# Methodology: How fast does a B2B CRM go stale?

By Cristian Frunze, Founder of Ken · Oct 4, 2026 · [Ken Research](/research) No. 3 · Sales teams

The full methodology behind [How fast does a B2B CRM go stale?](/research/crm-decay.md): the data, every method, every number and every limitation. The same text as the [methodology PDF](/research/crm-decay.pdf).

Pre-registered on 2026-10-16, before any data was pulled.

## 1. Question and headline

What share of B2B decision makers change company, title or both within 12 months, and what monthly loss does that imply for a CRM?

Of 355,160 decision makers in a role on 2023-10-01, **13.2%** had a different company or a different title at the same company on 2024-10-01 (95% CI 10.8% to 15.6%, delete-one-range jackknife). 9.6% changed company and 3.6% changed title at the same company (a lower bound). Assuming a constant monthly rate, that is a monthly decay of 1.17% (95% CI 0.95% to 1.40%), or **117 contacts a month in a 10,000-contact CRM** (95% CI 95 to 140).

## 2. Data

All people and company data is **Ken internal data**: Ken's dataset of public professional profiles, work histories and company records. It covers profiles in six countries (United States, United Kingdom, Canada, Australia, Ireland, New Zealand). For each person it holds a list of role spells with title, company, company link (when the employer has a company record), start and end month. Company records hold a size band and employee count. Ken internal data also includes a job-change table built by comparing two snapshots of the profile data, used here only as a cross-check.

Properties that matter for this study:

- Profiles are self-reported. People add new roles late, leave short roles off and often edit a current role instead of adding a new one after a promotion.
- Only currently active profiles exist. People whose profiles were deleted are not in the data.
- Role dates are at month precision.
- Company size is the current band, not the band at the reference date.

## 3. Sample

We drew 20 evenly spaced slices of the profile index, each covering 0.05% of the profile id space (1% in total). Every non-deleted profile in a slice is in the sample: 2,679,597 profiles with 2,030,786 role spells that overlap 2022-10-01 to 2025-10-01. Spreading the slices over the id space avoids tying the sample to one period in which profiles were added. Slice sizes vary from 7,358 to 345,536 profiles, because older parts of the id space are sparse.

Data cleaning: 57 spells that end before they start were dropped. 478 spells with no end date that are not marked current were treated as open, like every other spell with no end date.

## 4. Definitions

- **A role covers a date** d when it started on or before d and has no end date or ends after d.
- **Decision maker.** Titles are normalized (lower case, punctuation to spaces) and given a level by the shared sales title taxonomy. Decision makers are levels manager, director, VP, C-level and founder or owner. In that taxonomy, titles where "manager" names an individual-contributor job (account manager, product manager, project manager and similar) are not managers.
- **Reference role.** Among a person's decision-maker roles covering the reference date, the one with the latest start; ties go to the higher level, then to the role listed first on the profile.
- **Same company.** Company links are equal, or, when either role has no company link, the normalized company names are equal.
- **Status 12 months later.** *Unchanged*: a role at the same company covering the follow-up date has the same normalized title. *New title, same company*: still at the company but under a different title. *Changed company*: no role at the same company covers the follow-up date. This includes people with another role (26,816 in the primary cohort) and people with no role covering that date (7,272). *Any change* is changed company plus new title.
- **Function** comes from the same taxonomy, collapsed into reporting groups, plus study rules for finance, legal, operations, healthcare, education, retail and hospitality, property and construction, creative and design, and level-only titles.
- **Frontline supervisory titles** (post hoc sensitivity, added after a pilot): store, shift, restaurant and similar managers, supervisors, team leads and assistant managers.

## 5. Estimation

- Yearly rate p for each outcome, one row per person.
- Wilson 95% intervals at the person level, reported for every split.
- Because the sample is 20 slices and rates differ a lot between slices, the headline intervals use a **delete-one-slice jackknife** over the 20 slices. That interval is about ten times wider than the Wilson interval and is the one we quote.
- Monthly decay m = 1 - (1 - p)^(1/12), assuming people leave at a constant monthly rate, and contacts lost per month per 10,000 = 10,000 x m. Interval ends are converted the same way.
- Groups under 50 people are suppressed in every output.

## 6. Results

### 6.1 Primary cohort (2023-10-01 to 2024-10-01)

| Outcome | Rate | 95% CI (jackknife) | 95% CI (Wilson) |
|---|---|---|---|
| Any change | 13.2% | 10.8% to 15.6% | 13.1% to 13.3% |
| Changed company | 9.6% | 7.6% to 11.6% | 9.5% to 9.7% |
| New title, same company (lower bound) | 3.6% | 3.1% to 4.1% | 3.5% to 3.6% |

Status counts: unchanged 308,349, new title 12,723, changed company with another role 26,816, changed company with no role on the follow-up date 7,272.

### 6.2 Reference dates and sensitivities

| Reference date | Follow-up | People | Any change | 95% CI (Wilson) | 95% CI (jackknife) | Excluding frontline titles | Company in companies index |
|---|---|---|---|---|---|---|---|
| 2023-04-01 | 2024-04-01 | 352,042 | 13.4% | 13.3% to 13.5% | 11.0% to 15.8% | 13.0% | 17.5% |
| 2023-10-01 | 2024-10-01 | 355,160 | 13.2% | 13.1% to 13.3% | 10.8% to 15.6% | 12.9% | 17.3% |
| 2024-04-01 | 2025-04-01 | 357,897 | 13.0% | 12.9% to 13.1% | 10.6% to 15.4% | 12.8% | 17.2% |
| 2024-10-01 | 2025-10-01 | 359,031 | 12.3% | 12.2% to 12.4% | 9.8% to 14.7% | 12.2% | 16.5% |

### 6.3 Company link

| Group | People | Any change | 95% CI (Wilson) | Changed company | New title, same company |
|---|---|---|---|---|---|
| Reference role has a company id | 232,844 | 17.1% | 17.0% to 17.3% | 11.8% | 5.3% |
| No company id | 122,316 | 5.7% | 5.6% to 5.8% | 5.3% | 0.4% |

For people whose reference role has a company id, the jackknife interval is 14.2% to 20.0%, and the monthly loss is about 155 contacts per 10,000.

### 6.4 By seniority (2023-10-01)

| Level | People | Any change | 95% CI (Wilson) | Changed company | New title, same company |
|---|---|---|---|---|---|
| Manager | 171,173 | 15.9% | 15.7% to 16.1% | 11.4% | 4.5% |
| Director | 79,322 | 15.3% | 15.0% to 15.5% | 11.0% | 4.3% |
| VP | 20,141 | 16.8% | 16.3% to 17.3% | 11.5% | 5.3% |
| C-level | 25,667 | 8.0% | 7.6% to 8.3% | 6.6% | 1.4% |
| Founder or owner | 58,857 | 3.5% | 3.4% to 3.7% | 3.2% | 0.3% |

Excluding frontline titles, the manager rate is 16.1% (130,998 people).

### 6.5 By function (2023-10-01)

| Function | People | Any change | 95% CI (Wilson) | Changed company | New title, same company |
|---|---|---|---|---|---|
| Data and analytics | 1,439 | 24.3% | 22.2% to 26.6% | 14.0% | 10.4% |
| Marketing | 10,330 | 23.7% | 22.9% to 24.5% | 16.9% | 6.8% |
| Product | 5,915 | 20.1% | 19.1% to 21.1% | 13.1% | 7.0% |
| HR and recruiting | 7,395 | 18.5% | 17.7% to 19.4% | 12.8% | 5.8% |
| Sales | 21,714 | 18.1% | 17.6% to 18.6% | 13.2% | 4.9% |
| Legal | 1,959 | 17.9% | 16.2% to 19.6% | 11.8% | 6.0% |
| Creative and design | 4,561 | 17.7% | 16.7% to 18.9% | 14.1% | 3.6% |
| Customer success and support | 3,912 | 16.6% | 15.5% to 17.8% | 11.2% | 5.4% |
| Finance | 10,508 | 16.3% | 15.6% to 17.0% | 10.7% | 5.6% |
| Other | 83,282 | 16.2% | 16.0% to 16.5% | 11.5% | 4.7% |
| Retail and hospitality | 9,560 | 16.0% | 15.3% to 16.7% | 13.2% | 2.8% |
| Engineering and IT | 16,270 | 16.0% | 15.4% to 16.6% | 10.7% | 5.3% |
| Operations | 28,949 | 15.6% | 15.2% to 16.0% | 10.9% | 4.7% |
| Education | 2,623 | 15.3% | 14.0% to 16.8% | 11.5% | 3.8% |
| Healthcare | 7,539 | 14.7% | 13.9% to 15.5% | 10.4% | 4.3% |
| Property and construction | 6,686 | 13.8% | 13.0% to 14.7% | 10.9% | 2.9% |
| Consulting | 2,764 | 13.5% | 12.3% to 14.9% | 9.8% | 3.8% |
| General management | 29,042 | 10.0% | 9.7% to 10.4% | 7.4% | 2.6% |
| Level only (for example "director") | 42,236 | 9.7% | 9.4% to 10.0% | 7.8% | 1.9% |
| Founder or owner | 58,476 | 3.4% | 3.3% to 3.6% | 3.2% | 0.3% |

"Other" holds 83,282 people (23.4% of the cohort) whose titles no rule assigns.

### 6.6 By company size (2023-10-01)

| Employees | People | Any change | 95% CI (Wilson) | Changed company | New title, same company |
|---|---|---|---|---|---|
| 1-10 | 18,461 | 10.5% | 10.0% to 10.9% | 9.2% | 1.2% |
| 11-50 | 23,138 | 18.6% | 18.1% to 19.2% | 14.5% | 4.1% |
| 51-200 | 25,361 | 20.4% | 19.9% to 20.9% | 15.1% | 5.4% |
| 201-500 | 17,300 | 19.1% | 18.6% to 19.7% | 13.4% | 5.7% |
| 501-1,000 | 13,709 | 19.2% | 18.6% to 19.9% | 13.0% | 6.2% |
| 1,001-5,000 | 30,409 | 18.3% | 17.9% to 18.8% | 12.1% | 6.2% |
| 5,001-10,000 | 13,571 | 17.1% | 16.4% to 17.7% | 10.7% | 6.4% |
| 10,001+ | 56,928 | 16.1% | 15.8% to 16.4% | 9.4% | 6.7% |
| Unknown (no company record or no band) | 156,283 | 7.9% | 7.8% to 8.1% | 6.8% | 1.1% |

### 6.7 By country (2023-10-01)

| Country | People | Any change | 95% CI (Wilson) | Changed company | New title, same company |
|---|---|---|---|---|---|
| Australia | 16,552 | 14.9% | 14.4% to 15.5% | 10.6% | 4.3% |
| United Kingdom | 41,142 | 14.7% | 14.4% to 15.1% | 10.3% | 4.4% |
| Ireland | 3,654 | 14.7% | 13.6% to 15.9% | 10.4% | 4.3% |
| Canada | 20,909 | 13.4% | 12.9% to 13.8% | 9.7% | 3.7% |
| United States | 269,979 | 12.8% | 12.7% to 12.9% | 9.4% | 3.4% |
| New Zealand | 2,924 | 12.4% | 11.3% to 13.7% | 8.8% | 3.6% |

### 6.8 By sample slice (2023-10-01)

| Id range | People | Any change | No company id | Any change, with company id |
|---|---|---|---|---|
| 0 | 479 | 9.4% | 40.7% | 13.7% |
| 1 | 17,006 | 13.2% | 32.7% | 17.6% |
| 2 | 27,221 | 11.2% | 36.6% | 15.4% |
| 3 | 25,083 | 10.8% | 37.5% | 15.0% |
| 4 | 25,027 | 10.7% | 38.1% | 15.2% |
| 5 | 26,647 | 11.2% | 37.4% | 15.8% |
| 6 | 26,928 | 11.3% | 37.6% | 15.9% |
| 7 | 27,487 | 11.3% | 37.3% | 15.7% |
| 8 | 39,768 | 11.0% | 38.0% | 15.5% |
| 9 | 36,229 | 6.7% | 27.3% | 8.0% |
| 10 | 17,833 | 17.2% | 28.4% | 20.7% |
| 11 | 6,054 | 7.7% | 49.7% | 12.1% |
| 12 | 2,442 | 14.0% | 39.7% | 19.6% |
| 13 | 26,795 | 19.7% | 29.4% | 23.2% |
| 14 | 22,871 | 21.5% | 28.1% | 24.8% |
| 15 | 3,358 | 21.7% | 33.4% | 25.1% |
| 16 | 5,628 | 28.0% | 31.1% | 32.6% |
| 17 | 4,448 | 22.8% | 36.1% | 26.4% |
| 18 | 11,015 | 19.2% | 31.5% | 22.0% |
| 19 | 2,841 | 24.4% | 36.0% | 25.6% |

## 7. Pre-registered hypothesis

We expected 18% to 28% any change. The hypothesis is rejected by its own test: the pre-registered Wilson interval (13.1% to 13.3%) lies entirely below 15%. With the wider jackknife interval (10.8% to 15.6%) the conclusion is the same for the 18% to 28% range, but the 15% falsification line falls just inside the interval.

1. C-level company change below manager: holds (6.6% against 11.4%). The full ordering we guessed does not: VPs (11.5%) change company about as often as managers.
2. Under 50 employees above 10,001+ on company change: holds (12.2% against 9.4%).
3. 2024-10-01 below 2023-10-01: holds (12.3% against 13.2%).

## 8. Cross-check with the job-change table

The job-change table in Ken internal data comes from one comparison of two snapshots (every event has detection month 2026-06). It flagged 1,125,921 company changes by people whose previous seniority was a decision-maker level, against 73,268,675 current decision-maker profiles: 1.54% in one comparison window. It counts only company changes, and only those caught between two snapshots a few months apart, so it is a lower bound on a few months of company changes. That is consistent with, and below, the 1.17% monthly rate from the cohort.

## 9. Verification

An independent re-analysis, written from the study plan, the shared taxonomy and the data extract alone without seeing our code, recomputed the headline numbers. Its first run matched cohort size and the headline rates at reported precision. It also found 5,296 people with tied reference roles, a case the plan did not specify. Our first run broke those ties in an arbitrary order, which moved the with-company-id group by 554 people and the VP rate by 0.1 points. We fixed the tie-break (role listed first on the profile, the same rule the re-analysis used) and reran. After the fix, all 14 compared counts match exactly: cohort size, every status count, the with-company-id group, every level and the 2024-10-01 cohort.

## 10. Alternative explanations and limitations

- **Late updates make every rate a floor.** Rates fall with later reference dates (13.4% to 12.3%). The primary cohort has two years of follow-up time, but some changes from 2023-2024 will still be missing.
- **Internal title changes are undercounted.** People often edit their current role after a promotion. We report new-title-same-company as a lower bound.
- **Quiet profiles.** 34.4% of the cohort have a reference role with no company link, and only 5.7% of them change. Some of these profiles may simply not be maintained. If so, the true rate is closer to the with-company-id figure (17.1%). We cannot separate unmaintained profiles from stable careers with this data, so we report both.
- **Sample slices differ.** Slice rates run from 6.7% to 28.0%, and the difference remains among people with a company link. Slices probably reflect when profiles were added and how often they are refreshed. The jackknife interval accounts for this between-slice variation, but the sample of slices is small.
- **Survivorship.** Deleted profiles are not in the data. People who leave the workforce or delete their profile are missing, which again pushes rates down.
- **Who counts as a decision maker.** Levels come from titles alone. The definition includes frontline supervisors. Excluding them changes the headline from 13.2% to 12.9%.
- **Constant monthly rate.** The monthly figure assumes changes are spread evenly through the year. It describes average loss, not the shape within a year.
- **Company size** is the current band, so a company that grew is counted at its current size.
- **Coverage.** Six English-speaking countries only. Non-English titles are classified less well.
- **Comparison with vendor figures.** Commonly quoted figures of 20% to 30% yearly B2B data decay usually also count bounced emails, changed phone numbers and closed companies. They are not directly comparable to job and title changes.
- **Design.** This is a descriptive cohort study. Differences between groups are associations, not causes.

## 11. Sources

- Ken internal data: Ken's dataset of public professional profiles, work histories and company records (people, company and job-change tables), queried 2026-10-16.
- Title taxonomy: the shared sales title taxonomy, built for the Ken Research study *Who are GTM engineers, and where do they come from?*