When the sale closes fourteen months after the click

Almost every piece of writing about marketing measurement assumes a business where somebody clicks an advertisement and buys something within a few days. If that is your business, the standard advice works. If your median deal takes eleven months and your largest take two years, the standard advice is not merely insufficient — it will actively mislead you, and it will do so with confident-looking numbers.

In a long-cycle business, this month's revenue and this month's marketing spend have almost nothing to do with each other. Every report that puts them in the same row is comparing strangers.

This article is about what to do instead. It is written for the mid-market and enterprise businesses where this is the normal condition rather than an edge case: capital equipment, professional services, healthcare procurement, software sold to committees, anything involving a tender.

The mismatch at the centre of everything

Period reporting divides revenue recognised in a month by spend incurred in that month. When the gap between cause and effect is a year, those two figures describe different populations, and their ratio has no meaning.

The consequence is not merely that the number is imprecise. It is that the number moves for reasons unrelated to performance. Increase spending and the ratio worsens instantly, because the denominator is immediate and the numerator is not. Cut spending and it improves. A business managed on that signal will systematically cut during growth and spend during decline.

This is not a hypothetical failure mode. It is one of the most common patterns in mid-market marketing, and it is usually diagnosed as a marketing performance problem rather than as an artefact of the reporting period.

Cohorts, which fix most of it

A cohort report groups enquiries by the month they arrived and follows each group forward, recording what has closed so far. Spend attaches to the cohort it generated, so cause and effect stay together as the months pass.

The mental shift is from asking what did we earn this month to asking what has the March cohort become. That second question has a stable, accumulating answer that improves as evidence arrives, rather than a volatile one that reflects the accident of when contracts were signed.

Read the first row on its own and the cohort looks like a failure. Read the whole curve and it is comfortably the best month of the year. In period reporting, only the first row is ever visible at the moment decisions are made about that spending.

The maturity curve is the asset

Once you have several cohorts of history, the shape of the curve becomes predictive: you can tell at three months whether a cohort is tracking above or below the pattern, long before it has finished closing.

This is what converts cohort reporting from an accounting exercise into a management tool. Knowing that cohorts typically reach thirty percent of their eventual value by month six means a cohort at fifteen percent is genuinely behind, and you can say so with evidence rather than anxiety.

Building the curve takes time — you need cohorts old enough to have substantially closed — which is the honest cost of this approach. Most businesses can construct it retrospectively from historical data in an afternoon, which is a strong argument for doing that before waiting a year to accumulate it prospectively.

The date range rule, which fails silently

The source export must reach back further than the oldest sale being reported, by at least the ninetieth percentile of cycle length. Exporting both files over the same window is the commonest cause of a catastrophic-looking first result.

The failure is silent because nothing errors. The match runs, the report renders, and a large share of revenue lands in the unattributed bucket, where it is interpreted as evidence that marketing is not working or that the tool does not work. Both conclusions are wrong and both are drawn frequently.

The diagnostic is simple: plot the age of the matched enquiries. If the distribution stops abruptly at the edge of your export window rather than tapering naturally, the window is the constraint, not the marketing.

Leading indicators, and how to keep them honest

Because revenue arrives late, long-cycle businesses need intermediate measures. The requirement is that each one is periodically validated against eventual revenue rather than assumed to predict it.

Qualified opportunities by source is the usual choice and it is a reasonable one. Pipeline value created by source is better where deal sizes vary widely. Both are only useful if the relationship between the indicator and closed revenue has been checked, and checked again after any change to how opportunities are qualified.

The failure mode is well known and worth naming: the indicator becomes the target, qualification criteria loosen to meet it, and the indicator continues to look healthy while its relationship to revenue quietly dissolves. Guard against it by re-running the correlation every couple of quarters and by never tying compensation solely to the intermediate measure.

Match at the account, not only the person

In long-cycle business-to-business sales, the person who enquired is often not the person who signs. Matching only on individual identifiers will miss connections that genuinely exist, and will understate exactly the channels that reach early-stage researchers.

The practical technique is to match on person first, then on account — company name normalised, email domain, or a company identifier where both systems hold one — and to treat account-level matches as a lower confidence tier that a human confirms rather than as automatic revenue.

That tiering matters. Account matching is powerful and it is also where false positives live, particularly in large organisations where several unrelated projects run at once. A match that says this enquiry from a procurement analyst became that contract signed by a director eighteen months later is plausible, valuable, and worth a human glance before it counts.

What to tell the board while you wait

Report the cohort view, the leading indicators with their validation, and an explicit statement of when revenue evidence for recent changes will exist. Boards accept lag when it is named in advance and lose confidence when it appears as an excuse afterwards.

The sentence that does most of the work is a simple one: the marketing changes made this quarter will begin producing revenue evidence in the third quarter of next year, and we will report the cohort curve monthly until then. Said before the results are wanted, it is a plan. Said after they are asked for, it sounds like an evasion.

It also protects the investment. The most common way good long-cycle marketing gets cancelled is that it is evaluated on a quarterly cadence against a signal that cannot possibly have arrived yet, by people who were never told that.

Optimisation signals for the ad platforms

Bidding algorithms learn from recent conversions, so a twelve-month cycle starves them. The workable approach is to send a validated mid-funnel milestone with an estimated value, while continuing to upload closed revenue as it arrives.

The milestone has to be chosen with care, because whatever you send becomes the target. A qualified opportunity is usually the right level: late enough to correlate with revenue, early enough to arrive within a learning window. A form fill is too early and a signed contract too late.

Keep the closed-revenue upload running alongside it even though it arrives too late to guide bidding directly. It is what tells you whether the milestone is still predicting revenue, and that check is the only thing preventing the whole arrangement from optimising toward a proxy that has drifted.

The reporting cadence that matches the business

Match the cadence to the cycle. Monthly operational review of cohort progression and leading indicators, quarterly review of channel allocation, annual review of the cohort curve itself. Reviewing allocation monthly in a twelve-month-cycle business is reviewing noise.

This is a harder change than it sounds, because monthly allocation review is a deeply entrenched habit and its absence feels like negligence. The argument that wins is arithmetic: show how much a single deal moves a monthly figure, and how little of a month's revenue was caused by that month's spending.

What should stay monthly is anything with a fast feedback loop — enquiry volume, response times, qualification rates, campaign delivery. Those genuinely change month to month and genuinely respond to intervention. The distinction to hold is between operational metrics and allocation metrics.

Attribution modelling is the wrong tool here

Dividing credit across a two-year journey involving a dozen people is a modelling exercise with no way to validate the result. Matching is verifiable and modelling is not, and long cycles widen that gap rather than narrowing it.

The temptation runs the other way. Long journeys have many touches, many touches suggest multi-touch attribution, and multi-touch attribution promises to sort it out. What it actually does is apply weights nobody can check to a path that is missing most of its real events, since the majority of an enterprise evaluation leaves no trace in any system you own.

The defensible claim in a long-cycle business is narrower and more useful: this deal came from this enquiry, which came from this source, and here are both records. Everything beyond that is inference, and it should be labelled as such when it reaches a slide.

CloseRev matches rather than models, supports account-level matching alongside person-level, and lets the source export reach as far back as the data allows — which is what a long-cycle business actually needs from the category.

Deals that die, and the survivorship problem

Long cycles produce many enquiries that neither close nor formally end; they simply stop. Counting only closed deals makes every channel look better than it is, because the cost of the enquiries that went nowhere never appears.

The practical remedy is a stale rule applied consistently: an opportunity untouched for a defined period is treated as lost for reporting purposes, whatever its status field says. Without one, the pipeline accumulates optimistic records indefinitely and the conversion rates computed from it drift upward every year for no real reason.

The rule needs to be longer than your ninetieth-percentile cycle or it will discard deals that were always going to be slow. Setting it at roughly one and a half times that figure is a defensible starting point, and it should be written down alongside the cycle-length figures it depends on.

Attribution when the buyer never identifies themselves

In long-cycle categories much of the evaluation happens without any identifiable interaction: analyst reports, peer conversations, communities, search results read and never clicked. That activity is real, influential, and permanently outside your data.

The honest response is to size the gap rather than to model it away. If seventy percent of your closed revenue matches to a recorded source, the remaining thirty percent is not a failure of the tooling; it is a measurement of how much of your market's decision-making happens where you cannot see. That figure is itself a useful thing to trend.

It also argues for treating self-reported source seriously in these categories, collected at a moment when the buyer will answer honestly, and kept as its own field rather than merged into the matched data. It is the only instrument that reaches the invisible part of the journey, and its weaknesses are at least well understood.

The one thing to do this week

Measure your own cycle length distribution before anything else. Not the average — the tenth, fiftieth and ninetieth percentiles. Every other decision in this article depends on those three numbers and most businesses have never calculated them.

It takes one query against closed deals: the days between enquiry and close, distributed. The result is frequently surprising, particularly the ninetieth percentile, which is usually far longer than the number everybody quotes in meetings. That gap explains a great deal of previously mysterious reporting.

Once you have it, the export window follows, the cohort maturity expectation follows, the reporting cadence follows, and the sentence you say to the board follows. Three numbers, one afternoon, and most of the confusion resolves.

There is one further reason to compute it yourself rather than trusting the figure everybody repeats. Cycle length is usually quoted from memory by whoever has been in the business longest, and that memory is anchored on the deals that were memorable, which are disproportionately the fast ones and the disastrous ones. The median deal is unremarkable by definition, which is exactly why nobody remembers how long it took.

Most long-cycle businesses have never calculated their own cycle length distribution, and are running their marketing reporting on a guess about it.

Questions people actually ask

How do you measure marketing ROI with a long sales cycle?

By cohort rather than by period. Group enquiries by the month they arrived, then follow each cohort forward as it closes over the following year or two. Comparing this month's spend against this month's revenue is meaningless when the two refer to different customers.

How far back should the lead export go for a long cycle?

At least to the ninetieth percentile of your cycle length before the oldest sale you are reporting on, and preferably further. If deals routinely take fourteen months, a two-year lead history is the minimum that will explain a year of revenue.

What should I report while waiting for deals to close?

Leading indicators that you have validated against eventual revenue — qualified opportunities by source, pipeline value created by source, and cohort progression. The discipline is to check periodically that they still predict revenue rather than assuming it.

Does attribution even work for enterprise sales?

Matching does; modelling largely does not. You can reliably establish that a closed deal came from a specific enquiry. Dividing credit across a two-year, twelve-person buying journey is a modelling exercise with no way to check the answer.

How do you handle deals involving many people at one company?

Match at the account level as well as the person level. In business-to-business the enquiry and the signature frequently come from different people, so a person-only match will miss the connection that actually exists.

How long before a marketing change shows up in revenue?

Roughly one full sales cycle before the first signal and two before it is trustworthy. Setting that expectation in advance is the single most useful thing you can do, because the alternative is a change being reversed before it could possibly have worked.

Found this useful?

• Share on X
• Share on LinkedIn
• Share on Reddit
• Share on Email