We give you a number you will not have to make excuses for at the sales meeting
Nine counting rules enforced by code, not by a report setting. A number that breaks a rule never reaches the report.
What people come to us with
«The report says one thing, the team another»
A head of sales opens “Sales analysis” and sees a collapse mid-funnel. The managers say the stage is a formality and everything is fine. Nobody can check, and decisions about people get made on gut feeling.
«The funnel gets rebuilt in Excel every month»
Export, pivot table, half a day of work — and the number is still arguable: the counting rules live in the head of whoever built the table, and next month they will be different.
«There is a percentage, but no trust»
A manager’s conversion is computed from three deals, the average time is ruined by one deal stuck since last year. It is easier to argue with such a report than to work from it.
The nine counting rules
The rules do not depend on whether you like the result. Some of them refuse to show a number — that is code behaviour too, not a failure. The examples below are computed on the pilot demo pipeline by the same functions the widget uses: not an illustration, the same calculation.
We show you the transition where sales actually stall
Result: one weak transition instead of an even slide down the whole funnel.
A cumulative calculation descends in an even staircase: you see that few deals made it to the end, and not where the team got stuck. We count each transition separately. On the demo pipeline the weakest transition is “KEY MEETING HELD → Flat reserved”: 7%. This is the step a cumulative figure smears across the whole funnel — and the meeting turns into “work harder” instead of one specific stage.
property developer account (anonymised) · “Sales to new clients” pipeline · July 2026 · computed from status-change events
Where you see it: the “Funnel” tab: every step has its own percentage — a share of the previous stage, not of the pipeline entry.
You stop losing conversion on stages that are not sales steps
Result: the same deal history yields an honest number instead of a collapse out of nowhere.
“No contact” is not a sales step. It is a parking stage for people nobody could reach. While it sits in the chain it swallows flow that later returns to the sale: the report shows a collapse that never happened, and a head of sales starts looking for someone to blame.
The same history, 32,478 deals. The only difference is whether parking stages count as funnel steps.
The markup is not made up: on the pilot, “No contact” has 1,289 entries and 70% of its exits go to lost; from “Deferred demand” 85% of exits go backwards through the funnel. The skip count depends on the same markup: by stage order there are 1,357 skips, but a skipped parking stage is not a skip — real skips of a sales stage number 30. We show both numbers.
property developer account (anonymised) · “Sales to new clients” pipeline · July 2026 · computed from status-change events
The heuristic proposes the markup, a person confirms it. Why — the review of our own mistake.
Where you see it: the “Funnel” tab: parking stages sit in a separate list with their own numbers, not as chain steps. We agree the parking list at connection and change it whenever you write to us.
We do not let one stuck deal ruin the picture for the whole team
Result: stage time does not jump because of a deal that has been hanging since last year.
Median, not average. One stuck deal must not distort the picture for the whole team.
The median is the middle value: half the deals passed the stage faster, half slower. A deal stuck for a year shifts it by one position, while it breaks the average entirely.
Where you see it: everywhere time is shown: “Funnel”, “Lead path”, “Managers”. The words “average time” appear on no screen.
We will not show a percentage that one deal can overturn
Result: instead of a striking figure you get a “not enough data” label, and there is nothing to argue about.
Fewer than 8 deals — no percentage. You see “not enough data” instead. Percentages from two deals are never reliable.
At the tail of the demo pipeline the base drops to 1 — there is no percentage there, just a label. An empty space is more honest than a striking number that jumps with a single deal.
property developer account (anonymised) · “Sales to new clients” pipeline · July 2026 · computed from status-change events
Where you see it: the “not enough data” label instead of a percentage — on “Funnel”, in “Lead path” and in the managers table.
We will not build a breakdown on a field nobody fills in
Result: you will not decide anything from a slice that describes data entry discipline rather than sales.
Completeness below 30% — no breakdown. We show how much is filled in and explain why the number would mislead. The threshold can only be bypassed via support: there is no “show anyway” button yet.
On the pilot we track 8 breakdown fields. The emptiest is “Finishing”, filled in for 1% of deals. A breakdown by it would show who fills in what, not sales.
A field whose completeness has not been computed yet is labelled exactly that — “not computed”. It is not replaced with zero: zero reads as “the field is empty”, which is a different claim.
How the threshold behaves in a report and what to do with a half-empty field — in the metrics docs.
property developer account (anonymised) · “Sales to new clients” pipeline · July 2026 · computed from status-change events
Where you see it: the “Data quality” tab: completeness of every field and a verdict — build, build with a warning, or do not build.
We separate the bot’s work from the manager’s
Result: the team median does not improve by itself just because automation makes part of the transitions.
Automation on its own line. Transitions made by a bot are excluded from the team median: automation conversion must not be credited to a person.
On the pilot, automation made 5.3% of all transitions. Hand them to people and the team median improves by itself, without a single call to a client.
property developer account (anonymised) · “Sales to new clients” pipeline · July 2026 · computed from status-change events
Where you see it: the “Managers” tab: the bot has its own row. “Data quality” shows the share of transitions made by automation.
We credit the work to whoever did it — even after they leave
Result: last quarter’s report does not rewrite itself when owners change in the CRM.
The current owner is the deal’s state today, not the author of the work. Count by it, and a head of sales who moved other people’s deals before a meeting takes their work, while a departed employee vanishes from the report with everything they did.
That is why the staff directory is cumulative: remove a person from amoCRM and their row stays in the history.
Where you see it: the “Managers” tab and the manager filter elsewhere: it selects transitions by the owner at the time of the event, not the current one.
We compare sellers with sellers, not with the whole office
Result: service and partner teams do not drag the median down, and sellers do not get an easier bar.
Support, the partner channel, back office and bots are not part of the sales median. Their work is different, and a single figure misleads both ways: it lowers the bar for sellers and imposes someone else’s on the service team.
The group type is set once during markup. If no selling groups are marked, we count over all active staff and say so in the report.
Where you see it: the “Managers” tab: deviation from the median is shown for selling groups; for the rest the column is empty — there is nothing to compare with.
We show a strange number and explain it instead of massaging it
Result: you see that a stage received deals not only from above — and exactly where they came from.
Deals enter a stage not only from above: they return from parking, roll back and move in from other pipelines. A value above a hundred per cent is not a calculation error but a description of what actually happens. We will not massage it into a tidy figure.
On the demo pipeline this is how “Taken into work” behaves: 432% of “NEW LEAD”. The difference is returns from parking and entries from other pipelines; there were 686 cross-pipeline transitions in the month.
property developer account (anonymised) · “Sales to new clients” pipeline · July 2026 · computed from status-change events
Where you see it: the “Funnel” and “Overview” tabs: the value is shown with a ⚠ sign and a note that the stage received deals not only from the previous one.
See how these nine rules look in a report
The demo is open with no sign-up and no access to your CRM: the same pipeline, the same rules, the same numbers as on this page.
The rule we drew from our own mistake
The heuristic proposes the parking markup, a person confirms it. We got there the hard way: on the full pilot history — 7 years — the heuristic declared a stage that closes deals into revenue a parking stage, and mid-funnel conversion came out several times lower than the truth.
A threshold fixed some of the false positives; the last one cannot be fixed by a threshold at all — the separating number would have to be tuned to a single account. We did not do that, and wrote up why: with numbers, a date and what it means for your account.
measured on the full history of the pilot account · 69,567 deals · 7 years · docs/РЕШЕНИЯ.md, пункты 37 и 40
What we do not promise
There is no “count it like everyone else” switch
If a number disappoints you, we show which events it is built from and what base it was computed on. We will not recount it by other rules to make it look better.
There is no “not a parking stage” button in the widget yet
The markup confirmation mechanism is written and covered by tests; the button is not. Until it is, we agree the parking list by email and set it on your account. How the markup works.
Everything here was computed on one account
One pilot, seven years of history, measurements with a date and a method. There will be no “property developers usually…” on this site until there are more accounts.
We will review your funnel by these same rules
To start the conversation we only need your CRM subdomain. What leaves your account for us — and what never does — is on the “Data and access” page.