Multi-touch attribution

Last-touch reports flatter the channels near the end of the journey and starve the ones near the start. A blog post or webinar that builds awareness then shows no revenue, gets cut, and the demo requests dry up a few months later.

A multi-touch view shows which channels work together and in what order. It helps marketing explain its contribution and helps sales see where good leads start. It also makes budget talks less political, because the same journey data is on the table for everyone.

It is not proof of cause. It spreads credit by a rule, and a different rule gives a different answer. Use it to ask better questions, not to settle them.

• Fix tracking first. Use UTM tags on campaign links, record key actions as events and tie them to one person, so a path does not break into separate visits.
• Send touchpoints into the CRM so closed deals carry their history.
• Choose the model to fit the sales cycle: linear for short, similar paths, position based or time decay for long ones.
• Compare against first-touch and last-touch to see which channels change most.
• Set the attribution window to match how long deals actually take.

Buyers rarely convert after one visit. A deal might follow a LinkedIn post, a blog article, a webinar and finally a demo request. Counting only the last click credits the demo form and ignores everything that made it happen. Multi-touch attribution records each interaction, or touchpoint, and divides the credit between them using a rule, called an attribution model.

Common rules:

• Linear: every touch gets equal credit.
• Time decay: touches closer to the sale get more.
• Position based, or U-shaped: the first and last touch get the most, with the rest shared in between.
• Data driven: software works out weights from historical paths.

For a 12,000 deal with those four touches, linear gives 3,000 to each. Last-touch attribution gives all 12,000 to the demo request.

• Trusting the model without checking tracking. If half the journeys are broken, the credit is spread over fiction.
• Treating the chosen weights as fact. Compare two or three models and see which conclusions hold in all of them.
• Running it on too few deals. With 10 deals a year, patterns are noise. Use first-touch and last-touch and read the deals.
• Ignoring offline touches. Meetings, referrals and calls are real influences that tracking does not see.
• Using it as the only evidence. A controlled test, such as pausing a channel in one region, says more about cause.

It only sees what is tracked. Cookie limits, consent choices, ad blockers, offline meetings and private sharing leave gaps. It also shows correlation, not proof that a touch caused the sale. For a small business with few deals, a simple first-touch and last-touch comparison is often enough, and a controlled test says more about whether a channel really works.