Why it survives in well-run companies
This is the part that surprises operating partners, and it should not. Leakage is not a symptom of incompetence. It is a symptom of growth: systems accumulate one subscription at a time, each one solving a real problem, and the gaps appear between them rather than inside any of them. Nothing is broken. Every individual process works. Revenue simply stops arriving at a rate nobody can point to.
And nothing in the reporting stack is designed to surface it. Marketing reports on cost per lead and will tell you, accurately, that the channels are performing. Sales reports on conversion of the pipeline it received. Neither reports on the population that never became a pipeline record, because that population leaves no trace. There is no angry customer, no bad review, no line item, no ticket. The usual conclusion is that the market got harder.
The five places it actually sits
- 01Response latency. The enquiry that arrived and was answered thirty hours later, by which point the decision had been made elsewhere. This is the largest single category in most companies and the one most often misdiagnosed as a staffing problem when it is a routing problem.
- 02Unmonitored channels. The inbox one person watched before they changed roles. The second phone line. The marketplace message queue. The chat widget nobody has opened in months. Each was deliberately created and each has since become invisible.
- 03Absent follow-up. A first reply went out, the prospect did not respond, and nothing was scheduled behind it. A conversation that stalls closes itself, silently, and the CRM records it as an open opportunity indefinitely.
- 04Failed routing and handoff. It arrived, it was answered, and it went to the wrong person, the wrong territory or the wrong location. The loss happens in the handoff, which is the one part of the process no system owns.
- 05Renewal and expansion drift. Existing revenue that lapses because the trigger was a person remembering rather than a system firing. Usually the cheapest leak to close and the last one anyone looks at.
The common structure is worth naming: in every one of the five, the failure is work that depends on a person remembering to do it. That is the single most reliable predictor of where a leak will be found, and it is why the diagnostic looks at handoffs before it looks at anything else.
How it gets measured, and the bar a figure has to clear
A leakage figure is only worth having if it can survive an operating partner doing arithmetic on it in front of you. That means three things have to travel with every number: the system it was measured in, the period it covers, and the conversion assumption applied to it. A figure without all three is a claim rather than a measurement, and it will be discounted to zero by the first person who asks where it came from.
The method is unglamorous. Trace every inbound path through the systems of record. Timestamp each enquiry against its first genuine human response — not the autoresponder. Split the population at whatever latency threshold is meaningful in that market, and compare conversion between the fast group and the slow group using the company’s own historical rates. Then apply the company’s own average deal value, and annualise.
Two disciplines make the difference between a defensible figure and a sales number. The first: anything that cannot be measured in a system the company owns does not go in the report at all. Interview evidence is used to find where to look, never as the source of a figure. The second: a recoverability haircut, stated explicitly. Not every lost enquiry was ever winnable, and a model that implies otherwise is claiming a 100% seal, which nobody believes.
The arithmetic, shown
Take a company doing $40M with a services mix, receiving 600 qualified enquiries a month, reaching about 35% of them inside an hour, closing 22% of the conversations it actually has, with an average first-year contract value of $9,000. Every one of those four inputs comes from the company’s own systems.
- 210 enquiries reached fast, closing at 22% — about 46 deals a month.
- 390 reached slowly. Apply a conversion penalty to the slow group taken from that company’s own historical split rather than from a benchmark — say those close at 10% — and that is 39 deals.
- Current total: roughly 85 deals a month. The ceiling, if every enquiry were reached inside the hour at the same rate, is 132.
- The gap is about 47 deals. Apply a stated recoverability haircut — not all of them were winnable — of 60%, and you are left with 28 deals a month.
- At $9,000 each, that is roughly $252,000 a month, or about $3.0M annualised.
Leakage is not the same as a growth problem
The distinction decides where the money should go, so it is worth being precise. A growth problem means demand is insufficient: the answer is marketing, pricing or product. A leakage problem means demand is sufficient and is being lost after arrival: the answer is operational. The two feel identical from the top of a P&L and they have nothing in common as interventions.
The test is quick. Compare conversion between your fast-response population and your slow-response population, using your own data. If the fast group converts materially better and the slow group is large, you have a leakage problem and additional marketing spend will be partially wasted — you will be paying to acquire demand that will leak out through the same gaps. If the two groups convert about the same, latency is not your constraint and the money belongs in demand generation.
Across more than 500 businesses taken through this process, the cumulative unrealized revenue identified exceeds $100M, which is roughly $200,000 per business. That figure is not annualised, it is totalled at the point the demand was lost, and most of those businesses sit well below the $3M to $15M range this firm now works in — so it is a statement about how common the pattern is, not a forecast of what any single company will find. What any single company will find is a question only its own systems can answer.