Last-click is the right default for more low-volume B2B firms than most attribution vendors would admit. It becomes wrong when a long buying journey makes the final recorded interaction merely the easiest one to see. In a B2B Google Ads programme, revenue attribution models are not competing estimates of the same truth: each answers a different question, and choosing one chooses which question your budget meeting will be about.
A model redistributes credit. It does not create or remove signed revenue.
A form fill is not revenue. Connecting acquisition records to closed CRM revenue is the prerequisite. The separate problem of finding the commercial return behind reported Google Ads ROAS also needs settling before the weighting debate begins.
Six revenue attribution models and their questions
The useful taxonomy starts with the management decision, not the sophistication of the calculation. The following six-model taxonomy assigns exactly one question and one decision to each model.
- First-click attribution
Question: Which recorded source introduced the account?
Decision it can inform: Whether to fund channels that create recorded entry into the buying journey.
First-click is useful when acquisition is the disputed job. It becomes misleading when an earlier recommendation, event or colleague introduced the company but never entered the tracking record.
- Last-click attribution
Question: Which recorded interaction captured demand immediately before the chosen conversion?
Decision it can inform: Which demand-capture campaigns, pages or nurture steps should receive short-term budget.
Last-click is blunt, explainable and difficult to manipulate. For short journeys, those qualities often matter more than theoretical completeness.
- Linear attribution
Question: Which recorded touches participated in the successful journey?
Decision it can inform: Which channels require continued coverage when no defensible touch hierarchy exists.
Linear attribution gives participation equal weight. It does not claim every touch had equal influence, despite reports often being read that way.
- Time-decay attribution
Question: Which recorded touches occurred closest to the revenue event?
Decision it can inform: Whether late-stage nurture and demand-capture activity deserves greater operational priority.
Recency is the assumption. A decisive early comparison page can therefore receive less credit than a routine reminder sent later.
- Position-based attribution
Question: Which recorded touches opened and closed the journey?
Decision it can inform: How budget should be balanced between creating entry and capturing the eventual conversion.
Position-based models suit teams that consider the middle important but not decisive. The weighting remains a policy choice, not an observed fact.
- Data-driven attribution
Question: Which eligible recorded Google Ads interactions should receive value for the selected conversion objective?
Decision it can inform: In-platform bidding and campaign allocation when reliable CRM-derived conversion values reach Google Ads.
DDA is an operational optimisation model, not an automatic cross-channel revenue truth. Its allocation implications belong in our dedicated explanation of DDA and ROAS.
Journey architecture also changes what can be recorded. Two examples from our case studies make the point: Lanteria’s broad capability set was routed for several stakeholder audiences, while AfriCap Hub used a catalogue, filtering and registration journey. Neither project publishes revenue figures; the relevance here is the different path each architecture creates.
Model choice is question choice.
One 8-touch B2B journey, three incompatible answers
A single illustrative deal makes the disagreement visible. Assume a B2B software buyer signs a £24,000 contract after these eight recorded touches:
| Touch | Recorded interaction |
|---|---|
| 1 | Non-brand Google Ads click |
| 2 | Organic search guide visit |
| 3 | LinkedIn paid click |
| 4 | Email article click |
| 5 | Webinar registration |
| 6 | Direct case-study return |
| 7 | Brand Google Ads click |
| 8 | Demo-booking email click |
Three deterministic models allocate the same £24,000 as follows:
| Model | Touch 1 credit | Touches 2–7 credit | Touch 8 credit | Total |
|---|---|---|---|---|
| First-click | £24,000 | £0 each | £0 | £24,000 |
| Last-click | £0 | £0 each | £24,000 | £24,000 |
| Linear | £3,000 | £3,000 each | £3,000 | £24,000 |
The linear arithmetic is reproducible:
£24,000 closed-won revenue ÷ 8 recorded touches = £3,000 per touch
First-click reports £24,000 for touch 1 versus £0 for touch 8. Last-click reports £0 versus £24,000. Linear reports £3,000 versus £3,000.
Each answer is internally correct because each answers its own question. First-click identifies entry. Last-click identifies capture. Linear records participation. Actual closed-won revenue remains £24,000 under all three.
DDA is deliberately absent from this numerical split. Inventing a DDA allocation without the relevant Google Ads history would create false precision and drift into mechanics covered by the separate article.
Credit moves; won revenue does not.
Choose a default by cycle length and touch count
Use two inputs from the latest 20 closed-won deals: median days from first recorded marketing touch to signature, and median recorded marketing-touch count. With fewer than 10 won deals, retain last-click and label the selection provisional.
Place the company into one of three bands. When the inputs reach different bands, the more complex band wins.
| Measurable trigger | Default model | Budget-meeting question |
|---|---|---|
| Sales cycle of 30 days or fewer and 3 touches or fewer | Last-click | Which activity captures demand? |
| Sales cycle of 31–90 days or 4–7 touches, provided neither input reaches the next band | Position-based | Are entry or capture being underfunded? |
| Sales cycle above 90 days or at least 8 touches | Linear | Which recorded channels require coverage? |
DDA sits outside this map because sales-cycle length does not determine whether Google Ads should use it for optimisation. Use DDA for that named platform decision when closed-won values are stable and imported; do not make it the board’s universal revenue model.
First-click and time-decay should usually be challenger views. Run one when its specific question is under dispute, then remove it from the recurring meeting once the decision is made.
Our claim is falsifiable: within the simple band, position-based attribution should not change the keep-or-cut decision twice across the next three monthly reviews; if it does, our claim is wrong and last-click should lose default status.
The honest limit here is that cycle length and touch count are confounded by three factors: consent loss, CRM hygiene and buying-committee size. The thresholds are operating rules, not natural laws.
Campaign design then returns to a B2B Google Ads strategy built around commercial intent, rather than asking attribution to repair weak targeting.
Complexity earns its place only when it changes a decision.
If the path from ad click to signed revenue still has gaps, Actualyse will trace each hand-off with you — book a call
The model changes the budget story, not the revenue
A finance lead should be able to reproduce the disagreement on one sheet of paper. Take an illustrative UK cybersecurity consultancy with these five labelled inputs:
- Company type: UK cybersecurity consultancy.
- Measurement period: One month.
- Non-brand Google Ads spend: £5,000.
- Brand Google Ads spend: £3,000.
- Actual closed-won revenue: £40,000.
Monthly total spend = £5,000 + £3,000 = £8,000
Assume two model outputs. First-click assigns £32,000 to non-brand and £8,000 to brand. Last-click reverses that allocation, assigning £8,000 to non-brand and £32,000 to brand. Both reconcile to the same £40,000.
First-click non-brand ROAS: £32,000 ÷ £5,000 = 6.40×
First-click brand ROAS: £8,000 ÷ £3,000 = 2.67×
Last-click non-brand ROAS: £8,000 ÷ £5,000 = 1.60×
Last-click brand ROAS: £32,000 ÷ £3,000 = 10.67×
Blended ROAS under either model: £40,000 ÷ £8,000 = 5.00×
At a declared 4.00× hurdle, first-click says non-brand passes at 6.40× versus brand failing at 2.67×. Last-click says brand passes at 10.67× versus non-brand failing at 1.60×.
The model has reversed the campaign decision without changing spend, contracts or total revenue. Neither allocation proves whether the buyer would have converted without brand or non-brand activity.
A clear decision contract should also anchor specialist Google Ads management, otherwise the most flattering model will quietly win.
A weighting rule can reverse allocation without creating revenue.
What no revenue attribution model can do
Missing evidence is where false confidence begins. No model can allocate credit to touches that were never recorded or could not be joined to the eventual deal.
Six examples include a colleague’s recommendation, a forwarded PDF, a private Teams message, a podcast mention, a trade-show conversation and a first visit where the buyer withheld consent.
Four common fixes fail for different reasons:
- Changing a 30-day window to 90 days can include older recorded touches, but it cannot recover an invisible recommendation or unidentified visit. More time does not manufacture missing evidence.
- Adding more UTM parameters improves labels on future links, but it does not join separate people, devices and offline conversations to one buying committee. Better labels are not better identity.
- Replacing last-click with DDA redistributes observed credit, but it cannot establish incrementality. Neither model reveals the counterfactual sale that would have happened without the touch.
- Expanding a report from 6 channels to 14 and changing its colour scheme alters presentation, but it does not repair overwritten CRM sources or absent closed-won values. Granularity cannot correct the underlying record.
No attribution model measures persuasion directly. Sales effort, pricing, procurement friction and competitor failure can determine the contract while receiving no marketing credit at all.
Unobserved influence cannot be recovered by redistributing observed credit.
FAQ
Four implementation decisions settle the common edge cases.
Can qualified pipeline be used before deals close?
Use closed-won revenue for irreversible budget cuts. Keep qualified pipeline as a separate early indicator: £1 of pipeline is not £1 of revenue. If no deal can close within 60 days, pipeline may guide reversible bid adjustments, but not channel removal.
How often should an attribution model change?
Freeze the selected model for 90 days. Reopen the decision earlier only after a material conversion-definition change or when the measured sales-cycle band changes. Frequent switching turns normal variance into model shopping.
What should we do when first-touch CRM fields are incomplete?
If first touch is missing from more than 10% of closed-won deals, suspend first-click and position-based reporting. Use last-click temporarily, then repair capture before restoring models that require reliable entry data.
Should renewals share the acquisition model?
Separate new-business and renewal revenue when renewals exceed 20% of booked revenue. Otherwise, old acquisition touches can keep receiving credit for commercial work performed by customer success or account management.
Governance prevents model-shopping from becoming budget policy.
Summary
Apply these six rules:
- Default to last-click at 30 days or fewer and no more than 3 recorded touches.
- Default to position-based attribution at 31–90 days or 4–7 recorded touches.
- Default to linear attribution above 90 days or from 8 recorded touches.
- Use DDA for Google Ads optimisation only when reliable closed-won values are imported.
- Reject entry-weighted models when more than 10% of won deals lack first-touch data.
- Freeze the chosen model for 90 days unless its outcome definition materially changes.
Budget questions deserve explicit rules.
Actualyse builds measurement and attribution setups that tie B2B ad spend to real revenue. Book a call to talk through where yours stands.
