What is marketing attribution? A practical guide

Updated 2026-11-02

Marketing attribution is the practice of connecting sales to the marketing that helped cause them. You take a payment, look back at how that buyer found you, and give credit to the channels, campaigns and pages involved.

The point is a better budget decision. If you know which marketing produces revenue, you can do more of it. If you know which marketing only produces traffic, you can stop paying for it.

This guide covers how attribution works, the main models, the data you need, where tracking falls short, and how to set it up in a week. It links out to deeper pieces on each topic.

What marketing attribution means in plain terms

Say someone reads your blog post on a Monday, clicks a LinkedIn post two weeks later, and subscribes after opening your newsletter. They pay $1,000. Three pieces of marketing touched that sale. Attribution is how you decide what each one gets credit for.

Each of those moments is a touchpoint. The full list, in order, is the customer journey. The rule that splits the credit is the attribution model.

Attribution answers questions like these:

• Which channels produce revenue, and which only produce visits?
• Which campaigns are worth repeating?
• Which landing pages turn into paying customers?
• How long does it take someone to buy after their first visit?

It's different from web analytics. Web analytics counts what happens on your site: visitors, pages, sessions and clicks. Attribution joins those visits to money. You need the first to do the second, and that's why most tools that do attribution sit on top of some form of analytics.

Why attribution matters for growing companies

Most marketing reports stop at leads or signups. Those numbers can go up while revenue stays flat. A channel can send 5,000 visitors a month and close nothing, while a smaller channel sends 200 visitors and pays for your whole team.

Here's an example with made-up numbers:

Judged by visitors, paid social wins. Judged by revenue, it's last. You can only see that once visits and payments are in the same report.

Attribution also settles arguments. Without it, the loudest person in the room decides where the budget goes. With it, you can point at a number and ask where it came from.

How marketing attribution works

Every attribution system does the same four jobs, even if the details differ.

1. Collect visits. A script on your site records each visit, along with where it came from. Source data comes from the referrer and from UTM parameters in your campaign links.

2. Recognize the person. A cookie links visits from one browser over time. When someone signs up or fills in a form, an identify call ties that anonymous history to a real person.

3. Collect sales. Payments come from your billing system. For SaaS that's usually Stripe. The sale is matched to the person, usually by email or customer ID.

4. Assign credit. The system looks back through a set window, finds that person's touchpoints, and splits the sale across them with a model.

The window matters. A lookback window is how far back the system searches for visits before a sale. With a 90-day window, a visit 120 days before the payment gets no credit. Set it too short and you miss your slow buyers. If your sales cycle runs four months, a 30-day window will hide most of the journey.

The main attribution models

A model is a rule for splitting credit. There are four common ones. We cover them in depth in marketing attribution models explained, including a $1,000 sale worked through each one.

First touch. All credit goes to the first visit. It shows what introduces people to you. It's useful for judging awareness work, and it ignores everything that closed the deal.

Last touch. All credit goes to the last visit before the payment. It shows what tips people over. It tends to flatter branded search, email and direct visits, because buyers use those right before they pay.

Linear. Credit is split evenly across every visit. It's simple and fair, and it treats a passing glance the same as a pricing page visit.

U-shaped. The first and last visits get 40% each, and the remaining 20% is split across the middle. With only two visits, each gets 50%. It rewards the moments that open and close the journey and still gives the middle some credit.

Here's one journey of four visits and a $1,000 sale under each:

Linear and U-shaped are both forms of multi-touch attribution. There are also data-driven models, which use statistics to estimate each channel's effect. They need a lot of conversions to work, and their output is hard to check by hand. For most small and mid-sized companies, simple rules you can explain are a better start.

Which model should you choose?

Don't hunt for the one true model. No model is correct, because none of them measures cause and effect. They're different lenses on the same journey.

A better approach is to compare models and look at where they agree and where they split:

• If a channel looks strong in all four, it's probably working.
• If a channel looks strong only in first touch, it's good at introductions. Don't cut it for a low last-touch number.
• If a channel looks strong only in last touch, it may be capturing demand that something else created.
• If a channel looks weak in all four, it's a candidate to cut.

If you need one default, use U-shaped for a sales cycle with several visits. Use first touch when you're mainly trying to learn what brings new people in. Keep the same model when you compare periods, or the trend will reflect the switch and tell you nothing.

Attribution for B2B and SaaS

B2B journeys are longer and involve more people. One person reads the blog, another attends the webinar, and a third signs the contract. Cookie-based tracking sees each of those as a separate visitor unless they're identified.

That creates three practical rules:

• Use a longer lookback window. Start with your typical time to close, then add a margin.
• Identify people early. Call identify at signup and on key forms, so the anonymous history joins the right person.
• Ask buyers directly. People know things your tracking doesn't.

Read more in our guide to B2B marketing attribution.

What attribution can't see

This is the part many vendors skip. Tracking has real blind spots, and you should plan around them.

Ad blockers. Some visitors block analytics scripts. Those visits never get recorded.

Multiple devices. Someone who reads on a phone and buys on a laptop looks like two visitors until they're identified.

Dark social. A link shared in a group chat or a Slack channel usually arrives with no referrer, so it shows up as Direct. Podcasts, conversations and AI assistant answers leave little or no trace too. See our guide to dark social.

Influence versus cause. Attribution shows that a touchpoint happened before a sale. It can't prove the touchpoint caused the sale. Some of those buyers would have bought anyway. If you need proof of cause, run a holdout test where one group doesn't see the marketing.

Visits outside the window. A payment with no visit inside the lookback window has nothing to credit. A good tool reports it as unattributed and doesn't quietly assign it somewhere.

The honest takeaway is that attribution is a strong signal and an imperfect record. Treat it as evidence for a decision. Don't treat it as a final score.

Add self-reported attribution

The simplest fix for tracking's blind spots is to ask. Add a "How did you hear about us?" field to your signup or checkout form.

Self-reported data and tracked data catch different things. Tracking is precise about clicks and misses word of mouth. Surveys capture word of mouth and depend on memory. Used together, each one checks the other. If 30 people in a month name a podcast and your tracking shows them as Direct, you've found a channel your reports were hiding.

Our how did you hear about us guide has answer options and placement advice. Self-reported attribution in Headstream reads a heard_about form field, groups free-text answers, and puts them next to each person's first tracked visit.

How to set up marketing attribution

You can get a useful first version running in a week.

Step 1: Tag your links. Agree on one set of UTM conventions and use them everywhere. Lowercase only. Decide what goes in source, medium and campaign. Messy tags are the most common reason attribution reports look wrong.

Step 2: Install tracking. Add your analytics script to the shared layout of your site, so every page is covered.

Step 3: Identify people. Send an identify call when someone signs up, so their earlier visits join their account.

Step 4: Connect revenue. Bring in payments from your billing system. Include refunds and disputes. A sale that gets refunded shouldn't stay in a channel's total.

Step 5: Add a survey question. Put a "How did you hear about us?" field on your signup form.

Step 6: Set your lookback window. Pick one that fits your sales cycle.

Step 7: Review monthly. Compare models and look for channels where they disagree. Pick one budget change. Check the result the next month.

Common attribution mistakes

Picking one model and treating it as truth. Last touch alone will push you to cut the channels that create demand.

Ignoring refunds. If a channel brings buyers who refund, its revenue is overstated until you subtract them.

Comparing tools without checking definitions. Two tools can count "conversions" differently. Read the definitions before you trust a difference.

Trusting a number you can't trace. If you can't get from a channel's total to the payments behind it, you can't tell whether it's right.

Never reconciling to billing. Your attributed revenue should tie back to your billing totals. If it doesn't, find out why. Our piece on why analytics and Stripe disagree walks through the usual causes.

How to choose attribution software

Most tools fall into a few groups: free analytics suites, simple analytics tools, product analytics tools, and B2B attribution platforms built around a CRM. We go through how to pick in best marketing attribution software and give a stage-by-stage decision guide in marketing attribution tools compared.

Whatever you look at, ask these questions:

• Does it connect directly to your billing system?
• Can you compare several models on the same data?
• Do its totals match your billing totals, and can you check that yourself?
• Can you export the raw data?
• Does it show you what it can't measure?

How Headstream does attribution

Headstream is web analytics plus Stripe revenue attribution for SaaS. You add one script tag and connect Stripe with read-only access. Connecting Stripe is in beta. Headstream imports your payment history, and charges, refunds and disputes go into an append-only ledger.

Revenue attribution runs every hour across first touch, last touch, linear and U-shaped. The default lookback is 90 days, and you can adjust it. A run publishes only if its credits add up exactly to the ledger. If they don't, the last good run stays up. Refunds and disputes are credited back to the original sale, so a channel loses credit when its buyers do.

You can check the work. Every headline number has a page that shows what's counted, and any payment can be traced to the visits it credited. See how we show our work and how each number is calculated. The free plan covers 10,000 events a month.

Headstream doesn't import ad spend or sync with a CRM today. If you need return on ad spend or account-level B2B reporting, you'll want another tool for that part.