041 | $666 | $500 | Learned |
| 5. Google search ad → callDay 24 · estimate booked | — | $10,000 | $2,000 | $4,093 | $4,000 | $3,000 | Learned |
| Meta / Google split | $10,000 / $0 | $0 / $10,000 | $4,000 / $6,000 | $3,461 / $6,539 | $4,666 / $5,334 | $3,500 / $6,500 | Varies |
Same customer, same $10,000, and Meta’s share runs anywhere from all of it to none of it. None of these models is “correct.” Each is a policy for sharing credit, and the policy you pick can change which platform looks profitable. That is the real reason to understand them: so a report never surprises you with its own assumptions.
The percentages below are the common conventions. Tools vary, and most let you adjust them.
All credit to the first recorded touch. Good for: seeing which channels introduce new customers. Misleads: ignores whatever actually got them to act, and depends on the first touch being captured at all.
All credit to the final touch before the conversion. Good for: knowing what closes, and easy to verify. Misleads: overcredits branded search and retargeting, which often catch people who had already decided.
Equal shares for every touch. Good for: a neutral first look at long journeys. Misleads: a passing impression and the call that booked the job count the same.
More credit to touches closer to the conversion, usually with a half-life such as seven days. Good for: short, promotion-driven cycles. Misleads: shortchanges the ad that started a months-long remodel decision.
Typically 40% to the first touch, 40% to the last, 20% shared among the middle. Good for: valuing both introduction and close. Misleads: the 40/20/40 split is a convention, not a measurement.
Typically 30% each to the first touch, the touch that created the lead and the touch that created the opportunity (here, the booked estimate), with 10% spread over the rest. Good for: businesses with clear funnel stages. Misleads: needs those stage moments recorded reliably, including offline ones.
An algorithm compares paths that converted with paths that did not and assigns credit from the difference. Good for: large accounts with lots of conversions. Misleads: a black box, and it only learns from the touches and conversions it was given.
Keep first touch and last touch side by side on every lead, and tie closed revenue to that lead. Two plain views you can check by hand often beat one fractional view nobody can audit.
Google Ads uses data-driven attribution as the default for most new conversion actions, with last click as the main alternative. Google retired the first-click, linear, time-decay and position-based options from Google Ads and Google Analytics 4 in 2023. Data-driven attribution in Google Ads shares credit among Google ad interactions only — it does not see your Meta ads, your yard signs or the neighbor who recommended you. That is why its conversion columns show fractions, as explained in Google Ads Says 30 Conversions. Your CRM Says 9.
GA4 also defaults to a data-driven model in its reports and can switch to last-click views. It sees more channels than an ad account does, but only on the website, only for visitors it can track, and not the calls or the closed jobs.
Meta Ads Manager does not ask you to pick one of these models. It counts a conversion for its own ads when it falls inside an attribution window measured from a click or a view; the common default is 7 days after a click and 1 day after a view, and the window options available have changed over time, so check what your ad set uses. View-through and modeled conversions are part of the count, which is a big reason Meta’s numbers run ahead of yours — see Facebook Says 40 Leads. You Got 12.
The shared limitation: each platform attributes within its own walls. In the example above, if both platforms are told about the sale, Google’s data-driven model can credit all of it to its own three touches while Meta, if the retargeting ad was clicked within its window, counts the same sale for itself. Add the two dashboards together and you made $20,000 from a $10,000 job. No platform model fixes that, because none of them can see the other platforms. More in Platform ROI Versus Click Metrics.
Go back to the example and delete touch 5, the phone call. That is what happens when a business has no call tracking: the estimate was booked by phone, and the model never saw it. Every column in the table now redistributes credit among four touches, and every one of them is wrong in a slightly different way. A more sophisticated model does not help; it just divides the missing data more precisely.
For a local service business, the gaps are rarely small. Calls, walk-ins and referrals leave no click. iPhone and Safari privacy features shorten how long ordinary tracking remembers a visitor, as covered in Why Your iOS and Safari Conversions Disappeared. People switch from phone to laptop. And the number that matters — whether the job closed, and for how much — usually lives in a CRM or field-service tool that no attribution model looks at by default.
Volume matters too. Statistical models need many conversions to learn reliable patterns, and a business closing a few dozen jobs a month gives them little to work with. Fixed-rule models like U-shaped do not have that problem, but their percentages are assumptions you are choosing, not findings.
So the order of work for most small businesses is: capture the first touch on every lead and never overwrite it; capture the most recent touch separately; cover every lead path, phone included; and match closed revenue to the lead. With those in place, a first-touch view and a last-touch view of the same revenue answer the two questions owners actually ask — what brings new customers in, and what gets them to act — and you can open any lead and check why it was credited. If you later want a U-shaped or W-shaped view, that clean data is exactly what it needs. The build order is laid out in Building a Simple, Powerful Attribution System, and capturing the offline half is covered in Tracking Offline and Online Touchpoints.
Pull last month's customers and see how many have a recorded source at all. If a large share are blank or “direct,” fix capture first — tracking numbers, first-party cookies, lead-form syncs. No model can credit a touch that was never recorded.
Report revenue both ways. Where they agree, you have a clear answer. Where they disagree — typically social introduces and search closes — you have learned how your channels work together, which is the whole point of multi-touch thinking.
Let each platform's model steer bidding inside that platform. Decide budget between platforms on spend versus closed revenue from your own records, credited once per sale.
If journeys are long and your data is complete, a U-shaped or W-shaped view can be a useful third lens. Keep the single-touch views alongside it, and write down which model each report uses.
AuriaTrack is built around the inputs this guide argues matter most. The tracking script stores UTMs, the Google click ID (gclid) and Meta’s fbclid in first-party cookies on your own domain, alongside a separate write-once first-touch record, so a visitor’s original source is not overwritten by later visits. Attribution codes come from platform campaign IDs rather than names, so renaming a campaign does not split its history.
Won deals sync in from GoHighLevel, Square, ServiceTitan, Housecall Pro, Jobber or a CSV and are matched to the lead that started them. The reporting shows ROAS at the campaign and platform level, cost per closed deal and a side-by-side platform comparison, all from revenue you can trace to a real customer. Closed-deal values also go back to Google and Meta through server-side Enhanced Conversions and the Conversions API, so their own models learn from paid jobs rather than form fills.
Revenue attribution reporting is in every package. Call tracking with dynamic number insertion — the fix for the missing touch 5 — is part of Ascent and Summit. See how attribution works → or compare the packages on the pricing page.
Want to see what your current numbers say? Run them through the ROAS calculator or start with the free 5 Numbers guide.
Book a demo and we'll show you every lead with its first touch, its latest touch and the revenue it became.