HubSpot Attribution Reporting: What Each Model Actually Tells You

HubSpot gives you a menu of attribution models. None of them measures anything. Each one is an opinion about which touch deserves the credit, expressed as arithmetic, and the model you choose decides which channel looks good this quarter.

So the argument about which model is the right one never ends, because there is no right one. There is the question you are trying to answer, and the model that answers it.

What follows: the three attribution reports and what each is for, the models and what each is claiming, how to read them when paid, organic, partner and outbound all touch the same deal, and the Australian accounting error that makes the ROI figure underneath the whole exercise wrong.

Three reports, three different questions

HubSpot's attribution reporting comes in three types, and they answer different things.

• Contact create attribution, which interactions get the credit for the contact existing at all.
• Deal create attribution, which interactions get the credit for a deal being created.
• Revenue attribution, which splits closed-won revenue across the interactions on the path.

Run all three. They will disagree, and the disagreement is the useful part: a channel can be excellent at creating contacts and hopeless at creating deals, which usually means the offer works and the qualification does not.

Timing matters as much as weighting. If your sales cycle runs six months, revenue attribution is a report on marketing you did two quarters ago, while contact create attribution is a report on last month. Teams that run revenue attribution on its own and defund the worst performer are often defunding the channel that filled the pipeline they are about to close.

The models, and what each one is claiming

One piece of vocabulary first, because it is where most people get lost. Several models weight "the creation interaction". That is the touch during which the contact or the deal was created. It sits wherever the record was created on the path, which is often nowhere near the beginning or the end.

HubSpot's model list moves as the product does, so open the model picker in your own portal and check it against this. These eight are the ones the arguments are usually about.

Model	How the credit splits	What it is good at
First interaction	100 per cent to the first touchpoint	Judging brand and acquisition work
Last interaction	100 per cent to the final touch before the conversion	Working out what closed it
J-shaped	20 per cent first, 60 per cent creation, 20 per cent shared evenly across the rest	Long paths where the creation event is the milestone
Inverse J-shaped	60 per cent first, 20 per cent creation, 20 per cent shared across the rest	Businesses that believe early exposure does the work
U-shaped	40 per cent first, 40 per cent creation, 20 per cent across the middle	Processes with a defined start and a defined creation event
W-shaped	30 per cent first, 30 per cent contact creation, 30 per cent deal creation, 10 per cent across the middle	B2B with three real milestones
Time decay	More credit the closer an interaction sits to the conversion	Long cycles that need repeated contact
Linear	Equal credit to every interaction on the path	Refusing to guess

First and last interaction: run them as a pair

First interaction hands everything to the touchpoint that introduced the person, so it tells you which campaigns and channels bring new people into your database, and which early efforts tend to precede a deal. Last interaction hands everything to the final touch before the conversion event, whether that event is a contact being created, a deal being created or revenue being realised.

Both are easy to explain and easy to mislead with. First interaction makes your awareness work look like the whole business. Last interaction makes your demo request form look like the whole business, when the form's actual job is to be last. Run the two on the same date range and read the gap between them, which is where the information is.

J-shaped and inverse J-shaped: the same three parts, opposite convictions

J-shaped gives 20 per cent to the first interaction, 60 per cent to the creation interaction and shares the remaining 20 per cent evenly across the other touches. For contacts, that means acknowledging the original source while putting most of the weight on the moment someone became a contact. For revenue, the closed-won amount lands mostly on the creation event, with the origin still carrying something.

Inverse J-shaped reverses the ends: 60 per cent to the first interaction, 20 per cent to the creation interaction, 20 per cent split across everything else. Closed-won revenue maps predominantly back to the channel that found the person.

Choose J-shaped when the creation event is what your team is measured on. Choose inverse J-shaped when you genuinely believe the introduction does the work and the rest is order taking. Then leave it alone, because switching between the two moves the first touch's share from a fifth to three fifths, and every channel ranking in your reporting moves with it.

U-shaped and W-shaped: the milestone models

U-shaped splits 40 per cent to the first interaction, 40 per cent to the creation event and 20 per cent across whatever happened in between. It suits a process with an obvious start and an obvious creation moment, and it leaves a little room for the middle without pretending the middle is decisive.

W-shaped adds the third milestone: 30 per cent to the first interaction, 30 per cent to contact creation, 30 per cent to deal creation, and 10 per cent shared across the interactions between the first touch and the deal. In B2B with defined lifecycle stages this is usually the most defensible model in the room, because all three of those moments genuinely have to happen for the revenue to exist.

Time decay: for cycles that need repeated contact

Credit increases the closer an interaction sits to the conversion, and earlier touches fade out. Built for long sales cycles where persistence is the job, and useful when you need to justify recent activity as the deciding factor. Be honest about the trade: it flatters whatever you did most recently, so it rewards activity and undervalues patience.

Linear: the honest option while you fix your tracking

Equal credit to every interaction, on the basis that you cannot know. With six touches on a path that is informative. With sixty, every channel becomes a rounding error and the report says nothing. It is a good default while your tracking is still being fixed, because it makes no claim you will have to defend in front of a CFO.

When paid, organic, partner and outbound all touch the same deal

Before anyone argues about weightings, check three things.

What is actually recorded. Every model works from the interactions sitting on the record. A partner referral that arrived as a phone call, or an outbound conversation that lived in a rep's inbox and never got logged, cannot be credited by any model, because it is not in the data. Most complaints that attribution is broken are logging problems wearing a modelling costume.

How many contacts your deals carry. Revenue attribution works from the interactions of the contacts associated with the deal. If your reps associate one contact per deal while five people in the buying group read your emails, you are reporting one person's journey and calling it the account's.

Whether your UTMs were ever written down. More on that below, and it is the cheapest of the three to fix.

Then use the technique that survives a budget meeting: run two models on the same date range, one that favours the start of the path and one that favours the end, and treat the difference as the finding. If last interaction says a paid campaign is your best channel and first interaction says it barely exists, you are paying to convert people who already knew who you were. There is more on joining that up in tracking your ads from click to closed deal, and on the reporting side of the paid stack in turning paid media into measurable pipeline.

Setting it up so the report is worth reading

Attribution reporting is only as good as the tagging underneath it, and tagging is a discipline problem.

Agree naming conventions for source, medium and campaign, and keep them short and lowercase. Source is where the click came from, medium is the type of traffic, campaign is the thing you are running. Use Google's Campaign URL Builder to construct the URLs, then click every tagged link once before the campaign goes out. A broken tagged URL is invisible until the report comes back empty.

1. Tag every link that leaves your control.
2. Keep one document of the tags in use, somewhere a new hire can find it.
3. Review it quarterly and delete what nobody uses.
4. Change the campaigns on the back of what it tells you, which is the only reason to do any of it.

Google Analytics is worth running alongside HubSpot for cross-channel context, with one caveat: it is a Google product, so it reads Google platforms generously. Treat it as a second opinion. Analytics 360 adds more if you have the budget for it.

Third-party attribution tools are worth it when you need modelling these models do not do. Most portals are not there yet, and buying one before the tagging and the associations are right just gives you a more expensive version of the same missing data.

One practical note if you report by quarter, financial year or renewal period: the deal has to carry that period as a property, or every report becomes a manual date-range exercise. Here is the method for automatically creating deals based on reporting period properties.

The GST mistake underneath every ROI number

No law governs attribution modelling. There is no regulator, no Australian standard, no compliance answer, and anyone implying otherwise is selling something.

The Australian problem here is accounting. GST is 10 per cent. Ad spend invoiced by an Australian entity includes it. Deal amounts in HubSpot are whatever your team typed, and most sales teams type the GST exclusive figure, because that is the revenue number the P&L runs on. Divide one by the other and the spend side of the sum is inflated by ten per cent, so every channel reads worse than it is, every month, by the same amount. Record deal amounts GST inclusive against GST exclusive spend and you get the opposite error, with the channel reading better than it is.

Ten per cent is not a rounding error at the point where someone is deciding whether a channel survives the next budget.

So pick one convention and write it at the top of the report. We take GST out of both sides, because that matches how finance reports revenue and it makes the attribution number and the P&L number arguable in the same meeting. If you sell in more than one currency, check the exchange rate HubSpot is using to convert into your company currency before you compare an offshore channel with a domestic one.

If you want real market context for any of this, the ABS publishes actual figures on how Australian businesses use technology. That is a better source than the unattributed percentages that circulate in attribution articles.

What attribution will never tell you

The conversation at a conference. The recommendation in a group chat. The person who heard your name on a podcast, typed it into Google three weeks later and arrives in your reports as direct traffic with no history at all.

Attribution reports what was recorded, so give it the decisions it can carry: stop the channels that are obviously returning nothing, defend the ones that obviously are, and give up on allocating the last five per cent. If you want someone to sit inside your portal and get the associations, the tagging and the deal amounts right before you model anything, that is most of what our RevOps and HubSpot work involves.

Here is the exercise worth running this week. Pull contact create attribution and revenue attribution for the last twelve months, same date range, side by side. If the two lists rank your channels differently, which one is next year's budget built on?