Which Attribution Model Should You Use? Match the Model to the Revenue Question

Attribution debates usually stall because teams argue about models before agreeing on the question.

The model is not the decision.

The decision is which conversion you are trying to explain, and which credit rule makes that explanation honest.

Once you name the question, the model choice mostly makes itself.

This article walks through the main attribution models, the revenue question each one answers well, and the tradeoffs to accept when you pick it.

It also covers the report types beneath the models.

Choosing a model without the right conversion event often produces a confident report answering the wrong thing.

Start with the conversion, not the model

Attribution reporting always measures a conversion.

Before comparing first touch against linear or anything else, decide what event you want to explain.

Three data sources define that choice in HubSpot's attribution reporting.

Contact create attribution measures what generated contacts.

Deal create attribution measures what generated deals.

Deal revenue attribution measures what influenced deal revenue.

HubSpot describes these as roughly the top, middle, and bottom of the funnel for businesses operating with a funnel model (HubSpot Knowledge Base).

That ordering matters more than the model.

A first-touch view of contact creation tells you which channels start relationships.

A revenue attribution view tells you which interactions sit alongside closed revenue.

If your leadership question is 'where should we invest to fill the top of the funnel,' a bottom-of-funnel revenue report will mislead you no matter which model you apply to it.

So the working sequence is: name the question, pick the conversion event, then pick the credit rule.

Teams that reverse this order tend to pick a model first because it sounds sophisticated, then discover the report cannot answer what they actually asked.

The models and the questions they answer

HubSpot supports five attribution models: first touch, last touch, linear, time decay, and empirical (HubSpot Knowledge Base).

Each encodes a different belief about which interactions deserve credit.

First touch: which channels start relationships?

First touch gives all credit to the first recorded interaction.

Use it when your question is about discovery: which content, campaigns, or channels bring people into your world at all.

It is useful for evaluating top-of-funnel investments such as organic content, podcast appearances, or awareness campaigns whose value shows up later in the journey.

The tradeoff is obvious but worth stating.

First touch ignores everything that happened after the first interaction.

A channel that starts many relationships but never contributes to progression will look strong, and a channel that quietly moves late-stage deals forward will look invisible.

Treat first-touch reports as one lens, not a budget allocation tool on their own.

Last touch: what closed the deal?

Last touch gives all credit to the final interaction before conversion.

It answers a closing question: what was happening right before someone converted?

This is useful for evaluating bottom-of-funnel assets such as pricing pages, demos, or retargeting, and for understanding what sales-enablement content correlates with conversion.

Its weakness mirrors first touch.

Last touch over-credits whatever happens to be nearest the finish line, which is often a branded search or a direct visit from someone your marketing had already warmed up.

If you use last touch for budget decisions, you will systematically starve the earlier work that created the demand.

Linear: how does the whole journey contribute?

Linear gives equal credit to every interaction.

It answers a breadth question: across all the touchpoints in a customer journey, how is involvement distributed?

It is a reasonable default when you genuinely do not know which interactions matter more and want a balanced picture rather than a strong assumption baked in.

The tradeoff is that equal credit is itself an assumption.

A blog read in month one and a demo request in week twelve get the same weight.

Linear is honest about its simplicity, making it good for cross-functional reviews.

It is weak for optimizing spend between early and late-stage programs.

Time decay: what mattered most recently?

Time decay gives more credit to interactions closer to the conversion.

It answers a recency question: among everything that happened, which interactions clustered near the decision?

It suits longer sales cycles where late-stage activity is genuinely more decision-relevant, but you still want earlier touches represented rather than erased.

Compared with last touch, time decay is a softer version of the same instinct.

It keeps early interactions on the report while acknowledging that a touchpoint from a year ago probably influenced the deal less than last week's evaluation call.

Empirical: what does our own data say?

The empirical model is the outlier.

Instead of applying a fixed rule, it weighs interaction types based on how often they occur across your historical conversion paths.

Interaction types that appear less often receive more weight, because they are more distinctive; all individual interactions of the same type receive equal weight.

HubSpot notes that this model replaces the earlier U-shaped, W-shaped, J-shaped, and inverse J-shaped models (HubSpot Knowledge Base).

Use empirical when you have enough historical interaction data for meaningful weighting.

It reflects your actual journey patterns rather than a generic rule.

The tradeoff is interpretability.

Explaining to a stakeholder why a form submission outweighs a page view is harder when the answer is 'the data says so.'

It also inherits the patterns of your past; if your tracking has gaps, the model learns from those gaps.

Match the report type to the leadership question

Because each model can be applied to different conversion events, the practical pairing looks like this:

• 'Which channels create new contacts?' Contact create attribution, often read through first touch or linear.
• 'Which interactions help deals get created?' Deal create attribution, useful for judging whether marketing activity feeds pipeline, not just audience.
• 'Which interactions sit alongside revenue?' Deal revenue attribution, typically read with time decay or empirical when the journey is long.

Deal create and deal revenue attribution require Marketing Hub Enterprise.

Contact create attribution is available on lower tiers (HubSpot Knowledge Base).

If your subscription limits the report types available, that constraint may shape your roadmap before any model philosophy does.

Practical tradeoffs operators should weigh

One model for one meeting, several for real decisions

A single model rarely serves every audience.

First touch flatters demand generation; last touch flatters closing motions; linear spreads accountability.

A useful operating habit is to run the same period through two or three models and look for interactions that earn credit consistently.

Touchpoints that matter under multiple credit rules are more robust candidates for investment than ones that only shine under a single lens.

Watch what the data includes

Attribution reports are only as complete as the interactions they capture.

HubSpot attribution reports include up to twenty million interactions per report after sampling (HubSpot Knowledge Base).

High-frequency types like page views and email interactions may be sampled.

Lower-volume types like form submissions, calls, and meetings are not.

Sampling preserves the date range and stops high-volume types from dominating credit.

Very high-volume accounts should check which types were sampled before drawing fine-grained conclusions.

Reports can also include up to one hundred event input types, so the set of interactions you include is itself a modeling decision (HubSpot Knowledge Base).

Excluding sales meetings from a revenue attribution report will change the story, whatever model you apply.

Tracking gaps shape every model

No model fixes missing data.

If a channel is not tracked, it earns no credit under any rule; the only difference is how visibly the absence shows up.

Before trusting any attribution report, verify your key channels actually record interactions.

Be cautious with hard-to-track channels like dark social or word of mouth.

Attribution is a lens on recorded behavior, not a census of everything that influenced the buyer.

Interpretability is a real cost

A model your stakeholders cannot explain is a model they will quietly ignore.

Linear and first touch are easy to defend in a meeting precisely because their assumptions are simple.

Empirical may be more faithful to your data but harder to socialize.

Choose with your audience in mind, not only your analytical preferences.

A short decision path

1. Write down the revenue question in one sentence.
2. Pick the conversion event that matches it: contact creation, deal creation, or deal revenue.
3. Choose the credit rule that matches the shape of the question: discovery (first touch), closing (last touch), balance (linear), recency (time decay), or data-driven weighting (empirical).
4. Check which interaction types are included and whether sampling applies to your account.
5. Where the stakes are high, compare the same period under a second model and note where the story changes.

Attribution model choice is not a permanent commitment.

Revisit it when your funnel changes, when you add new channels, or when the leadership question shifts from demand creation to revenue efficiency.

The model that served your growth stage will not automatically serve your efficiency stage.

Where attribution fits in a revenue operations system

Attribution reporting works best when the underlying journey data is clean and connected across your tools.

If you are building that foundation, it helps to understand what revenue attribution means and how marketing touchpoints connect to qualified pipeline before tuning models.

For the underlying definitions, our attribution model glossary entry and revenue attribution glossary entry cover the vocabulary in more depth.

When acting on attribution findings, connecting the right systems matters as much as reporting.

The model you choose is a statement about which part of the journey you are willing to see.

Pick the question first, and the model becomes a consequence rather than a debate.

How Meshline can help. Connect automation, Organic Marketing (demand generation), and customer lifecycle management (Revenue Intelligence).

Bring topic planning, content publishing and performance feedback into the conversation about your workflow. Book a Meshline demo.