Marketing attribution analysis

Marketing attribution analysis is the method of assigning credit for a conversion across the marketing touchpoints that led to it, so each channel can be judged on the revenue it actually influenced. A buyer rarely converts after a single interaction. They might read a blog post, click a paid search ad a week later, open an email, and finally convert from a retargeting ad. Attribution decides how the value of that sale is divided across those four touchpoints.

The reason it matters is budget. Without attribution, the last click before purchase takes all the credit, which systematically flatters channels at the bottom of the funnel and starves the channels that created the demand in the first place. A blog post that introduced the buyer gets nothing, while a retargeting ad that simply caught them at checkout gets everything. Reallocating budget on that basis quietly defunds the work that actually drives growth.

Attribution is a model, not a measurement. There is no single correct answer to how much credit a touchpoint deserves, only different rules for splitting it. The job of the analyst is to choose a model that reflects how buyers actually behave, understand its blind spots, and read the results with those blind spots in mind. The same set of conversions can make a channel look excellent or wasteful depending purely on the model applied.

The calculation assigns a credit weight to each touchpoint in a conversion path, then sums the weighted value to each channel. The weights are decided by the attribution model. The models below differ only in how they distribute those weights across the path.

1. 1

   First-touch attribution

   All credit goes to the first touchpoint in the path. This rewards the channels that create awareness and start journeys, but ignores everything that nurtured and closed the deal, so it overstates top-of-funnel channels.

2. 2

   Last-touch attribution

   All credit goes to the final touchpoint before conversion. It is simple and common, but it overstates closing channels like branded search and retargeting while erasing the channels that built the demand.

3. 3

   Linear attribution

   Credit is split evenly across every touchpoint in the path. A four-touch journey gives each channel a quarter. It is fair and easy to explain, but treats a casual first visit and the conversation that closed the deal as equally important.

4. 4

   Time-decay attribution

   Touchpoints closer to the conversion receive more credit than earlier ones. This suits longer sales cycles where recent interactions reflect genuine buying intent, though it still underweights the channels that opened the journey.

5. 5

   Position-based attribution

   A fixed share, often 40 percent, goes to the first and last touch each, with the remaining 20 percent split across the middle. It credits both demand creation and deal closing, which is why many teams treat it as a sensible default.

A worked example shows how much the model matters. A 10,000 pound deal touches a blog post, a paid search ad, an email, then a retargeting ad. Under last-touch, retargeting earns the full 10,000 pounds. Under linear, each channel earns 2,500 pounds. Under position-based, the blog post and retargeting take 4,000 pounds each, with paid search and email sharing the remaining 2,000 pounds. Same deal, three completely different verdicts on which channel deserves more budget.

A metric tree decomposes attributed revenue into the channels that earned it, and each channel into the inputs that produce its conversions. This turns an attribution report into a diagnostic, because a fall in one channel can be traced to the specific input that caused it rather than blamed on the channel as a whole.

The first level splits total attributed revenue across the main channel groups. Each channel then decomposes into the chain that produces its credited conversions: the volume of touchpoints it generated, the rate at which those touchpoints progressed buyers, and the value of the conversions it influenced. The leaf level holds the operational levers, such as spend, click-through rate, landing-page conversion rate and average deal value.

The tree separates a channel problem from an attribution artefact. If paid revenue drops, the tree shows whether the cause is lower spend, weaker creative, a worse converting page, or simply a model change that moved credit elsewhere. Each of those leads to a different decision and a different owner.

There is no benchmark for how much credit a channel should receive, because that depends on the model and the business. The useful benchmarks are about the gap between models and the quality of the underlying tracking. The size of the swing between first-touch and last-touch for a channel tells you how much your budget decisions depend on the model you happened to pick.

A practical health check is to run two models side by side and compare. If a channel looks profitable under last-touch but barely registers under first-touch, it is a closer rather than a creator, and cutting it may quietly hurt the channels upstream. The wider the swing, the more carefully that channel needs reading. Stable channels that rank similarly under several models are the ones you can fund with confidence.

Improving attribution analysis is partly about better data and partly about better judgement. Cleaner tracking makes the credit split trustworthy, and reading results across more than one model stops you from over-trusting any single view.

The metric tree approach starts by finding the channel with the largest gap between its attributed revenue and its true incremental contribution, then drilling into the branch that explains it. If a channel is credited heavily but holdout tests show little effect, the tree points at the touchpoint volume and conversion inputs that inflate its credit.

KPI Tree lets you connect each channel branch to the team that owns it, with RACI ownership on every node so the accountable owner for paid, organic, email and referral is explicit. When attributed revenue for a channel moves, the platform pushes that change to its owner rather than leaving it to surface in a monthly report. The verified impact loop then checks whether a budget shift actually moved total conversions, which is the difference between reallocating to a real driver and reallocating to a channel that only looked good under the chosen model.

1. 1

   Defaulting to last-touch and forgetting it is a choice

   Last-touch is the easiest model to set up, so teams adopt it by default and then read its output as objective truth. It systematically rewards closing channels and defunds demand creation. Treat it as one view among several, not the answer.

2. 2

   Comparing channels under different models

   Judging organic by first-touch and paid by last-touch is comparing two different things. Credit only means something when every channel is measured under the same rule. Mixed models produce a ranking that is pure artefact.

3. 3

   Ignoring untracked touchpoints

   Offline events, dark social and word of mouth rarely make it into the path. The credit they should have earned gets handed to whatever trackable touch sits nearest, overstating tracked channels and understating real demand drivers.

4. 4

   Treating attribution as proof of causation

   A channel appearing in many winning paths is correlated with conversions, not proven to cause them. Without holdout tests, you cannot tell whether a channel is driving sales or just present when they happen.

5. 5

   Changing the model and reading the swing as performance

   Switching attribution models reshuffles credit across channels even when nothing about the channels changed. Reading that reshuffle as a channel improving or declining leads to confident decisions built on an accounting change.