- The ROAS a platform reports for a B2B account is almost always cost-per-lead wearing a different label: it counts form submissions inside the attribution window, while contracts close months after that window shuts. The two systems are counting different events, so comparing the figures directly is meaningless.
- Define return before optimising it. The workable approach is to assign each funnel stage a value derived from historical close rates, and to import closed-won amounts back into the platform so automated bidding learns from outcomes rather than from form fills.
- While the numerator is mismeasured, the denominator carries spend that never had a chance: off-target search terms, non-buyer audiences, invalid clicks and automated traffic. High click costs and a thin conversion base make this more damaging in B2B than in consumer categories.
- Sequence matters more than any individual tactic: repair measurement, then remove clearly wasted spend, then address landing page and form friction, and only then change bidding. Reversed, every later step optimises against a ratio that is already wrong.
- Read results as cohorts fixed by the month of the click, matured, and compared against cohorts of the same age. A one-month-old cohort compared against a year-old one will always look worse, purely because it has had less time.
The reported number is not measuring revenue
An ad platform credits a conversion when the tracked action fires inside its attribution window. That window is measured in days; the median B2B sales cycle is measured in months. The gap is not a precision problem but a units problem: the platform is counting form submissions and demo requests, the CRM is counting contract value, and the two events sit one or two quarters apart on the timeline.
The consequence runs in both directions. Upward, the flattering ratio in the report is really a measure of form-generating efficiency, and presenting it as return on revenue tends to collapse the first time finance reconciles it. Downward, automated bidding learns from the same signal: it will work hard to produce more forms, and more forms is not more contracts. When low-quality leads are cheaper to acquire, the system will honestly optimise toward them, because that is what the objective function it was given actually says.
This is why B2B accounts so often produce a contradictory picture: cost per lead is falling on the platform side while the sales team reports that lead quality is deteriorating. Both statements can be true at the same time, and usually are. The general definitions and health baselines in our overview of how to improve return on ad spend still apply, but in B2B they only start working once the measurement has been aligned.
Define what counts as return before optimising it
Before touching a single campaign setting, the account needs a definition of return that everyone agrees on. There are three practical options, each with a cost. Pure revenue counts only signed contract value: the most accurate, and far too slow to steer day-to-day decisions. Pipeline counts opportunity value at a chosen sales stage: much faster, but it imports the sales team's subjective judgement into an advertising metric. Stage-weighted value assigns each funnel stage a figure equal to its historical close rate multiplied by average contract value.
The third is usually the workable compromise: it preserves the fact that leads differ in value while still producing a number before any contract is signed. The cost is that it depends on historical close rates, and those drift as the product, the pricing and the market change. The coefficients need recalculating each quarter rather than being set once and inherited for a year.
Once the definition is settled, the decisive step is sending that value back to the ad platform. The major platforms all accept offline conversion import keyed on the click identifier, which lets the bidding model learn from real outcomes instead of form fills. The engineering effort here is routinely underestimated — the click identifier has to be captured at form submission, carried through the CRM to the closed deal, and uploaded on a schedule — but it is the only action in the chain that changes what the system is learning. Without it, every later optimisation is tuning against the wrong objective.
Industry averages quoted for B2B return are almost always comparing incompatible definitions and are not portable. The usable reference is internal: the same account, on the same definition, over a previous comparable period, with the acceptable threshold derived from your own gross margin and renewal rate rather than a published figure.
The denominator: spend that never had a chance
Return is a ratio, and nearly all of the attention goes to the numerator. The denominator deserves the same scrutiny, because some share of the spend was destined not to convert at the moment the click happened — not converting slowly, but never converting at all.
It typically comes from four places: search terms matched to unrelated intent, placements plainly outside the target audience, invalid clicks that the platform has already identified and credited back, and automated traffic. The first two converge with routine review of search-term and placement reports, which is ordinary operational work. The last two need to be looked at separately: the invalid click filtering and automatic credits in Google Ads remove part of it, but the platform's reporting rarely lines up cleanly with your own traffic data, so both sides have to be read together.
What makes B2B distinctive is the leverage. When a keyword costs a great deal per click and produces single-digit qualified leads per month, even a modest share of clicks that were never prospective buyers accounts for a substantial share of the outcome. The standard methods for identifying that traffic are covered in bot traffic detection, while the basic categories of click fraud and the definition of invalid traffic determine which portion the platform handles automatically and which portion is left to you.
A defensible order of operations
The most common question on this topic is which action works fastest, and the honest answer is that the fastest route is not reversing the order. The sequence below is not a matter of taste; each step depends on whether the output of the previous one can be trusted.
- Repair measurement. Carry the click identifier through to the CRM and send closed-won values back to the platform. Until that is done, any optimisation is only adjusting form-generating efficiency.
- Remove the clearly wasted spend: search terms, placements, geography, device, and invalid or automated traffic. This changes the denominator immediately and does not depend on the previous step being finished.
- Fix the receiving end. The landing page and the form are the first filter on lead quality, and changing them costs far less than restructuring the account.
- Change bidding last. Value-based bidding is only meaningful once the platform can see real outcomes; switching strategy before then is just another way of chasing form fills.
Steps two and three can run in parallel; steps one and four cannot be swapped. The reasoning matches the general approach in our piece on return on ad spend optimisation strategies, with the difference that on B2B timescales the first two steps carry considerably more weight than they do in consumer accounts.
The landing page and the form as the first quality filter
A B2B landing page carries a job that a consumer page does not: it has to improve not only the submission rate but the quality of who submits. These two goals pull against each other. Stripping the form to its minimum raises submissions while admitting a large volume of people with no purchasing intent; a long form keeps genuine buyers out. The resolution is not to pick one, but to move the qualification after the submission: collect the minimum necessary, then qualify in a later step.
The usual improvements are covered in improving landing page conversion rate. One caution is specific to B2B: the conversion base is thin, and many tests never reach a sample size that supports a conclusion, which ends with a statistically empty difference being treated as a finding. Estimating the required duration first, using sample size and significance for landing page tests, avoids a great deal of wasted experimentation; the mechanics of running the test itself are in our landing page A/B testing guide.
Reading the numbers without fooling yourself
Reports on long-cycle businesses carry a structural trap: the most recent period always looks worst. Not because performance declined, but because recent clicks have not had time to close. A calendar-month view of revenue booked this month divided by spend this month collapses at the start of every month and climbs back as time passes, entirely independently of campaign quality.
The correct reading is cohort-based: fix the traffic into a group by the month the click occurred, then track what that group closes over time. Comparisons are made only between cohorts of the same age — three-month-old against three-month-old. That is what separates "has not had enough time" from "quality declined", two situations that look identical in a calendar view.
Something usable is still needed before a cohort matures. Qualified-lead rate and progression to the second sales stage work as leading indicators, provided everyone reading the report understands they are proxies rather than return. Reporting a proxy as ROAS is among the most common self-inflicted injuries in this area. When separating types of invalid activity, the distinction between general and sophisticated invalid traffic determines whether the platform credits it automatically, and how click fraud patterns differ by industry covers the sector-level differences.
Frequently asked questions
Why is my B2B ROAS lower in the CRM than in the ad platform?
Because the two systems count different events at different times. The platform credits a conversion when the tracked action fires inside its attribution window, which is measured in days, and in B2B that action is almost always a form submission or a demo request rather than a payment. The CRM credits revenue when the contract is signed, which usually happens after the window has closed and often after the click has left the platform's reporting entirely. The platform figure is a cost-per-lead measure wearing a ROAS label; the CRM figure is closer to real return but arrives too late to steer bidding. Neither is wrong on its own terms. The error is comparing them as though they measured the same thing, or optimising against the fast one while being judged on the slow one.
What is a good ROAS for B2B?
There is no portable benchmark, and figures quoted as industry standards are usually comparing incompatible definitions. What is acceptable depends on gross margin, on whether contracts renew, and on how much of the reported revenue advertising can reasonably claim. A high-margin business with multi-year renewals can accept a first-order return well below one and still be profitable across the contract lifetime; a low-margin business with one-off purchases cannot. The useful reference is internal: the same account, same definition, previous comparable period, with the threshold derived from your own margin and retention rather than a published average.
What should B2B advertisers do first to improve return on ad spend?
Fix the measurement before touching the campaigns, because every optimisation afterwards inherits whatever the measurement says. In practice that means importing closed-won values back from the CRM so the platform learns from outcomes rather than form fills, and separating spend that had no chance of producing a deal from spend that produced a real but slow one. Reviewing search terms, placements and traffic quality removes cost that would never have converted under any bidding strategy, which changes the denominator immediately. Bidding and creative work are worth doing, but until outcome data flows back they act on a ratio whose numerator is still mismeasured.
How does invalid traffic affect B2B ROAS specifically?
It hits harder than in consumer categories because the cost per click is higher and the conversion base is thinner. When a keyword is expensive per click and yields a handful of qualified leads a month, a modest share of clicks that were never prospective buyers becomes a large share of the outcome. The distortion is worse than the arithmetic implies: automated bidding learns from the conversion feed, so a segment that generates clicks and no closed business degrades its own ratio — but only if the outcome signal actually reaches the platform. Ad platforms filter part of the invalid activity and credit it back automatically, so the practical work is reading those reports alongside your own traffic data rather than assuming the filtering is complete.
How long should I wait before judging a B2B campaign on ROAS?
Long enough for a cohort to have had a realistic chance to close — at least one full sales cycle, measured from your own CRM rather than from a rule of thumb. Judging at thirty days when the median deal takes ninety guarantees that every recent cohort looks worse than every older one, purely because it has had less time. The workable approach is to fix cohorts by the month of the click, let them mature, and compare cohorts at the same age instead of a young one against a mature one. In the interim, qualified-lead rate and stage-two progression carry the signal, provided everyone reading the report understands they are proxies rather than return.