What Multi-Touch Attribution Is
What is multi-touch attribution?
Multi-touch attribution is a method of dividing the credit for a conversion, such as a new lead, a deal or revenue, across every marketing and sales interaction that preceded it, instead of giving all the credit to a single touch. Each model applies a different rule for that split, so the same deal can make different channels look most valuable.
A B2B buyer rarely converts on the first visit. They click an ad, read a few articles, attend a webinar, ask for a demo and talk to sales over several weeks. Single-touch attribution picks one of those moments and ignores the rest. Multi-touch attribution tries to reflect the whole path. The catch is that there is no objectively correct way to split the credit: every model is a rule someone chose, and the rule you choose shapes where the budget goes.
The Nine Models, Side by Side
Most of the RevOps agencies that rank for this topic build on HubSpot, which publishes the exact rule behind each model in its attribution reporting documentation. Here they are in one table.
| Model | How credit is split | Tends to favour |
|---|---|---|
| First interaction | 100% to the first interaction | Awareness channels: paid social, display, events |
| Last interaction | 100% to the last interaction before conversion | Closing touches: branded search, email, direct |
| Linear | Equal share to every interaction | Channels that appear often, however minor |
| U-shaped | 40% first, 40% lead conversion, 20% across the rest | Whatever starts the journey and creates the lead |
| W-shaped | 30% first, 30% contact creation, 30% deal creation, 10% across the rest | Touches at the three hand-off points |
| Full path | 22.5% each to first, contact creation, deal creation and last; 10% across the middle | The full funnel, including the closing touch |
| Time decay | More credit to recent touches, on a 7-day half-life | Late-stage sales and retargeting |
| J-shaped | 20% first, 60% conversion, 20% across the rest | Conversion-stage touches |
| Inverse J-shaped | 60% first, 20% conversion, 20% across the rest | Demand creation |
The table is easier to judge when you can see one deal move. The simulator applies those published rules to an example six-touch journey. Watch what happens to the LinkedIn ad, the first touch, as you switch from inverse J-shaped to time decay.
Interactive simulator
One deal, nine attribution models
An example $24,000 deal with six touches. Pick a model to see how HubSpot's published rules split the credit.
U-shaped: 40% first, 40% lead conversion, 20% split across the rest.
- LinkedIn ad click42d before close · First touch40.0% · $9,600
- Blog post visit30d before close5.0% · $1,200
- Webinar attended21d before close5.0% · $1,200
- Demo form submitted14d before close · Contact created40.0% · $9,600
- Discovery call7d before close · Deal created5.0% · $1,200
- Case study email1d before close · Last touch5.0% · $1,200
Model rules from HubSpot's attribution reporting documentation. The journey is illustrative. In HubSpot, W-shaped is available on deal and revenue reports, and full path on revenue reports only.
On a 42-day journey, time decay gives the first touch almost nothing, because a 7-day half-life halves its weight six times over. Inverse J-shaped gives it 60%. Neither is wrong. They answer different questions: “what starts demand?” and “what closes it?” The practical advice is to pick the model that matches the decision you are making, write that choice down, and compare at least two models before moving budget.
Why Google Retired Four Models
Which attribution models did Google remove from Google Ads and GA4?
In 2023 Google retired the first click, linear, time decay and position-based attribution models from Google Ads and Google Analytics 4. Conversion actions using them were switched to data-driven attribution from September 2023, with last click remaining as the only rules-based alternative. Google said these four models accounted for under 3% of Google Ads web conversions.
Google's own announcement argued that rules-based models do not adapt to changing customer journeys, and that data-driven attribution, which uses machine learning across the advertiser's conversion data, does. That may be true inside Google's properties, but it has a limit worth understanding: an ad platform can only model the touches it can see. It sees ad clicks and site visits. It does not see the discovery call, the procurement review or the deal that closed for half the expected value.
That is why B2B teams increasingly run attribution in the CRM rather than in the ad platform. The CRM is where the deal, its value and its close date actually live. Platform attribution is useful for optimising bids within one platform; CRM attribution is what you need to compare platforms against each other and against organic channels.
Where Attribution Data Breaks
Every model is only as good as the touches it receives. In practice, most attribution errors are data errors rather than model errors. These are the common failure points:
- Inconsistent UTM parameters. “LinkedIn”, “linkedin” and “li” become three channels. A written naming convention, enforced by a link builder rather than by memory, prevents most of this.
- Source lost at the form. If the form does not capture the UTM values and the referring page into hidden fields, the contact arrives in the CRM as “direct” or “offline”.
- Mobile tracking limits. Since Apple's App Tracking Transparency framework arrived with iOS 14.5 in 2021, apps must ask permission before tracking users across other companies' apps and websites, which reduces what ad platforms can observe.
- Offline and sales touches. Calls, meetings and referrals only count if someone logs them, and anything logged days later by hand is less reliable than an automatic capture.
- Contacts not linked to deals. Revenue attribution needs every buying-group contact associated with the deal. A deal linked to one contact credits only that person's journey.
- Duplicate and junk leads. Duplicates double-count touches; spam submissions inflate lead counts and make cheap channels look better than they are.
How to Set Up Multi-Touch Attribution in a CRM
How do you set up multi-touch attribution for a B2B SaaS company?
Define a UTM naming convention, capture UTM values and the first page visited on every form, store them on the contact and keep them unchanged, associate every buying-group contact with the deal, and record the real closed value. Only then choose a model, and compare it against at least one other before moving budget.
Name things once
One written UTM convention for source, medium and campaign, used by everyone.
Capture at entry
Hidden form fields store UTMs, landing page and referrer on the contact.
Connect the deal
Associate every buying-group contact with the deal and record its real value.
Then pick a model
Choose the model that fits the decision, and compare it with a second one.
- 1
Write the UTM convention down
Lower-case values, a fixed list of sources and mediums, and a campaign naming pattern. Generate links with a builder so nobody types them by hand.
- 2
Persist original source on the contact
Store first-touch values in fields that never get overwritten, and keep latest-touch values separately.
- 3
Log sales touches automatically
Sync meetings and calls from calendar and dialler tools rather than relying on manual entry.
- 4
Associate contacts and companies with deals
Revenue attribution only counts journeys of contacts linked to the deal, so make association part of the sales process.
- 5
Quarantine junk before reporting
Flag spam and duplicates before they reach attribution reports, and keep the quarantine reversible.
- 6
Review quarterly against closed-won deals
Check a sample of won deals by hand. If the reported path does not match what sales remembers, fix the capture, not the model.
Loomstrat builds this layer for SaaS teams through its RevOps and growth systems service, and runs it in its own products: UTMLoop for UTM link tracking, Loomstrat CRM for source capture and Loomstrat Analytics for closed-loop reporting. Our closed-loop ROAS guide covers the reporting side in detail.
Lead Scoring That Sales Trusts
What is lead scoring and how should it work with attribution?
Lead scoring ranks leads by how likely they are to buy, usually by combining fit (does the company match your ideal customer?) with engagement (are they showing buying behaviour now?). Attribution explains which channels produced revenue; scoring decides which current leads sales should contact first. Both depend on the same captured data.
Most lead scores fail for one of two reasons: they reward activity that does not predict buying, such as opening newsletters, or they are never checked against deals that actually closed. A useful score keeps the two dimensions separate.
| Fit score | Engagement score | |
|---|---|---|
| Question it answers | Could this company become a good customer? | Is this person showing buying intent now? |
| Typical signals | Industry, company size, role, region, tech stack | Pricing page visits, demo requests, product sign-ups, replies |
| How it changes | Rarely, when firmographic data changes | Constantly, and should decay when activity stops |
| Common mistake | Scoring job titles nobody in the buying group holds | Scoring newsletter opens as heavily as demo requests |
There is also a structural reason not to over-trust engagement. Research by John Dawes of the Ehrenberg-Bass Institute for LinkedIn's B2B Institute, summarised in Marketing Week, estimates that up to 95% of B2B buyers are not in the market at any given time, because companies change major suppliers only every few years. A high fit score with low engagement is usually a future buyer, not a bad lead: route it to nurture rather than to sales, and let engagement decide when it moves.
Finally, test the score. Every quarter, compare the scores leads had when they arrived with which ones became closed-won deals. If high-scoring leads do not close more often than low-scoring ones, change the weights.
Speed to Lead: What the Research Shows
How quickly should sales respond to a new lead?
Within an hour. Research published in Harvard Business Review found that firms that tried to contact a lead within an hour were nearly seven times as likely to qualify it as those that waited even one more hour, and over 60 times as likely as firms that waited 24 hours or longer.
The study, “The Short Life of Online Sales Leads” by James Oldroyd, Kristina McElheran and David Elkington, had two parts. An audit sent web leads to 2,241 US companies and timed the replies. A separate analysis of about 1.25 million leads produced the seven-times and sixty-times findings. The audit showed how few companies were fast:
Almost a quarter of companies never replied at all. The research is from 2011, but the mechanism has not changed: a lead is most reachable just after it has asked to be contacted. The fix is operational rather than motivational: route leads automatically by territory or score, alert the owner instantly, and track response time as a visible metric with an SLA.
RevOps for SaaS: Agency, Build or In-House?
Teams searching for a “HubSpot RevOps agency” usually need one of three different things, and they are priced and staffed differently.
| Option | What you get | Best when |
|---|---|---|
| CRM-specialist agency | Configuration of your existing CRM: properties, workflows, reports | Your data and tools are fine; the CRM is just not set up |
| Custom build | Integrations, data pipelines and reporting your CRM cannot do alone | Data lives in several systems, or the CRM reports cannot answer your questions |
| In-house RevOps hire | Ongoing ownership of process, data quality and reporting | You have enough volume to need someone full-time, permanently |
The common pattern for SaaS companies is to fix the data capture first (that is a project), then hire or contract for ongoing ownership (that is a role). Doing it the other way round leaves a new RevOps hire spending their first months cleaning data instead of improving the funnel.
Frequently Asked Questions
What is multi-touch attribution?
A method of dividing credit for a conversion, such as a lead, a deal or revenue, across every interaction that came before it, instead of giving all credit to one touch. Each model uses a different rule, so the same deal can make different channels look most valuable.
Which attribution model is best for B2B SaaS?
There is no single best model. W-shaped and full-path models suit B2B because they credit the hand-off points: first touch, contact creation and deal creation. Choose the model that matches the decision you are making, and compare it with a second model before moving budget.
What is the difference between U-shaped and W-shaped attribution?
In HubSpot, U-shaped gives 40% to the first interaction, 40% to the lead conversion and 20% to the rest. W-shaped gives 30% each to the first interaction, contact creation and deal creation, and 10% to the rest.
Why did Google remove attribution models?
Google retired first click, linear, time decay and position-based models in 2023, saying rules-based models do not adapt to changing journeys and that under 3% of Google Ads web conversions used them. Data-driven attribution became the default, with last click as the alternative.
How does time decay attribution work in HubSpot?
It gives more credit to interactions closer to the conversion, using a 7-day half-life: an interaction 8 days before conversion receives half the credit of one 1 day before.
What should a B2B lead score include?
Two separate parts: a fit score based on firmographics such as industry, size and role, and an engagement score based on buying behaviour such as pricing visits and demo requests, which decays when activity stops. Review the weights quarterly against closed-won deals.
How fast should sales follow up on a lead?
Within an hour. Research in Harvard Business Review found firms contacting leads within an hour were nearly seven times as likely to qualify them as firms that waited one more hour, and over 60 times as likely as those waiting 24 hours or more.
Do I need a RevOps agency or a custom build?
If your data is already in one CRM and it is simply not configured, a CRM-specialist agency is usually enough. If data sits in several systems or the CRM cannot answer your questions, you need integrations or a custom reporting layer built.
Attribution models get most of the attention, but the model is the last decision, not the first. Clean capture, deals linked to the right contacts, real closed values and a quarterly check against what sales actually saw will improve the numbers more than switching from one model to another.