Contents
In short. The two figures usually answer different questions. A period calculation divides the deals closed in a month by the leads created in that same month, yet many of those closed deals came from earlier months. A cohort calculation takes the leads created in one month and asks how many became customers after the same amount of time. When the sales cycle is long and lead volume changes, the period figure swings for reasons unrelated to how well the sales team works. For decisions, calculate by cohort, compare cohorts of the same age and label any cohort whose sales cycle is still running.
Why one business sees both 3% and 12%
Usually the reports simply answer different questions. An ad platform or lead sheet counts the month’s leads. The CRM, the system that stores contacts and deals, may count deals by close date. A third report may include only deals created and closed inside the chosen dates. Each has its own numerator and denominator.
The second cause is age. Yesterday’s lead has not had time to become a customer; a three-month-old lead has. Mix them, and the result depends on the share of young leads.
Systems do select data differently. Per Bitrix24 documentation, a deal falls into a reporting period if it was added, edited or closed in it, or is still in progress, with separate date filters for records created in the period. One month in two reports from one system can hold different deals.
This month’s deals and this month’s leads are different people
31 of the 34 deals closed that month came from earlier months.
A period calculation divides the deals closed in a month by the leads of the same month. A cohort calculation takes the leads created in one month (that group with a shared starting point is the cohort) and counts how many became deals. The deal belongs to the month of the lead, not the month it closed.
An illustrative example with invented numbers. Before April, 200 leads arrived each month, and of each 200, 2, 6, 5, 3 and 1 become customers in months 0 to 4 (month 0 is the lead’s own month). In June and July leads tripled to 600; in August there were 300. The July cohort was weaker in its first two months. Snapshot: 31 August, each lead and deal counted once.
The period figure ranges from 3.5% to 11.3%, although the April, May and June cohorts behave identically. When leads rose in June the denominator jumped at once, while their deals came later. In August leads fell, but June and July deals were still closing: 31 of the 34 deals closed that month came from earlier months.
A third figure, all April to August deals over all their leads, 93 of 1,900 (4.9%), blends cohorts aged zero to four months and does not help either. A period calculation works when the cycle is much shorter than the period and volume is steady: the editors’ inference, not a platform rule.
| Item | Deals closed in month / leads in month |
|---|---|
| April | 8.5% |
| May | 8.5% |
| June | 3.5% |
| July | 5.5% |
| August | 11.3% |
Illustrative example, snapshot on 31 August. Figures from the table below.
Scroll right to see all columns
| Month | Leads in month | Deals closed in month | Of these, from the same month's leads | Period conversion |
|---|---|---|---|---|
| April | 200 | 17 | 2 | 8.5% |
| May | 200 | 17 | 2 | 8.5% |
| June | 600 | 21 | 6 | 3.5% |
| July | 600 | 33 | 6 | 5.5% |
| August | 300 | 34 | 3 | 11.3% |
How long to wait before judging a cohort
Cohort age is the time from the cohort date to the snapshot. Analytics tools define a cohort by a first event: Google Analytics cohort exploration offers an inclusion criterion such as first touch and a day, week or month granularity; Mixpanel and Amplitude use the date of the starting event; Yandex builds Metrica cohorts by first visit. These describe users. For deals you need CRM dates.
Maturity is an editorial method, not a standard. A cohort is mature when its age covers the time in which most eventual deals normally close. Take older finished cohorts and find the age by which a chosen share of the total closed. The 90% threshold is chosen for the example, not from documentation. April reaches it in month 3, so on 31 August the April (age 4) and May (age 3) cohorts are mature; June, July and August are not.
A calendar month is crude: a lead from the month’s last day had less time in month 0. Counting days from each lead’s create date, for example deals within 30, 60 and 90 days, is more exact. The free funnel calculator on this site (/en/tools/funnel/) asks whether the cohort has completed the usual sales cycle; a maturity rule lets you answer deliberately.
| Item | April cohort, cumulative |
|---|---|
| Month 0 | 11.8% |
| Month 1 | 47.1% |
| Month 2 | 76.5% |
| Month 3 | 94.1% |
| Month 4 | 100% |
2, 8, 13, 16 and 17 of 17 eventual deals. The 90% maturity threshold is chosen for the example.
Compare cohorts of the same age, not of the same month
In the table each row is a month’s cohort, each column an age, each cell the cumulative share of leads that became deals; a dash means that age is not reached. Read down the columns: within one the age is the same.
At month 0 every cohort stands at 1.0%. At month 1 the April, May and June cohorts reach 4.0% and July 3.0%: 18 deals instead of the 24 that 4.0% of 600 leads would give. The gap is visible only because the comparison is made at one age.
For an immature cohort you can derive a rough reference by assuming it follows the shape of a mature one. The June cohort has 39 deals at month 2, and the April cohort had closed 76.5% of its total by that age. So 39 / 0.765 gives about 51 deals, or 8.5% of 600 leads. This is not a forecast: it holds only while the cohorts really are alike.
Scroll right to see all columns
| Lead cohort | Leads | Month 0 | Month 1 | Month 2 | Month 3 | Month 4 |
|---|---|---|---|---|---|---|
| April | 200 | 1.0% | 4.0% | 6.5% | 8.0% | 8.5% |
| May | 200 | 1.0% | 4.0% | 6.5% | 8.0% | - |
| June | 600 | 1.0% | 4.0% | 6.5% | - | - |
| July | 600 | 1.0% | 3.0% | - | - | - |
| August | 300 | 1.0% | - | - | - | - |
Each stage is a subset of the one before
A funnel is a chain of stages: lead, contact, qualified lead, target step (a meeting, viewing or demo) and deal. Count unique leads of one cohort; each stage sits inside the previous one and cannot be larger. If your CRM lets a deal skip a stage, a lead counts in a stage if it reached that stage or any later one.
Google Analytics distinguishes a closed funnel, entered only at the first step, from an open one, entered at any step. For leads choose closed: every stage counts the same cohort. The funnel calculator on this site checks the nesting and rejects a stage larger than the one before.
Illustrative example: the April cohort, 200 leads, cycle complete. 80.0% of leads reach contact, 56.3% of contacts qualify, 44.4% of qualified leads reach the target step, and 42.5% of those become deals. From original leads to deal: 8.5%.
| Item | Value |
|---|---|
| Leads | 200 leads |
| Contact | 160 leads |
| Qualified | 90 leads |
| Target step | 40 leads |
| Deals | 17 leads |
Illustrative example, cycle complete.
Do not drop a young cohort, and do not average it in
Do not drop young cohorts from the report, but do not average them with mature ones.
For an immature cohort you only know deals are at least what you see now; the total will grow. Statisticians call such observations right-censored: the event has not happened by the end of observation. SAS documentation notes such data cannot be analyzed by simply ignoring them. For sales: do not drop young cohorts from the report, but do not average them with mature ones.
Mixpanel marks still-running periods with an asterisk and, for unsegmented queries, averages only completed periods, since recent users have not reached later intervals and would depress the last values. Amplitude also flags incomplete days with an asterisk. Label every cohort’s age and show mature and immature cohorts apart.
The numbers show where leads are lost, not why
The July cohort lags at month 1 after volume tripled. The reason could be a different lead mix or source, a slow first reply, seasonality, statuses entered late or a changed lead definition. The numbers alone cannot tell these apart.
Yandex warns that cohorts of different size and composition lead to wrong conclusions, and that external events and seasonality affect behavior. The stage with the largest reduction, as in the calculator, is a reason to investigate, not proof of a cause: here the biggest loss, 70 leads, is between contact and qualified.
The order: check dates and statuses, split the cohort by source, form one hypothesis and test it with a small change. If people read "qualified" differently, that is a separate task: align it using the article on agreeing how to judge lead quality.
Team task: one table of cohorts by age
For an analyst or marketer with CRM access. The result is one table of cohorts by age with a maturity label. The sequence is the editors’ proposal, not platform documentation.
- Define a lead (a unique contact), the lead date, a deal and the closed-won stage.
- Export one row per lead: create date, source, stage entry dates, close date and outcome; check whether create dates were edited by hand.
- Assign each lead to its month cohort and calculate its age at the snapshot date.
- From older finished cohorts, find the age by which your chosen share of the total closed: that is the maturity rule.
- Build the cumulative conversion table by age, leaving unreached cells empty.
- Compare cohorts only within a column; enter a mature cohort’s stage counts into the funnel calculator.
- For each lag, record hypotheses, the check, an owner and the next snapshot date.
Where this calculation stops working
All numbers are invented and describe no market or company. There is no universal conversion benchmark: the sources lack a method and date to justify one. Results depend on product, price, channel, lead definition and cycle length.
The model assumes one deal per lead and leaves out repeat purchases, refunds and several deals from one contact. Small cohorts give unstable shares, and a difference of a few deals is not a finding. Google Analytics, Metrica, Mixpanel and Amplitude describe users, not CRM deals. The calculator reviews one cohort and only the data you enter; tracking cohort age and comparing cohorts is your job.
Key decisions
- calculate conversion for the leads of one month, not for the deals closed in that month
- compare cohorts of the same age and label cohorts with an unfinished sales cycle
- use the table to find a stage to investigate, and establish the cause of losses separately
What to discuss with your team
- The report states which date selects the deals and which date selects the leads
- Every lead cohort is labeled with its age at the snapshot date
- Immature cohorts are shown separately and not averaged with mature ones
- Stage counts are nested: each stage is no larger than the previous one
- The maturity rule is written down and based on older finished cohorts
- Any cause of losses is recorded as a hypothesis with a check, an owner and a date
Terms in plain language
- Cohort
- A group sharing a starting point, here all leads created in one month, followed together over time
- Cohort age
- Time between the cohort date and the snapshot date, in months or days
- Mature cohort
- A cohort at least as old as the time in which most deals normally close, by a written rule
- Period conversion
- Deals closed in a period divided by leads of the same period, regardless of when each deal's lead arrived
- Nested stages
- Funnel stages where each one is a subset of the previous one and so cannot be larger
Sources
- Google Analytics: Cohort explorationsupport.google.com · checked 31/10/2026
- Google Analytics: Funnel explorationsupport.google.com · checked 31/10/2026
- Yandex Advertising educational article: cohort analysis and building cohorts in Metrica (in Russian, updated 23 Sep 2024)b2b.yandex.ru · checked 31/10/2026
- Mixpanel: Retention report documentationdocs.mixpanel.com · checked 31/10/2026
- Amplitude: Interpret retention analysisamplitude.com · checked 31/10/2026
- HubSpot: Default deal propertiesknowledge.hubspot.com · checked 31/10/2026
- Bitrix24 Helpdesk: Reporting period in CRMhelpdesk.bitrix24.com · checked 31/10/2026
- SAS/STAT: Overview of survival analysis, right-censoringsupport.sas.com · checked 31/10/2026