£18,000 goes into Google Ads in January. It produces 60 valid enquiries and no signed contracts by month-end. The campaign therefore appears to have generated 0x revenue.
Five months later, three deals from that January cohort close for £102,000. Its realised lead generation ROAS is 5.67x. Waiting until June to judge January is accurate but too late for February's budget decision. Calling it 0x in February is prompt but wrong.
The practical answer is a lead value model. Give each open lead an expected value based on what similar leads historically became. As evidence accumulates, replace that estimate with a stage-weighted value, then with actual won revenue. Forecast and fact stay separate.
Lead generation ROAS needs more than one number
Ordinary ROAS is simple:
Revenue attributed to the campaign ÷ campaign spend = ROAS
The calculation only becomes final when the sales cohort is mature. Until then, zero revenue does not mean zero return. It means the outcome is unresolved.
A useful report separates three views:
- Stage-weighted pipeline ROAS: expected revenue represented by open leads and opportunities, divided by the spend that acquired the cohort.
- Forecast cohort ROAS: won revenue plus the expected value of everything still open, divided by cohort spend.
- Realised revenue ROAS: net revenue from closed-won deals, divided by cohort spend.
Cost per lead still matters, but it measures acquisition efficiency, not commercial return. A £200 lead that never qualifies is expensive. A £900 lead that reliably produces £30,000 contracts may be cheap.
The answer is not to assign the full average contract value to every form fill. If the average deal is £40,000 and 5% of valid enquiries become customers, a new enquiry has an expected revenue value of £2,000, not £40,000. Applying the full amount would inflate expected return twenty-fold.
Lead valuation also does not repair attribution. It assumes the spend and leads in the cohort have already been identified correctly. If that premise is doubtful, start with why reported Google Ads ROAS can mislead and the wider B2B measurement framework for long sales cycles. The job here is narrower: valuing an unresolved lead without pretending it is revenue.
Build lead values from closed deals
A defensible model starts with lead-created cohorts old enough for the normal outcome period to have elapsed, not percentages chosen in a CRM settings screen.
First, choose one revenue basis. Net first-year contract revenue, excluding VAT, refunds and pass-through costs, is usually more defensible than speculative lifetime value. Recurring revenue can be included beyond year one only when retention data supports it. Use the same basis at every stage.
Then calculate the empirical probability of a lead at each stage eventually becoming closed-won:
Stage-to-win rate = mature cohort records that reached the stage and later won ÷ all mature cohort records that reached the stage
The modelled value is:
Stage value = stage-to-win rate × expected net revenue if won
If a credible opportunity amount exists, use that amount. Before one exists, use the mean won revenue for the same cohort, offer or deal band.
Suppose 8% of historically resolved marketing-qualified leads became customers and their mean first-year revenue was £36,000. Each comparable lead at that stage is worth:
8% × £36,000 = £2,880 expected revenue
Use cumulative stage-to-win rates. If 50% of sales-qualified leads reach proposal and 40% of proposals win, the sales-qualified-to-win probability is 20%, not 50%. Many CRM defaults use tidy values such as 10%, 25%, 50% and 75%. Unless those values reproduce observed outcomes, they are labels rather than evidence.
Do not automatically replace a skewed mean with the median. Expected portfolio value is based on an arithmetic mean. If one £300,000 deal distorts a book of £20,000 contracts, separate enterprise and mid-market deals, cap the outlier in a downside case, or show both cases. Changing to the median answers a different question.
Segment only where economics genuinely differ and the sample supports it. Offer, company size and commercial intent can justify separate values. Individual campaigns or keywords rarely have enough wins to support their own probabilities. A B2B keyword strategy organised around commercial intent can preserve useful groupings without inventing precision at keyword level.
Stage-weighted pipeline ROAS: a worked example
Take an illustrative B2B consultancy with £36,000 mean first-year revenue per win. Its mature CRM data shows these eventual win rates:
| Current status | Records in campaign cohort | Historical win rate | Value per record | Weighted value |
|---|---|---|---|---|
| Disqualified or lost | 30 | 0% | £0 | £0 |
| Valid enquiry, not yet qualified | 10 | 4% | £1,440 | £14,400 |
| Marketing-qualified lead | 10 | 8% | £2,880 | £28,800 |
| Sales-qualified lead | 7 | 20% | £7,200 | £50,400 |
| Proposal issued | 3 | 35% | £12,600 | £37,800 |
| Total | 60 | £131,400 |
The campaign cost £18,000. At this 45-day snapshot, none of the opportunities has closed. Its stage-weighted pipeline ROAS is:
£131,400 ÷ £18,000 = 7.3x
That is expected revenue, not £131,400 in the bank. Multiplying all 30 open records by the £36,000 average would produce a fictional £1.08 million pipeline and a 60x return. Stage weighting recognises that most open leads will not buy.
Now let the cohort mature. Three deals close for £42,000, £36,000 and £24,000:
(£42,000 + £36,000 + £24,000) ÷ £18,000 = 5.67x realised ROAS
The day-45 model predicted £131,400; £102,000 materialised. Actual revenue was 77.6% of the forecast. One cohort could be ordinary variance. If mature cohorts repeatedly land near that ratio, reduce the probabilities or investigate where the model is optimistic.
This comparison is the point of modelling. The forecast supports an earlier decision; realised revenue calibrates the next forecast.
Keep pipeline ROAS honest
A stage model can become another flattering dashboard unless five controls are enforced.
Count each record once, at its current stage. A sales-qualified lead in the example is worth £7,200, not the sum of every value it passed on the way there. Adding £1,440, £2,880 and £7,200 would report £11,520 for the same lead.
If systems record every progression as a separate value event, send only the incremental uplift: £1,440 at valid enquiry, another £1,440 on becoming marketing-qualified, then £4,320 on becoming sales-qualified. A lost lead also needs its remaining expected value written back to zero. Recording only positive progress guarantees inflation.
Never rename expected value as revenue. Pipeline ROAS, forecast cohort ROAS and realised ROAS should appear as separate lines. A board pack that says “£131,400 generated” before anything has closed is inaccurate, however plausible the forecast.
Keep numerator and denominator attached to the same cohort. Do not divide this quarter's open pipeline, some of which came from earlier activity, by this quarter's spend. Review leads according to when their acquisition spend occurred, then let that cohort mature.
Check for double weighting. Some CRM pipeline fields already multiply opportunity amount by a probability. Applying the stage rate again discounts the same deal twice. Store or identify both the unweighted amount and the weighted forecast.
Show uncertainty. A 20% win rate based on five resolved opportunities is not equivalent to one based on 500. When samples are thin, pool similar stages and publish low, base and high cases. Stale proposals should receive a lower value only if historic results show that age reduces win probability.
Keep every as-of forecast rather than recalculating old snapshots with newer probabilities. Otherwise, the reporting quietly rewrites history and conceals forecast error.
A credible report states the cohort date, revenue basis, stage definitions, resolved sample sizes and model version. If an external report omits those inputs, use this practical audit of agency ROAS reporting to challenge the number rather than its chart design.
Turn lead generation ROAS into budget decisions
There is no universal good ROAS. Finance needs to set the required return from margin, sales cost, cash flow and the contribution the company expects to retain.
A simple revenue hurdle starts with the maximum share of revenue the business will spend on advertising:
Required revenue ROAS = 1 ÷ allowable ad-spend share of revenue
If advertising may consume no more than 20% of first-year revenue, the hurdle is 5x.
That target turns expected lead value into a maximum cost per lead:
Maximum CPL = expected revenue per valid lead ÷ required ROAS
In the worked example, a new valid enquiry is worth £1,440. At a 5x target, the maximum CPL is £288. The campaign's initial CPL was £300, so lead volume alone made it look marginal. Its later stage mix produced 7.3x expected pipeline ROAS, and its final 5.67x realised ROAS cleared the hurdle.
Do not scale from a base case alone. Reducing every open-stage probability in the example by 20% gives:
(£131,400 × 80%) ÷ £18,000 = 5.84x
That still clears a 5x hurdle, although not a 6x hurdle. The decision follows:
- Scale when a conservative case clears the required return and the data is sufficiently mature.
- Hold and test when the base case clears it but the conservative case does not.
- Rework or pause when a mature base case remains below it.
ROAS is still a revenue measure, not profit. Where offers have materially different margins, add an expected contribution view: weighted revenue multiplied by contribution margin, divided by spend. Label it separately.
The same commercial rules should govern an internal team, ongoing Google Ads management tied to qualified pipeline or a fixed-scope Google Ads project. Ongoing optimisation only becomes rational once lead definitions and values are stable; a contained repair is more proportionate when those inputs are the main problem.
FAQ
What is lead generation ROAS?
It is closed-won net revenue from a lead-acquisition cohort divided by the advertising spend that produced those leads. While deals remain open, report stage-weighted pipeline or forecast cohort ROAS and label it as expected, not realised.
How do you calculate the value of a B2B lead before it closes?
Multiply the historical probability of a comparable lead winning from its current stage by the expected net revenue if it wins. Use mature outcomes, a consistent revenue period and an actual opportunity amount where one is available.
Should pipeline ROAS use the full value of open opportunities?
No. Full pipeline value assumes every opportunity will close. Multiply each unweighted opportunity amount by its empirical stage-to-win probability, then sum the results. Count each record at one current stage.
How often should stage values be updated?
Review them when enough new outcomes have matured to change the evidence, and after material changes to pricing, offer or qualification. Quarterly may suit a high-volume business; a lower-volume firm may need a longer interval. Stability is better than reacting to every isolated win.
Summary
- Treat unresolved lead value as a forecast, never as booked revenue.
- Derive stage values from mature win rates and a consistent net-revenue basis.
- Calculate pipeline ROAS from mutually exclusive current stages.
- Compare forecast cohorts with realised revenue to calibrate the model.
- Set CPL and budget limits from the company's required commercial return.
Actualyse builds measurement and attribution setups that tie B2B ad spend to real revenue. Book a call to talk through where yours stands.
