1. Same cohort, same periods
A typical mistake in bottleneck diagnosis happens before the first number: volumes from different periods are put in relation to each other. This month’s quotes come from last month’s inquiries; this month’s closings from quotes sent six weeks ago. A rate built from such numbers measures the calendar, not the sales process.
That’s why the diagnosis works with a cohort: all inquiries from a closed period – for example a month or quarter – and their path up to today. The period has to be far enough back that most decisions have been made. If your sales cycle lasts 60 days, look at a cohort that is at least 90 days old.
2. What you measure per stage
- Volume per stage
- How many cases in the cohort reached this stage? Count, don’t estimate.
- Conversion
- Volume of the next stage divided by volume of this stage. Always in pairs between adjacent stages – and once across the whole path.
- Dwell time
- Median of the days a case spent in the stage. The median is robust against individual long-runners.
- Response time
- Time from receipt of the inquiry to the first personal response, also as a median. Automatic acknowledgments don’t count.
- Loss reasons
- For each lost case, one reason from a fixed list (e.g., no need, timing, budget, competitor, no response, canceled internally). Free text only as a supplement.
3. A mini example (fictional)
| Stage | Quantity | Conversion to the next stage |
|---|---|---|
| Inquiries | 100 | 30% (30 of 100 qualified) |
| Qualified | 30 | 50% (15 of 30 with a quote) |
| Quotes | 15 | 20% (3 of 15 won) |
| Closings | 3 | Total: 3% (3 of 100) |
At first glance, the quote stage at 20% is the weakest. But be careful: a low rate at one stage doesn’t yet say the cause lies there. Perhaps too many unsuitable inquiries slipped through qualification – then the problem only shows up at the quote but arises one stage earlier. Perhaps dwell time in the quote stage is high because quotes are sent but never discussed. The rate shows you where to look. It doesn’t show what to change.
4. The diagnostic decision tree
The tree leads from observation to hypothesis. It never ends at a cause, but always at a testable assumption and a question you settle in an interview or a test.
QuestionDo all the numbers come from the same cohort, and has the sales cycle for it run its course?
- No: Stop. First define the cohort and wait for the measurement point – otherwise you're comparing today's inquiries with yesterday's closings.
- Yes:
QuestionIs the volume at the first stage (inquiries) too small to even mathematically reach the target volume?
- Yes: HypothesisVolume bottleneck at the entry. Check: origin of inquiries, response among existing customers, demand built up. Caution: First check whether later stages can even process the existing volume.
- No:
QuestionIs the median response time to new inquiries more than one business day, or is it not measured at all?
- Yes: HypothesisResponse bottleneck. Check: Who sees new inquiries and when, is there a backup, is the first response logged?
- No:
QuestionWhich stage shows the lowest conversion AND the longest median dwell time at the same time?
- Inquiry → qualified: HypothesisQualification bottleneck. Check: Do audience and message fit, are there fixed qualification questions, is “not a fit” recorded properly?
- Qualified → Quote: HypothesisNeeds-clarification bottleneck. Check: Is the decision-maker reached, is the need confirmed in writing, does creating the quote take too long?
- Quote → decision / close: HypothesisDecision bottleneck. Check: Is a scheduled next step missing, is there a loss reason “no response,” are terms or alternatives unclear?
- No tier stands out clearly: HypothesisNo stage bottleneck proven. Check: loss reasons by frequency, comparison with a second cohort, data quality of the stage assignment.
Every end node is a hypothesis – not a diagnosis. It is confirmed or rejected only through interviews and the prioritized test (see below).
5. Why a rate doesn’t prove a bottleneck
- A stage can look weak because the stage before it was too generous. The rate then measures qualification errors, not quote quality.
- A stage can look strong because only sure things arrive there – and the real selection quietly happened earlier.
- Small cohorts fluctuate. With 15 quotes, a single closing shifts the rate by almost seven percentage points.
- External events (season, price change, staff turnover) affect individual cohorts. Compare at least two cohorts before you call something a pattern.
That’s why the diagnosis has two stages: first the number shows where you look. Then a conversation or a test confirms or disproves the hypothesis.
6. Interview questions
To the team
- When do you decide that an inquiry is qualified – and what exactly do you ask to determine that?
- What happens after a quote is sent? Is there a fixed date to discuss it?
- For which lost deals in the latest cohort was the reason already visible before the quote?
- Which piece of information do you miss most often when a customer doesn’t decide?
To lost customers (three to five conversations are enough for first patterns)
- What tipped the decision – and when in the process was that clear?
- Was there a point where you missed an answer or a document?
- Would a different sequence or a different contact person have changed anything?
7. Prioritized test plan
Each hypothesis becomes a test with a metric, a period and a stop criterion. Prioritize by expected effect and effort; test one thing per stage, not three at once.
| Hypothesis | Test | Metric | Period | Stop if |
|---|---|---|---|---|
| Quotes are sent, not discussed | Send every quote only with a meeting scheduled to discuss it | Cohort closing rate, dwell time in “Quote” | A monthly cohort, measured after 90 days | Meeting is declined by > 50% of customers |
| Qualification lets unsuitable inquiries through | Four fixed qualification questions before every quote | Share of quotes among qualified inquiries; loss reason “no need” | A monthly cohort | Number of quotes drops, closings drop too |
| Late response costs inquiries | Response within a defined time window, one owner | Median response time; qualification rate | Four weeks | Response time drops, qualification rate stays unchanged |
The worksheet below has columns for cohort, stage, volume, conversion, dwell time, response time, loss reasons, hypothesis and test – as a CSV you can open in any spreadsheet.
Downloads
To take away
Markdown opens in any text editor or note-taking tool. CSV files are semicolon-separated and open directly in Excel, Numbers, or LibreOffice.
