How to Measure Marketing When a B2B Sales Cycle Runs Nine Months
Which lead indicators can stand in for revenue on a nine-month B2B sales cycle, how to validate them by cohort, and how to test and model pipeline in Meridian or Robyn.
On this page
- Why long B2B cycles break the attribution habits ecommerce built
- Which lead indicators can stand in for revenue
- A lag table from first touch to closed-won
- How to validate each indicator against closed-won revenue by cohort
- A holdout design that reads pipeline instead of revenue
- Modeling pipeline as the KPI in Meridian or Robyn
- What to report while the revenue is still months away
- Sources
Marketing measurement for a B2B company with a nine-month sales cycle has to run on lead indicators, because the revenue a campaign helps create closes roughly 270 days after the first touch, when the budget for the next two quarters is already set. The workable approach has three parts: pick indicators that sit early in the cycle, prove that each one predicts closed-won revenue by cohort, and test whether marketing causes those indicators with a holdout that reads pipeline. Closed-won revenue still decides whether the program worked. It arrives late, so the job in the meantime is to build indicators you have earned the right to trust.
Why long B2B cycles break the attribution habits ecommerce built
B2B long sales cycle marketing measurement differs from ecommerce attribution in one structural way: the click and the revenue sit months apart, and ad platforms connect them inside windows measured in days or weeks. In Google Ads, the click-through conversion window for Search and Display campaigns can be set from 1 to 30, 60 or 90 days depending on the conversion source, with a 30-day default, according to Google. Offline conversion import has the same ceiling. Google's FAQ says "We retain the GCLID for only 90 days," and when a sale happens later than that, it advises uploading an earlier offline conversion event that falls within the 90 days (Google).
On a nine-month cycle, then, the closed deal never makes it back to the platform as a click-matched conversion. Whatever the bidding system optimizes is already a proxy, usually a qualified lead. The same FAQ calls a separate conversion action for each funnel stage best practice and notes that you would usually bid to only one stage. An ecommerce team can let the platform read revenue directly, while a B2B team has to choose its proxy on purpose and check it.
The buyer pool adds a second gap. LinkedIn's B2B Institute, which cites its joint research with the Ehrenberg-Bass Institute, states that "95% of your potential buyers aren't ready to buy today" and gives the example that "80% of companies change banking services once every 5 years" (LinkedIn). Advertising that reaches those future buyers can show up in revenue a year or more later, outside any click window. On cycles this long no attribution model sees the whole path, so choosing one is a reporting decision more than a measurement one. Our explainer on attribution models and their limits covers how each model assigns credit, our attribution model decision guide matches models to data volume and sales cycle, and the Attribution Doctor runs a simple media-mix model on monthly spend and revenue by channel in your browser.
Which lead indicators can stand in for revenue
A useful lead indicator has three properties. It happens early enough to act on, it happens often enough to count within a quarter, and it converts to closed-won revenue at a stable rate. MQL counts tend to fail the third test, since form fills are cheap to inflate. The candidates that hold up more often sit closer to the buying committee.
| Indicator | What it measures | What can distort it |
|---|---|---|
| Engaged target accounts | Several people from one account active within a set window | Bot traffic, existing customers, job seekers |
| Branded search and direct visits from target accounts | Demand the market creates without a paid click | Sales outreach and events driving the same searches |
| Hand-raisers (demo requests, booked meetings) | Explicit buying intent | Incentive offers that pull in poor-fit buyers |
| Qualified opportunities created | Deals sales has accepted as real | Rep targets tied to opportunity counts |
| Qualified pipeline value | Size of the accepted deals | Optimistic deal sizing early in the cycle |
Keep at least one brand-side indicator in the set. LinkedIn's summary of Les Binet and Peter Field's study, which draws on IPA Databank B2B cases from 1998 to 2018, recommends a 50/50 budget split between long-term brand building and short-term sales activation (LinkedIn). A measurement plan built only on hand-raisers will keep moving money toward activation, because activation is all it can see.
A lag table from first touch to closed-won
Before you choose a reading date for each indicator, you need the lag from first touch to each stage. The table below is a hypothetical example for a software company with a median cycle of nine months, and every figure in it is an assumption for illustration.
| Milestone | Median days from first touch | 80th percentile days | Share of first-touch accounts that reach it |
|---|---|---|---|
| First known touch | 0 | 0 | 100% |
| Engaged account | 21 | 45 | 30% |
| Hand-raiser | 75 | 130 | 8% |
| Qualified opportunity created | 110 | 170 | 5% |
| Proposal stage | 190 | 250 | 2.5% |
| Closed-won | 270 | 360 | 1.2% |
Read it from the bottom up. In this example a qualified opportunity appears about 160 days before the revenue it predicts, and an engaged account about 250 days before. To build your own, pull two to three years of CRM history covering every opportunity, won and lost, and date each one from the first known touch by anyone at the account. Report medians and 80th percentiles, since a handful of deals that stall for a year will drag an average out.
How to validate each indicator against closed-won revenue by cohort
Validation means grouping accounts by the quarter of their first touch, measuring the indicator at a fixed point, then waiting for the cohort to mature and measuring the revenue it produced. The fixed reading point matters. If you count every opportunity a cohort ever creates, older cohorts look better simply because they had more time.
The hypothetical check below counts qualified opportunities created within 180 days of first touch and reads revenue 18 months after first touch.
| First-touch cohort | Qualified opportunities by day 180 | Closed-won revenue by month 18 | Revenue per qualified opportunity |
|---|---|---|---|
| Q2 2024 | 42 | $1.47M | $35.0k |
| Q3 2024 | 51 | $1.69M | $33.1k |
| Q4 2024 | 38 | $1.25M | $32.9k |
| Q1 2025 | 60 | $1.32M | $22.0k |
Three steady cohorts make the indicator usable, at roughly $33k to $35k of revenue per opportunity in this example. The Q1 2025 drop is the pattern to watch for: opportunity counts rose and revenue did not follow, which usually means the stage definition loosened, often because someone's quota became the indicator. Run the same check by channel and by segment, since an indicator can hold for enterprise accounts and fail for mid-market. Repeat it each quarter as another cohort matures. The ratio you validate here becomes the conversion factor for the holdout readout and the MMM, much as the correction factor does in the multiplier method for calibrating platform attribution.
A holdout design that reads pipeline instead of revenue
Validated indicators put a value on pipeline. Whether marketing caused that pipeline is a separate question, the one incrementality testing answers. On a nine-month cycle a revenue holdout would run for most of a year and then wait for deals to close, so a pipeline holdout reads the validated indicator instead.
A hypothetical design for an account-based program:
- Take the target account list (2,000 accounts in this example) and randomly assign 20% to a control group, stratified by segment and by whether the account already has an open opportunity.
- Exclude the control accounts from paid targeting wherever your channels allow it, and agree with sales that those accounts get worked exactly like the rest.
- Run the program for one quarter.
- Read the primary metric at the 80th percentile lag of the indicator you chose. With the lag table above, qualified opportunities would be read about 170 days after launch.
- Convert the difference in opportunities per 100 accounts to revenue with the validated ratio, then confirm against actual closed-won revenue when the cohort matures.
Choose the earliest validated indicator that still produces enough events. If the 400 control accounts produce eight opportunities, the interval around any difference will be wide, and engaged accounts or hand-raisers may work better as the primary metric with opportunities as a secondary read. Channels that cannot target an account list, such as broad video or audio, need a geographic design; our guide to a quarterly geo-holdout test covers that version.
Modeling pipeline as the KPI in Meridian or Robyn
A marketing mix model on a long cycle usually works better with pipeline as its KPI, and the reason is in the documentation. Meridian truncates each channel's effect at the max_lag setting. Google recommends a range of 2 to 10 for geometric decay and 4 to 20 for binomial decay, and notes that with large values there is often too little data to estimate lagged effects precisely. On weekly data, 20 periods is 140 days, about half of a nine-month cycle. Qualified opportunities, at a median of 110 days in the lag example, fit inside that reach, while closed-won revenue at 270 days would be cut off.
Meridian accepts a KPI that is not revenue. In that case Google recommends supplying revenue_per_kpi, or an estimate of it, so that incremental outcome is expressed in money. Without it, and with ROI priors on media, Meridian defaults to a total paid media contribution prior with a mean of 40% and a standard deviation of 20%, and ROI is read as incremental KPI units per unit of spend. The validated revenue per qualified opportunity from your cohort work is the natural value for revenue_per_kpi.
Robyn takes a single dependent variable typed as either revenue or conversion, according to Meta, so qualified pipeline value fits the first type and an opportunity count fits the second. The same page strongly recommends Weibull PDF adstock for products with a longer conversion window, because it can model a lagged effect that geometric decay cannot. Both tools accept experiment results for calibration, which is where the pipeline holdout feeds back in. Google's calibration guidance says that experiments and MMMs "often have different estimands and measurement goals" and asks whether an experiment ran long enough to capture long-term effects (Google), so express the holdout in the model's KPI and weeks before you use it as a prior. For a side-by-side view of the two tools, see Google Meridian vs Meta Robyn.
What to report while the revenue is still months away
Each quarter, a report that holds up in a budget meeting has three parts: the leading indicators with the revenue ratio each one has earned, the newest matured cohort's actual closed-won revenue next to what its indicators predicted, and the latest holdout readouts converted at those ratios. The gap between predicted and actual revenue for the matured cohort is the health check on the whole system, and when it widens, re-validate the indicators before you move budget on them. Building the cohort tables and models behind that report is part of our data analytics work.
Sources
- Google, Google Ads Help, "About conversion windows"
- Google, Google Ads Help, "Offline conversion imports FAQs"
- LinkedIn, B2B Institute article on the 95-5 rule
- LinkedIn, B2B Institute, "5 Principles of Growth in B2B Marketing"
- Google, Meridian documentation, "Set the max_lag parameter"
- Google, Meridian documentation, "When the KPI is not revenue (total paid media contribution prior)"
- Google, Meridian documentation on ROI priors and calibration
- Meta, Robyn documentation, features page
