How to Measure Demand Generation ROI in 2026

An MQL is a prediction that a lead will turn into revenue. In most companies, marketing sets the threshold and is then graded on how many it produces. That is a scoring system marking its own homework, and it explains why MQL counts so often rise in the same quarter that pipeline falls.

Measuring demand generation ROI properly means four numbers instead of one: sourced and influenced pipeline under definitions agreed with sales, pipeline to closed revenue conversion by source with the sales cycle lag built in, cost per opportunity and CAC payback rather than cost per lead, and a periodic incrementality read on what would have happened anyway. Then you close the loop, so the ad platforms optimize toward deals rather than form fills.

Why the MQL stopped working as a return metric

Trade coverage through 2026 has tracked a steady shift in B2B from MQL volume toward sourced revenue and influenced pipeline as the headline measure. The reason is structural rather than fashionable.

MQL volume responds to the definition, not to the market. Loosen the scoring rules and the number goes up. Buy cheaper traffic and the number goes up. Neither produced a single additional deal. A metric that improves when quality falls is not a return metric, whatever else it is useful for.

This is not an argument for deleting the MQL. It remains a reasonable operational signal for routing, follow-up speed and sales capacity planning. It stops being reasonable the moment it is presented to a board as evidence of return.

The four numbers that replace it

1. Sourced and influenced pipeline, defined in advance

Sourced pipeline is opportunity value where marketing created the first qualifying interaction. Influenced pipeline is opportunity value where marketing touched the account at any point before close. Sourced is narrow and defensible. Influenced is broad and trivially inflatable, because nearly every deal has a marketing touch somewhere in it.

Report both, and write the rules down with sales before the first report rather than after the first disagreement. An influenced pipeline figure without a stated definition is the fastest way to lose finance's trust in everything else on the page.

2. Pipeline to revenue conversion, by source

Pipeline is not revenue. Two sources can generate identical pipeline value and convert at wildly different rates, and the gap usually shows up in deal size, cycle length and discount depth rather than in win rate alone.

Track conversion from opportunity to closed won by source, and track the three things that move alongside it: average deal size, median cycle length and realized discount. A source that converts well but only on heavily discounted deals is not the source you thought it was.

3. Cost per opportunity and CAC payback

Cost per lead rewards volume. Cost per opportunity rewards qualification. CAC payback, the number of months of gross margin required to recover fully loaded acquisition cost, is the one finance already understands, because it is a working capital question rather than a marketing one.

Load the cost properly. Media, agency fees, tooling, content production and the demand generation salaries. A CAC that counts only media spend flatters every channel equally and tells you nothing about which one to scale.

4. Incrementality, at least twice a year

Every attribution model answers the question of who gets the credit. None of them answers whether the revenue would have arrived anyway. Branded search is the standard example: it converts beautifully and a meaningful share of it was always going to happen.

A geo holdout, where spend is withheld in matched regions and the difference measured, is the cheapest causal read available to most B2B teams. It costs real pipeline in the holdout regions, which is why it tends to get proposed and never scheduled. Put it in the plan with a date and a budget line, or it will not happen.

Close the loop: send revenue back to the platforms

Most B2B accounts optimize toward form fills because a form fill is the last thing the platform saw. The bidding system is then working hard to buy more of the wrong thing. Fixing this is unglamorous plumbing with a larger effect than most creative changes.

In Google Ads, closed revenue goes back through offline conversion uploads using enhanced conversions for leads or Google Ads Data Manager, with daily uploads recommended. Value-based bidding needs two or more unique values to have anything to rank between, so lead generation businesses typically send lead scores, values tied to form answers, margin figures or predicted lifetime value rather than a flat number.

The thresholds are where long cycle B2B gets stuck. Target ROAS asks for at least 15 conversions in the last 30 days and conversion delays under seven days, which a six month sales cycle does not satisfy on closed revenue. The practical answer is to optimize toward a scored intermediate event that correlates with closing, such as a qualified opportunity, and to validate that correlation rather than assume it.

One naming trap worth flagging. Google has a campaign type called Demand Gen, which is not the same thing as the discipline. Its value bidding requirements are stricter than standard campaigns: 50 conversions with value in 35 days at campaign level including 10 in the last seven days, plus 100 across the account, and no bid or creative changes for 14 days after launch.

Three methods, and when each earns its keep

Multi-touch attribution is for in-platform optimization. It is directional, it only sees touches it can track, and it is structurally blind to dark social, private communities, events and the conversation that happens before anyone visits your site. Use it to steer, not to settle arguments.

Incrementality testing is for causal truth on one question at a time. It is the only method that can tell you a channel is not working. It is expensive and slow, which is the honest reason it is rare.

Marketing mix modeling is for strategic allocation across channels at a level attribution cannot reach. It needs years of history and enough variation in spend to learn from. Below real scale it produces confident numbers from insufficient data, which is worse than no model.

Self-reported attribution, the how did you hear about us field, is the cheapest of the four and the most biased. People over-report what they remember and under-report advertising. Its value is in finding channels your tracking cannot see. A large gap between self-reported and tracked attribution is a signal to investigate, not a discrepancy to reconcile away.

The lag nobody budgets for

If the median cycle is six months, this month's spend and this month's closed revenue are two unrelated things sitting next to each other on a slide. Comparing them produces a number that moves for reasons nobody can explain, which is how marketing budgets get cut in the quarter the previous quarter's work was about to land.

Report in cohorts instead. Group spend and the pipeline it created by the period the spend happened, then let each cohort mature and report it as it does. The first few quarters of this feel uncomfortably slow. They are also the first honest numbers the business has had.

A cadence that survives a board meeting

• Monthly: leading indicators only. Spend, opportunities created, cost per opportunity, pipeline coverage against target.
• Quarterly: matured cohorts. Pipeline to revenue conversion by source, deal size, cycle length, CAC payback.
• Twice a year: one incrementality test, or a modeling refresh if you are at the scale to support one.

The discipline that makes this work is refusing to report a number at a frequency faster than it can actually change. Most demand generation reporting fails on that single point.

What to stop reporting

• MQL volume as a headline number
• Last-touch attribution presented as causation
• Cost per lead with no quality measure attached to it
• Influenced pipeline with no written definition of influence
• Platform-reported ROAS or conversions as the company's revenue number

Growth is not a marketing problem. It is a capital allocation system. Demand generation ROI is simply the part of that system where the capital goes out before the evidence comes back, which is exactly why the measurement has to be built before the spend, not after someone asks.

A note on sources: Google Ads bidding thresholds and upload guidance here are drawn from Google Ads help documentation and PPC Land's reporting, checked in November 2026. Platform requirements change often, so verify them in your own account. The shift away from MQL volume is drawn from B2B trade coverage through 2026 and is described here as a direction of travel, not as a measured percentage.

FAQ

Frequently asked