# Methodology: Who gets cut first in a B2B layoff? Recruiters, then SDRs

By Cristian Frunze, Founder of Ken · Oct 14, 2026 · [Ken Research](/research) No. 8 · Hiring and org design, Sales teams

The full methodology behind [Who gets cut first in a B2B layoff? Recruiters, then SDRs](/research/who-gets-cut-first-in-layoffs.md): the data, every method, every number and every limitation. The same text as the [methodology PDF](/research/who-gets-cut-first-in-layoffs.pdf).

Ken Research, October 2026. Data: Ken internal data.

## 1. Summary

**Question.** When B2B companies shrank in 2022-2024, which functions lost the biggest share of their people: recruiting, marketing, sales, customer success, support or engineering? How does it differ by company size and industry, and which functions grew back first?

**Answer.** Recruiting. Across 5,482 software and IT companies with 50 or more people in Ken internal data whose headcount fell 10% or more from a peak between 2022 Q1 and 2024 Q4, and that were still hiring afterwards, recruiters were 2.188% of headcount at the peak and 4.074% of the people lost: a raw cut index of 1.862. Recruiting also shrank at the 5,565 software and IT companies that did not shrink (by 9.23% over the same quarters, while their total headcount grew 12.77%). Net of that normal churn, recruiting's cut index is 1.262 (95% interval 1.108 to 1.416), the highest of the six functions: marketing 1.139, customer success 1.084, sales 1.068, engineering 1.056 and support 0.805. Inside sales, SDRs and BDRs were cut hardest (1.632) and account managers least (0.856).

Recruiting was also the slowest to come back. In 20.8% of the episodes where it had been cut, recruiting was back at its peak headcount within 8 quarters of the company's low point, against 41.2% for sales, 38.9% for engineering and 36.1% for the companies as a whole.

**Hypotheses.** Of the seven pre-registered hypotheses, three were right and four wrong on the pre-registered measure (section 7). The two clearest misses: engineering was not the least-cut function (support was), and recruiting's net cut index was below the 1.5 we expected.

**One change after the first run.** The pre-registered net cut index divides each function's share of the excess loss by its share of headcount at the peak. That version is not 1 when a company cuts every function in proportion to the size it would otherwise have had, so it mixes normal churn back in. The headline net cut index divides by the function's share of *expected* headcount instead (section 5.2). Both are reported everywhere, and the hypotheses are scored on the pre-registered one as written. On the pre-registered version recruiting's net index is 1.016.

**Verification.** An independent re-analysis, written from the study plan alone without seeing our code, recomputed the study from its definitions and agreed with it on 123 of 127 comparable numbers at the published precision and on the other 4, bootstrap interval ends, within sampling noise (section 8). The comparison found one real difference: the first run counted each recruiter once per company and quarter, while the study plan, like the shared build for every other function, counts a person once per title level. Recruiting is now counted the study plan's way. Its headline net index moved from 1.265 to 1.262, and no conclusion changed.

## 2. Data

### 2.1 Ken internal data

Ken internal data is Ken's dataset of public professional profiles, work histories and company records. Its people data holds only people whose current country is the US, the UK, Canada, Australia, Ireland or New Zealand (earlier roles can be anywhere), so every headcount in this study means people in these six countries. A company with staff elsewhere looks smaller than it is, and a company that moved jobs out of these countries looks as if it cut them. Profiles were last updated up to 2026-04-29.

### 2.2 The company headcount history

The study uses the shared company headcount history, built for the Ken Research study *Did AI kill the SDR role?* (build finished 2026-10-17). It counts a person at a company at a quarter end when one of their role spells there covers that date: the spell started on or before it, and ended on or after it (the end month counts in full) or is current, with current spells running to the 2026-04-30 snapshot. Spells with no start date, or with no end date that are not current, are dropped. Each job title is normalized (lower case, punctuation removed) and given a function and a level by the shared sales title taxonomy, built for the Ken Research study *Who are GTM engineers, and where do they come from?*, plus a customer-support rule added by the SDR study. Of the 555,053,752 role spells in the extract, 342,882,318 cover at least one quarter end from 2015 Q1 to 2026 Q1 and are kept.

This study reads quarter ends 2021 Q1 to 2026 Q1. 2026 Q1 is the last complete quarter before the snapshot, and every time series ends there.

### 2.3 One extra pass: recruiters and joiners

The shared function label `recruiting_hr` mixes recruiters with human resources, HR business partner, people partner, staffing and sourcing titles. One pass over the person-level spells (on Ken's servers; no person-level row leaves them) split the label with the pre-registered recruiter rule (recruit, recruiter, recruiting, recruitment, talent acquisition, talent partner, talent advisor, talent lead, talent sourcer, talent sourcing, talent associate, sourcer, headhunter). Of the 4,804,996 kept spells with the label, 2,567,747 have a recruiter title, and recruiters are 50.5% of the people with a `recruiting_hr` title at the study's companies. The pass applies the same counting rules as the shared build and reproduces it exactly: 0 of 3,645,789 company-quarter cells differ, with 17,713,797 person-quarters on both sides. The split was listed title by title for every title with 50 or more spells. Human resources titles, "staffing coordinator", "staffing manager", "sourcing specialist" and "strategic sourcing manager" stay out of recruiting. Recruiting and the HR remainder are then counted like every other function, once per person, company, quarter and title level (section 4). The first run counted each recruiter once per company and quarter; that version is kept as a check row ("Recruiting, each person once") and is discussed in section 8.

The same pass counted joiners: people whose earliest spell at a company starts in a given quarter (76,985,401 joins at known companies from 2021 Q1 to 2026 Q1). A promotion inside a company starts a new spell but is not a join.

### 2.4 Late reporting

People add new jobs to their profiles months late. The SDR study measured this with Ken's monthly job-change feed: of 1,822,157 job changes first detected in June 2026, 41.1% had started in 2026 Q1, and 41.5% had started before 2025 Q4. So the latest quarters undercount new jobs, and some jobs appear more than a year late. This matters most for the regrowth measures, which end in 2025-2026, and for falls that end in 2024 Q4 (sections 3.2 and 6.9).

## 3. Sample

### 3.1 Companies

173,609 companies with a known company id had at least 50 people in Ken internal data at some quarter end from 2022 Q1 to 2024 Q4. 151,784 of them have a company record, and 12,203 of those carry one of the 18 software and IT industry labels that the SDR study audited (software development, the software product labels, IT services and consulting and the other IT system labels, data infrastructure and analytics, technology, information and internet, internet marketplace and social networking platforms, and computer and network security). The 21,825 companies without a record have no industry and are left out.

### 3.2 Contraction episodes

For each company we take the largest peak-to-trough fall in headcount inside 2022 Q1-2024 Q4, counting only peaks of 50 or more people. The peak is the last quarter at the peak level before the trough, and the trough is the earliest quarter of the largest fall. A company is a contraction episode when that fall is 10% or more. It is in the primary sample if it is still operating: at least one person joins it in the 4 quarters after the trough. Shutdowns, acquisitions and moves to a new company page also look like falls, but nobody joins afterwards; 82.6% of the software and IT companies with a fall pass this rule, and 88.4% of all companies with a fall do.

| | Software and IT | All companies |
|---|---|---|
| Companies with a record and 50+ people at some quarter end 2022 Q1-2024 Q4 | 12,203 | 151,784 |
| Fell 10% or more from a peak of 50+ people | 6,638 | 58,223 |
| ... of which nobody joined in the 4 quarters after the trough | 1,156 | 6,753 |
| Contraction episodes (primary sample) | 5,482 | 51,470 |
| Comparison companies (no fall of 10% or more) | 5,565 | 93,561 |
| Median fall, peak to trough (%) | 23.5 | 18.0 |
| Median people at the peak | 97.0 | 85.0 |
| People at the peak, summed | 1,670,162 | 12,315,737 |
| People lost, peak to trough, summed | 438,146 | 2,638,570 |
| Median quarters from peak to trough | 6.0 | 6.0 |
| Episodes with a peak of 100+ people (robustness f) | 2,662 | 21,064 |
| Episodes with a fall of 10-20% | 2,223 | 29,521 |
| Episodes with a fall of 20-30% | 1,258 | 10,645 |
| Episodes with a fall of 30-50% | 1,347 | 8,132 |
| Episodes with a fall of over 50% | 654 | 3,172 |

Peaks cluster at the start of the window and troughs at its end. In software and IT, 1,180 episodes peak in 2022 Q1, the first window quarter, so many companies had peaked earlier. 2,729 of the 5,482 software and IT episodes are at their lowest in 2024 Q4, the last window quarter, so many falls went on after the window. Robustness check (h) keeps only episodes whose trough came by 2024 Q1. 99.8% of the software and IT episodes have their headquarters in the six countries.

### 3.3 Comparison companies

Comparison companies are companies with a record whose largest fall from a 50-person peak inside the window is under 10%: 5,565 in software and IT and 93,561 in all industries. Each episode is compared with the comparison companies that had 50 or more people at its peak quarter, in the same size band (people at that quarter) and the same industry group, over the same two quarters. A cell with fewer than 20 comparison companies falls back to the size band across industries. All 5,482 software and IT episodes had a full cell (between 3,861 and 5,339 software and IT comparison companies per peak quarter); across all industries 51,469 of 51,470 episodes did, and one fell back.

### 3.4 Industry groups

The industry on each company record is one of the standard industry labels. The 420 labels of the study's companies were mapped to seven groups before any cut index was computed. Healthcare providers (hospitals, practices and care services) are in the consumer and public group because they sell to patients; the healthcare and life sciences group holds biotech, pharma, medical devices and laboratories. Construction, transport and logistics, wholesale and utilities are in manufacturing, industrial and energy. Staffing and recruiting firms are in business and professional services, where recruiters are the product rather than a support team. Records with no industry form their own matching cell and count only in the all-company results.

| Industry group | Industry labels | Panel companies | Episodes | Comparison companies |
|---|---|---|---|---|
| Software and IT | 17 | 12,203 | 5,482 | 5,565 |
| Other technology, telecom and media | 40 | 5,427 | 2,408 | 2,625 |
| Financial services | 17 | 11,088 | 3,095 | 7,384 |
| Business and professional services | 39 | 19,448 | 7,345 | 11,249 |
| Manufacturing, industrial and energy | 150 | 29,940 | 8,625 | 19,913 |
| Healthcare and life sciences | 5 | 3,571 | 1,388 | 1,837 |
| Consumer, public sector and the rest | 151 | 69,511 | 22,885 | 44,663 |
| No industry on the record | 0 | 596 | 242 | 325 |

### 3.5 Size bands

Size is the number of people in Ken internal data at the peak quarter: 50-199, 200-999, 1,000-4,999 and 5,000 or more. In software and IT the four bands hold 4,174, 1,097, 176 and 35 episodes. Robustness check (e) uses the size band on the company record instead, which describes the company in 2026 and counts staff in every country.

## 4. Functions

The functions are groups of the shared title labels, summed over levels (as set in the study plan):

- **Recruiting:** recruiter titles inside `recruiting_hr` (section 2.3).
- **Marketing:** marketing, demand generation, growth marketing and marketing operations.
- **Sales:** SDR and BDR, account executives, account management, general sales titles, business development and partnerships, sales engineering, revenue operations, enablement and GTM roles. SDR/BDR, account executives and account managers are also shown alone.
- **Customer success:** customer success titles.
- **Support:** customer, technical, product, IT and help-desk support titles (the rule from the SDR study), outside sales, social care, health and education contexts.
- **Engineering:** software engineering titles (explicit technical titles and bare developer or software titles), other titles with "engineer" (systems, network, hardware and similar) and growth engineering. Software engineering alone is shown too.
- **For context:** HR (the rest of `recruiting_hr`; the shared taxonomy only catches "human resources", "HRBP" and "people partner", so short titles such as "HR manager" sit in all other roles), product, data and analytics, operations, founders, interns and students, and all other roles.

Total headcount counts each person once per company and quarter. A function counts each person once per company, quarter, title function and level, so a person holding two titles at once can count twice (recruiting and HR, split from `recruiting_hr` by this study, follow the same rule); summed over all functions, the person-function counts are 3.5% above total headcount. Levels come from titles: individual contributors (no level word, or senior, staff, principal) and managers and above (manager, lead, head, director, VP, C-level); founders and interns are in neither.

## 5. Methods

### 5.1 Raw cut index

For a function f, with each episode's peak quarter P and trough quarter T, pooled over the episodes in a cell:

- H_f = sum over episodes of the function's headcount at P; H = the same for total headcount;
- L_f = sum of (function headcount at P minus at T); L = the same for total headcount;
- raw cut index = (L_f / L) / (H_f / H).

It compares the function's share of the people lost with its share of headcount at the peak. Equivalently, it is the function's percentage fall divided by the company's percentage fall. Above 1, the function lost people faster than the company. Pooling weights each episode by its size; robustness check (d) weights every episode equally.

### 5.2 Net cut index

Rebuilt headcount cannot tell a layoff from a resignation, and functions grow and shrink at different rates for ordinary reasons. Recruiting, for example, shrank in 2022-2024 even at companies that did not shrink. So each episode gets an expected level for every function: its headcount at the peak times one plus the change of the same function at the matched comparison companies (same industry group, same size band, same quarters P to T, pooled). The excess loss is the expected level minus the actual headcount at the trough.

- **Net cut index (headline):** (sum of the function's excess losses / sum of total excess losses) / (expected function headcount / expected total headcount). Equivalently, the function's shortfall from its expected level divided by the company's shortfall from its expected level. It equals the raw index when the comparison companies do not change, and 1 when a company cuts every function in proportion to the size it would otherwise have had.
- **Pre-registered net index:** (sum of the function's excess losses / sum of total excess losses) / (H_f / H). It divides by the share of headcount at the peak instead of the expected share. It equals the headline version times (1 + the function's comparison change) / (1 + the company's comparison change), so it understates the cut in functions that were shrinking anyway and overstates it in functions that were growing anyway.

For recruiting in software and IT: recruiting fell 48.84% from peak to trough, while at the comparison companies it fell 9.23% and their total headcount grew 12.77%. Recruiting's shortfall from its expected level was 43.64%, against 34.59% for the company as a whole, so the headline net cut index is 1.262. Recruiting was 2.222% of the excess losses, against 1.762% of expected headcount (headline) and 2.188% of peak headcount (pre-registered, 1.016). The change was made after seeing the first run, because the pre-registered version does not do what the brief asks: net out normal churn. That is why both versions are reported everywhere.

If an episode's cell has nobody in a function at the peak quarter, the function's expected change is the cell's total-headcount change; that happened for 6 episodes for SDR/BDR and 199 for customer success managers, and for none of the six main functions.

### 5.3 Intervals

95% intervals come from a company cluster bootstrap: 500 draws, each giving every company (episodes and comparison companies alike) a Poisson(1) weight, then recomputing the comparison changes and the indices; seed 68. They cover sampling variation across companies, not the measurement issues in section 9.

### 5.4 Regrowth

For each episode and function with at least 5 people at the peak and fewer at the trough, we count the quarters after the company's trough until the function is back at its peak-quarter headcount. The headline is the share back within 8 quarters, for episodes whose trough came by 2024 Q1, so all 8 quarters are observed by 2026 Q1. Also reported: the share back within 4 quarters (every episode), the share not cut at all (function at least as large at the trough as at the peak), and Kaplan-Meier estimates censored at 2026 Q1. The same measure for total headcount gives the company benchmark.

### 5.5 Timing

To see which function falls first, each function's headcount is summed over episodes for every quarter from 4 before to 8 after the company's peak and shown as a percentage of its sum at the peak quarter, for episodes whose peak came by 2024 Q1 (so 8 later quarters are observed). Exploratory, with no hypothesis.

### 5.6 Robustness

(a) The brief's episode definition without the still-operating rule. (b) Falls of 10% to 50% only. (c) Engineering as software engineering only (in every table). (d) Every episode weighted equally: the mean over episodes of each function's share of the loss over the mean share of headcount, and the same with excess losses and expected headcount. (e) Size from the company record. (f) A peak of at least 100 people, for episodes and comparison companies. (g) Regrowth censored at 2025 Q3, and with at least 10 people at the peak. Added after the first run, not pre-registered: (h) only episodes whose trough came by 2024 Q1.

### 5.7 Suppression

A published figure needs at least 50 people (in the function at the peak, summed over the cell's episodes) and at least 20 episodes. Smaller cells are marked suppressed. No individual person or company is named.

## 6. Results

### 6.1 The six functions in software and IT

| Function | People at peak | Share of peak (%) | Share of loss (%) | Change peak to trough (%) | Change at comparison companies (%) | Raw cut index (95% interval) | Net cut index (95% interval) | Pre-registered net index (95% interval) |
|---|---|---|---|---|---|---|---|---|
| Recruiting | 36,551 | 2.188 | 4.074 | -48.84 | -9.23 | 1.862 (1.617-2.100) | 1.262 (1.108-1.416) | 1.016 (0.888-1.160) |
| Marketing | 70,949 | 4.248 | 5.056 | -31.22 | 13.46 | 1.190 (1.128-1.247) | 1.139 (1.077-1.193) | 1.146 (1.071-1.213) |
| Sales | 253,087 | 15.153 | 16.080 | -27.84 | 14.45 | 1.061 (1.000-1.125) | 1.068 (1.005-1.124) | 1.084 (1.003-1.155) |
| Customer success | 35,411 | 2.120 | 2.040 | -25.24 | 19.59 | 0.962 (0.878-1.044) | 1.084 (1.005-1.177) | 1.149 (1.044-1.285) |
| Support | 77,554 | 4.644 | 4.280 | -24.18 | 5.08 | 0.922 (0.841-1.007) | 0.805 (0.719-0.902) | 0.750 (0.655-0.860) |
| Engineering | 350,904 | 21.010 | 21.430 | -26.76 | 15.39 | 1.020 (0.942-1.099) | 1.056 (0.990-1.120) | 1.081 (1.003-1.160) |
| Sales: SDR/BDR | 23,817 | 1.426 | 2.500 | -45.98 | 24.04 | 1.753 (1.616-1.882) | 1.632 (1.504-1.760) | 1.795 (1.591-2.014) |
| Sales: account executives | 52,855 | 3.165 | 3.553 | -29.45 | 18.96 | 1.123 (0.999-1.242) | 1.177 (1.053-1.291) | 1.241 (1.078-1.393) |
| Sales: account managers | 38,250 | 2.290 | 1.745 | -19.99 | 13.65 | 0.762 (0.679-0.855) | 0.856 (0.792-0.926) | 0.863 (0.787-0.944) |
| Engineering: software only | 284,175 | 17.015 | 18.050 | -27.83 | 16.09 | 1.061 (0.980-1.146) | 1.094 (1.025-1.163) | 1.126 (1.044-1.212) |
| Recruiting, each person once (check) | 36,060 | 2.159 | 4.051 | -49.22 | -9.73 | 1.876 (1.633-2.114) | 1.265 (1.113-1.419) | 1.012 (0.887-1.156) |
| HR (rest of the shared label) | 8,884 | 0.532 | 0.607 | -29.95 | 3.69 | 1.142 (1.059-1.227) | 0.938 (0.880-1.027) | 0.863 (0.796-0.975) |
| Product | 33,640 | 2.014 | 1.962 | -25.56 | 19.00 | 0.974 (0.920-1.033) | 1.082 (1.030-1.135) | 1.142 (1.069-1.214) |
| Data and analytics | 32,150 | 1.925 | 2.309 | -31.47 | 20.76 | 1.200 (1.104-1.319) | 1.250 (1.162-1.350) | 1.339 (1.207-1.495) |
| Operations | 29,181 | 1.747 | 1.841 | -27.64 | 13.69 | 1.054 (0.989-1.122) | 1.051 (0.997-1.118) | 1.060 (0.992-1.147) |
| Founders | 11,040 | 0.661 | 0.569 | -22.60 | 9.12 | 0.861 (0.702-1.014) | 0.840 (0.689-0.992) | 0.813 (0.644-0.992) |
| Interns and students | 11,825 | 0.708 | 1.520 | -56.34 | -1.11 | 2.148 (1.907-2.417) | 1.615 (1.422-1.841) | 1.416 (1.110-1.699) |
| All other roles | 776,451 | 46.490 | 39.657 | -22.38 | 12.04 | 0.853 (0.827-0.879) | 0.888 (0.857-0.921) | 0.882 (0.845-0.922) |
| Whole company | 1,670,162 | 100.000 | 100.000 | -26.23 | 12.77 | 1.000 | 1.000 | 1.000 |

Recruiting's raw cut index (1.862) is far above every other function's; netting out normal churn brings it to 1.262, still the highest of the six. Marketing (1.139) was cut harder than its size, customer success (1.084), sales (1.068) and engineering (1.056) a little harder, and support (0.805) less. Inside sales the cuts fell on new business: SDRs and BDRs 1.632 and account executives 1.177, against 0.856 for account managers. Interns and students were cut almost as hard as SDRs (1.615), and data and analytics staff were also cut hard (1.250).

### 6.2 By company size (software and IT)

Net cut index:

| | 50-199 | 200-999 | 1,000-4,999 | 5,000+ |
|---|---|---|---|---|
| Episodes | 4,174 | 1,097 | 176 | 35 |
| Recruiting | 1.022 | 1.298 | 1.287 | 1.463 |
| Marketing | 1.176 | 1.079 | 1.036 | 1.045 |
| Sales | 1.087 | 1.085 | 1.030 | 0.849 |
| Customer success | 1.053 | 0.963 | 0.912 | 1.126 |
| Support | 0.848 | 0.833 | 0.815 | 0.620 |
| Engineering | 1.031 | 1.019 | 1.059 | 1.186 |
| Sales: SDR/BDR | 1.559 | 1.523 | 1.427 | 1.427 |
| Sales: account executives | 1.234 | 1.234 | 1.130 | 0.781 |
| Sales: account managers | 0.843 | 0.796 | 0.921 | 0.809 |

Raw cut index:

| | 50-199 | 200-999 | 1,000-4,999 | 5,000+ |
|---|---|---|---|---|
| Episodes | 4,174 | 1,097 | 176 | 35 |
| Recruiting | 1.249 | 1.733 | 1.964 | 2.562 |
| Marketing | 1.269 | 1.171 | 1.108 | 1.091 |
| Sales | 1.145 | 1.111 | 1.006 | 0.863 |
| Customer success | 0.915 | 0.915 | 0.830 | 1.149 |
| Support | 0.902 | 0.890 | 0.933 | 0.960 |
| Engineering | 1.042 | 1.008 | 1.034 | 1.008 |
| Sales: SDR/BDR | 1.910 | 1.669 | 1.570 | 1.323 |
| Sales: account executives | 1.308 | 1.282 | 1.008 | 0.690 |
| Sales: account managers | 0.743 | 0.691 | 0.802 | 0.812 |

Recruiting has the highest net index of the six at 200-999 people (1.298), 1,000-4,999 (1.287) and 5,000 or more (1.463). At 50-199 people marketing is highest (1.176) and recruiting is close to 1 (1.022). Support has the lowest net index of the six in every band. The 5,000+ band holds only 35 episodes.

All companies, by size (net):

| | 50-199 | 200-999 | 1,000-4,999 | 5,000+ |
|---|---|---|---|---|
| Episodes | 42,413 | 7,657 | 1,190 | 210 |
| Recruiting | 1.298 | 1.376 | 1.399 | 1.192 |
| Marketing | 1.195 | 1.188 | 1.163 | 1.158 |
| Sales | 1.094 | 1.101 | 1.037 | 0.877 |
| Customer success | 1.434 | 1.243 | 1.152 | 1.166 |
| Support | 1.008 | 0.994 | 0.970 | 0.890 |
| Engineering | 1.229 | 1.184 | 1.122 | 1.106 |
| Sales: SDR/BDR | 1.904 | 1.726 | 1.549 | 1.418 |
| Sales: account executives | 1.305 | 1.306 | 1.287 | 0.717 |
| Sales: account managers | 0.991 | 0.945 | 1.014 | 0.960 |

### 6.3 By industry (all companies)

Net cut index:

| | Software and IT | Other technology, telecom and media | Financial services | Business and professional services | Manufacturing, industrial and energy | Healthcare and life sciences | Consumer, public sector and the rest |
|---|---|---|---|---|---|---|---|
| Episodes | 5,482 | 2,408 | 3,095 | 7,345 | 8,625 | 1,388 | 22,885 |
| Recruiting | 1.262 | 1.514 | 1.504 | 1.192 | 1.578 | 1.417 | 1.343 |
| Marketing | 1.139 | 1.232 | 1.197 | 1.233 | 1.333 | 1.238 | 1.188 |
| Sales | 1.068 | 0.908 | 1.055 | 1.093 | 0.973 | 0.987 | 0.993 |
| Customer success | 1.084 | 1.402 | 1.385 | 1.280 | 1.665 | 1.584 | 1.593 |
| Support | 0.805 | 0.753 | 0.953 | 0.997 | 0.948 | 0.866 | 1.017 |
| Engineering | 1.056 | 1.030 | 1.123 | 1.087 | 1.179 | 1.027 | 1.127 |
| Sales: SDR/BDR | 1.632 | 1.512 | 1.538 | 1.585 | 1.381 | 1.054 | 1.818 |
| Sales: account executives | 1.177 | 0.619 | 1.207 | 1.277 | 1.083 | 0.764 | 1.187 |
| Sales: account managers | 0.856 | 1.011 | 1.000 | 0.993 | 0.968 | 0.944 | 1.032 |

Raw cut index:

| | Software and IT | Other technology, telecom and media | Financial services | Business and professional services | Manufacturing, industrial and energy | Healthcare and life sciences | Consumer, public sector and the rest |
|---|---|---|---|---|---|---|---|
| Episodes | 5,482 | 2,408 | 3,095 | 7,345 | 8,625 | 1,388 | 22,885 |
| Recruiting | 1.862 | 1.930 | 1.787 | 1.502 | 1.702 | 1.806 | 1.392 |
| Marketing | 1.190 | 1.324 | 1.254 | 1.278 | 1.403 | 1.336 | 1.181 |
| Sales | 1.061 | 0.879 | 1.089 | 1.019 | 1.015 | 0.976 | 1.100 |
| Customer success | 0.962 | 0.889 | 1.346 | 0.972 | 1.376 | 1.488 | 1.319 |
| Support | 0.922 | 1.099 | 1.044 | 1.080 | 1.069 | 0.931 | 1.194 |
| Engineering | 1.020 | 0.828 | 0.894 | 1.027 | 1.152 | 1.025 | 0.981 |
| Sales: SDR/BDR | 1.753 | 1.125 | 1.636 | 1.572 | 1.148 | 1.090 | 1.754 |
| Sales: account executives | 1.123 | 0.443 | 1.275 | 1.247 | 1.121 | 0.624 | 1.124 |
| Sales: account managers | 0.762 | 0.881 | 0.966 | 0.921 | 0.985 | 0.824 | 0.971 |

Recruiting was cut harder than its size in every industry group, from 1.192 in business and professional services to 1.578 in manufacturing, industrial and energy. It has the highest net index of the six in software and IT, in other technology, telecom and media (1.514) and in financial services (1.504). In the other four groups customer success is highest (1.280 to 1.665), but customer success teams there are small: under half a percent of headcount at the peak, against 2.12% in software and IT. Support has the lowest net index of the six in every group except consumer and public, where sales is lowest.

### 6.4 All companies

| Function | People at peak | Share of peak (%) | Share of loss (%) | Change peak to trough (%) | Change at comparison companies (%) | Raw cut index (95% interval) | Net cut index (95% interval) | Pre-registered net index (95% interval) |
|---|---|---|---|---|---|---|---|---|
| Recruiting | 167,081 | 1.357 | 2.308 | -36.46 | 2.56 | 1.702 (1.616-1.805) | 1.352 (1.294-1.418) | 1.268 (1.206-1.340) |
| Marketing | 375,695 | 3.051 | 3.827 | -26.88 | 11.32 | 1.255 (1.224-1.286) | 1.219 (1.195-1.244) | 1.241 (1.214-1.268) |
| Sales | 1,072,786 | 8.711 | 9.349 | -22.99 | 9.32 | 1.073 (1.045-1.109) | 1.050 (1.023-1.080) | 1.050 (1.017-1.087) |
| Customer success | 54,568 | 0.443 | 0.521 | -25.21 | 21.62 | 1.177 (1.101-1.259) | 1.368 (1.310-1.445) | 1.522 (1.438-1.632) |
| Support | 404,640 | 3.286 | 3.621 | -23.61 | 4.43 | 1.102 (1.071-1.140) | 0.954 (0.920-0.991) | 0.911 (0.873-0.953) |
| Engineering | 1,023,976 | 8.314 | 8.815 | -22.71 | 14.08 | 1.060 (1.011-1.118) | 1.146 (1.102-1.195) | 1.196 (1.142-1.254) |
| Sales: SDR/BDR | 39,805 | 0.323 | 0.606 | -40.16 | 23.04 | 1.874 (1.753-1.995) | 1.825 (1.719-1.924) | 2.053 (1.886-2.222) |
| Sales: account executives | 125,580 | 1.020 | 1.216 | -25.55 | 14.85 | 1.192 (1.092-1.298) | 1.250 (1.156-1.338) | 1.313 (1.197-1.420) |
| Sales: account managers | 147,408 | 1.197 | 1.109 | -19.85 | 11.53 | 0.926 (0.884-0.976) | 1.000 (0.956-1.056) | 1.020 (0.968-1.093) |
| Engineering: software only | 583,483 | 4.738 | 5.491 | -24.83 | 16.14 | 1.159 (1.094-1.234) | 1.253 (1.199-1.312) | 1.331 (1.264-1.405) |
| Recruiting, each person once (check) | 164,311 | 1.334 | 2.288 | -36.74 | 2.25 | 1.715 (1.629-1.819) | 1.355 (1.296-1.421) | 1.267 (1.203-1.339) |
| HR (rest of the shared label) | 101,895 | 0.827 | 0.821 | -21.25 | 8.42 | 0.992 (0.919-1.050) | 0.972 (0.919-1.018) | 0.964 (0.909-1.015) |
| Product | 79,782 | 0.648 | 0.727 | -24.05 | 19.78 | 1.123 (1.058-1.189) | 1.300 (1.249-1.351) | 1.424 (1.359-1.492) |
| Data and analytics | 108,262 | 0.879 | 1.102 | -26.86 | 19.27 | 1.254 (1.192-1.328) | 1.374 (1.327-1.427) | 1.499 (1.437-1.575) |
| Operations | 198,940 | 1.615 | 1.595 | -21.15 | 11.33 | 0.987 (0.939-1.035) | 1.037 (1.001-1.073) | 1.055 (1.016-1.097) |
| Founders | 56,147 | 0.456 | 0.317 | -14.91 | 7.20 | 0.696 (0.623-0.767) | 0.733 (0.666-0.801) | 0.719 (0.648-0.791) |
| Interns and students | 332,373 | 2.699 | 4.504 | -35.75 | 3.02 | 1.669 (1.595-1.746) | 1.337 (1.280-1.391) | 1.260 (1.199-1.313) |
| All other roles | 8,758,983 | 71.120 | 64.416 | -19.40 | 8.96 | 0.906 (0.898-0.913) | 0.925 (0.918-0.931) | 0.922 (0.915-0.928) |
| Whole company | 12,315,737 | 100.000 | 100.000 | -21.42 | 9.35 | 1.000 | 1.000 | 1.000 |

Across all industries, recruiting's raw index (1.702) is again the highest of the six, but customer success has a slightly higher net index (1.368, against 1.352 for recruiting), driven by the small customer success teams outside software.

### 6.5 Individual contributors and managers

| Population | Group | People at peak | Raw | Net (95% interval) | Pre-registered net |
|---|---|---|---|---|---|
| Software and IT | Recruiting, individual contributors | 28,515 | 1.917 | 1.276 (1.115-1.439) | 1.006 |
| Software and IT | Recruiting, managers and above | 7,893 | 1.652 | 1.220 (1.081-1.358) | 1.061 |
| Software and IT | Marketing, individual contributors | 31,324 | 1.203 | 1.121 (1.016-1.218) | 1.111 |
| Software and IT | Marketing, managers and above | 38,194 | 1.125 | 1.140 (1.084-1.200) | 1.176 |
| Software and IT | Sales, individual contributors | 174,741 | 1.072 | 1.086 (1.019-1.144) | 1.109 |
| Software and IT | Sales, managers and above | 77,125 | 1.016 | 1.020 (0.954-1.078) | 1.025 |
| Software and IT | Customer success, individual contributors | 25,120 | 0.995 | 1.119 (1.035-1.220) | 1.196 |
| Software and IT | Customer success, managers and above | 10,165 | 0.860 | 0.988 (0.903-1.083) | 1.031 |
| Software and IT | Support, individual contributors | 63,709 | 0.927 | 0.804 (0.705-0.913) | 0.748 |
| Software and IT | Support, managers and above | 13,791 | 0.893 | 0.802 (0.723-0.881) | 0.753 |
| Software and IT | Engineering, individual contributors | 288,542 | 1.013 | 1.068 (0.994-1.138) | 1.102 |
| Software and IT | Engineering, managers and above | 54,329 | 0.783 | 0.903 (0.847-0.965) | 0.925 |
| All companies | Recruiting, individual contributors | 130,228 | 1.746 | 1.359 (1.288-1.436) | 1.259 |
| All companies | Recruiting, managers and above | 35,737 | 1.521 | 1.324 (1.263-1.387) | 1.300 |
| All companies | Marketing, individual contributors | 202,073 | 1.262 | 1.219 (1.188-1.250) | 1.238 |
| All companies | Marketing, managers and above | 159,674 | 1.105 | 1.153 (1.123-1.181) | 1.191 |
| All companies | Sales, individual contributors | 749,560 | 1.074 | 1.046 (1.012-1.082) | 1.044 |
| All companies | Sales, managers and above | 316,844 | 1.034 | 1.038 (1.006-1.073) | 1.044 |
| All companies | Customer success, individual contributors | 38,690 | 1.217 | 1.402 (1.336-1.486) | 1.565 |
| All companies | Customer success, managers and above | 15,689 | 1.054 | 1.279 (1.210-1.364) | 1.414 |
| All companies | Support, individual contributors | 325,515 | 1.119 | 0.966 (0.928-1.007) | 0.922 |
| All companies | Support, managers and above | 79,024 | 1.030 | 0.900 (0.853-0.952) | 0.859 |
| All companies | Engineering, individual contributors | 848,130 | 1.028 | 1.143 (1.095-1.196) | 1.202 |
| All companies | Engineering, managers and above | 149,643 | 0.869 | 1.024 (0.985-1.067) | 1.070 |

In software and IT, managers were cut a little less than individual contributors in most functions. The gap is clearest in engineering, 0.903 for managers against 1.068 for individual contributors, and in customer success, 0.988 against 1.119. In marketing and support the two levels were cut about alike.

### 6.6 Who falls first

Headcount as a percentage of its level at the company's peak quarter, software and IT episodes with a peak by 2024 Q1:

| Function | -4 | -2 | -1 | peak | +1 | +2 | +4 | +8 |
|---|---|---|---|---|---|---|---|---|
| Recruiting | 82.00 | 96.69 | 100.69 | 100.00 | 88.95 | 77.25 | 62.93 | 54.52 |
| Marketing | 85.33 | 93.14 | 97.05 | 100.00 | 94.69 | 88.99 | 80.14 | 72.82 |
| Sales | 86.07 | 93.33 | 97.33 | 100.00 | 95.50 | 90.61 | 83.41 | 76.12 |
| Customer success | 81.85 | 92.09 | 96.74 | 100.00 | 96.33 | 92.33 | 85.86 | 79.22 |
| Support | 95.26 | 98.11 | 99.12 | 100.00 | 95.82 | 91.98 | 86.27 | 78.38 |
| Engineering | 86.82 | 93.07 | 97.06 | 100.00 | 95.48 | 90.69 | 84.45 | 78.55 |
| Sales: SDR/BDR | 73.83 | 87.59 | 94.80 | 100.00 | 89.99 | 79.81 | 66.85 | 59.33 |
| Whole company | 88.67 | 94.54 | 97.60 | 100.00 | 95.67 | 91.17 | 84.66 | 77.79 |

Recruiting turned first. It was at 100.7% one quarter before the company's peak and fell 23% in the two quarters after it, against 9% for the company. SDRs fell 20% over the same two quarters. Marketing fell a little faster than the company. Sales, customer success, support and engineering tracked the company.

### 6.7 Regrowth

Software and IT:

| Function | Eligible episodes | Not cut (%) | Back within 4 quarters (%) | Eligible, trough by 2024 Q1 | Back within 8 quarters (%, 95% interval) | Kaplan-Meier: back by 8 quarters (%) | Kaplan-Meier median (quarters) |
|---|---|---|---|---|---|---|---|
| Recruiting | 1,000 | 7.4 | 6.1 | 259 | 20.8 (15.9-26.3) | 13.1 | not reached |
| Marketing | 2,298 | 13.8 | 13.6 | 559 | 36.3 (32.4-40.3) | 22.2 | not reached |
| Sales | 3,628 | 14.9 | 14.6 | 933 | 41.2 (38.1-44.3) | 26.8 | not reached |
| Customer success | 1,231 | 23.8 | 11.8 | 322 | 31.7 (26.9-36.8) | 22.2 | not reached |
| Support | 1,794 | 20.4 | 11.5 | 411 | 29.2 (25.0-33.9) | 16.9 | not reached |
| Engineering | 4,209 | 12.5 | 13.4 | 1,153 | 38.9 (36.4-41.7) | 22.9 | not reached |
| Sales: SDR/BDR | 952 | 10.5 | 7.6 | 255 | 25.9 (20.7-31.4) | 16.2 | not reached |
| Sales: account executives | 1,417 | 14.0 | 12.9 | 349 | 37.8 (32.7-43.0) | 23.3 | not reached |
| Sales: account managers | 1,006 | 26.5 | 14.8 | 219 | 40.2 (33.4-46.7) | 23.8 | not reached |
| Engineering: software only | 3,897 | 12.4 | 13.1 | 1,048 | 38.2 (35.5-41.3) | 22.0 | not reached |
| Recruiting, each person once (check) | 989 | 7.6 | 5.8 | 257 | 20.6 (15.4-25.6) | 13.9 | not reached |
| HR (rest of the shared label) | 259 | 20.6 | 7.3 | 43 | 18.6 (8.2-29.7) | 15.2 | not reached |
| Product | 1,234 | 18.4 | 13.9 | 274 | 35.8 (30.5-41.1) | 23.9 | not reached |
| Data and analytics | 915 | 14.9 | 13.2 | 209 | 38.8 (32.3-44.8) | 20.5 | not reached |
| Whole company | 5,482 | 0.0 | 7.9 | 1,550 | 36.1 (33.7-38.4) | 20.8 | not reached |

Share back at peak within 8 quarters, by size band and industry group:

| Cell | Recruiting | Marketing | Sales | Customer success | Support | Engineering | Whole company |
|---|---|---|---|---|---|---|---|
| Software and IT, 50-199 | 25.5 | 34.8 | 41.3 | 33.2 | 29.8 | 39.8 | 37.8 |
| Software and IT, 200-999 | 19.8 | 38.3 | 40.6 | 28.3 | 30.9 | 35.8 | 29.1 |
| Software and IT, 1,000-4,999 | 13.3 | 44.8 | 37.9 | 29.2 | 17.9 | 30.0 | 19.4 |
| Software and IT, 5,000+ | suppressed | suppressed | suppressed | suppressed | suppressed | suppressed | suppressed |
| All companies, Software and IT | 20.8 | 36.3 | 41.2 | 31.7 | 29.2 | 38.9 | 36.1 |
| All companies, Other technology, telecom and media | 44.8 | 38.7 | 45.2 | suppressed | 27.4 | 40.0 | 39.9 |
| All companies, Financial services | 22.9 | 38.9 | 43.1 | 34.0 | 31.6 | 33.9 | 41.9 |
| All companies, Business and professional services | 26.9 | 43.9 | 45.5 | 25.5 | 32.3 | 43.3 | 46.9 |
| All companies, Manufacturing, industrial and energy | 26.9 | 38.8 | 47.6 | suppressed | 29.6 | 47.5 | 37.8 |
| All companies, Healthcare and life sciences | suppressed | 44.4 | 47.8 | suppressed | 16.3 | 39.2 | 37.1 |
| All companies, Consumer, public sector and the rest | 34.5 | 51.0 | 46.3 | 35.3 | 36.0 | 46.4 | 47.4 |

All companies:

| Function | Eligible episodes | Not cut (%) | Back within 4 quarters (%) | Eligible, trough by 2024 Q1 | Back within 8 quarters (%, 95% interval) | Kaplan-Meier: back by 8 quarters (%) | Kaplan-Meier median (quarters) |
|---|---|---|---|---|---|---|---|
| Recruiting | 4,197 | 12.4 | 8.8 | 905 | 27.0 (24.0-30.1) | 16.2 | not reached |
| Marketing | 12,816 | 18.7 | 19.5 | 3,458 | 44.5 (42.7-46.0) | 28.8 | not reached |
| Sales | 18,550 | 19.5 | 18.2 | 4,937 | 45.2 (43.8-46.5) | 28.9 | not reached |
| Customer success | 2,006 | 22.7 | 11.9 | 488 | 32.0 (27.9-36.1) | 20.9 | not reached |
| Support | 9,156 | 20.7 | 12.1 | 2,149 | 32.1 (30.0-34.1) | 17.8 | not reached |
| Engineering | 15,269 | 16.0 | 16.1 | 4,249 | 42.8 (41.3-44.3) | 26.1 | not reached |
| Sales: SDR/BDR | 1,513 | 12.6 | 8.9 | 382 | 25.1 (21.1-29.8) | 15.8 | not reached |
| Sales: account executives | 3,636 | 19.1 | 13.8 | 924 | 36.5 (33.4-39.6) | 23.1 | 15 |
| Sales: account managers | 4,354 | 25.4 | 15.1 | 998 | 38.3 (35.1-41.3) | 24.3 | not reached |
| Engineering: software only | 10,249 | 17.8 | 13.7 | 2,627 | 38.8 (37.0-40.6) | 22.3 | not reached |
| Recruiting, each person once (check) | 4,169 | 12.0 | 8.5 | 896 | 26.5 (23.7-29.5) | 15.6 | not reached |
| HR (rest of the shared label) | 2,679 | 23.6 | 11.8 | 530 | 32.5 (28.8-36.7) | 19.3 | not reached |
| Product | 2,815 | 19.6 | 12.3 | 600 | 31.7 (27.9-35.3) | 21.7 | not reached |
| Data and analytics | 3,111 | 18.7 | 15.4 | 669 | 42.0 (38.5-45.5) | 24.7 | not reached |
| Whole company | 51,470 | 0.0 | 13.6 | 16,896 | 43.7 (43.0-44.5) | 25.9 | not reached |

Within 8 quarters of the company's low point, sales was back at its peak headcount in 41.2% of eligible software and IT episodes, engineering in 38.9%, marketing in 36.3%, customer success in 31.7%, support in 29.2% and recruiting in 20.8%; the companies themselves were back in 36.1%. Recruiting is the least often back in every size band with enough data. Kaplan-Meier medians are not reached for any function: fewer than half of the episodes are back at peak by 2026 Q1.

### 6.8 Companies that did not shrink

At the 5,565 software and IT comparison companies, recruiting headcount peaked in 2022 Q3 at 109.0% of its 2022 Q1 level and by 2026 Q1 was 11% below 2022 Q1, while their total headcount grew 22%, engineering and sales 30% each, and support 8%. At the 5,482 software and IT episode companies, by 2026 Q1 recruiting was 47% below its 2022 Q1 level and total headcount 19% below. This is why the net cut index matters: much of recruiting's fall at shrinking companies happened at growing companies too.

### 6.9 Robustness

Net / raw cut index, software and IT:

| Variant (software and IT) | Episodes | Recruiting (net / raw) | Marketing (net / raw) | Sales (net / raw) | Customer success (net / raw) | Support (net / raw) | Engineering (net / raw) | Engineering: software only (net / raw) |
|---|---|---|---|---|---|---|---|---|
| Main | 5,482 | 1.262 / 1.862 | 1.139 / 1.190 | 1.068 / 1.061 | 1.084 / 0.962 | 0.805 / 0.922 | 1.056 / 1.020 | 1.094 / 1.061 |
| (a) Brief's definition, no still-operating rule | 6,638 | 1.217 / 1.750 | 1.139 / 1.196 | 1.067 / 1.068 | 1.097 / 1.004 | 0.812 / 0.913 | 1.049 / 1.019 | 1.081 / 1.054 |
| (b) Falls of 10-50% | 4,828 | 1.296 / 2.075 | 1.121 / 1.180 | 1.074 / 1.066 | 1.082 / 0.940 | 0.823 / 0.973 | 1.050 / 0.997 | 1.089 / 1.040 |
| (d) Each episode weighted equally | 5,482 | 1.125 / 1.435 | 1.129 / 1.204 | 1.083 / 1.107 | 0.960 / 0.775 | 0.886 / 1.012 | 1.020 / 1.012 | 1.048 / 1.047 |
| (f) Peak of 100+ people | 2,662 | 1.314 / 1.981 | 1.118 / 1.168 | 1.061 / 1.048 | 1.077 / 0.969 | 0.795 / 0.921 | 1.060 / 1.015 | 1.098 / 1.054 |
| (h) Troughs by 2024 Q1 (added after the first run) | 1,550 | 1.664 / 2.524 | 1.167 / 1.232 | 1.017 / 0.990 | 1.215 / 1.090 | 0.906 / 0.963 | 0.958 / 0.904 | 1.005 / 0.954 |

Size band from the company record (net, software and IT):

| Record size band (software and IT) | Episodes | Recruiting | Marketing | Sales | Customer success | Support | Engineering |
|---|---|---|---|---|---|---|---|
| under 50 | 350 | 1.440 | 1.244 | 1.041 | 1.233 | 0.706 | 1.266 |
| 50-199 | 2,676 | 1.253 | 1.173 | 1.123 | 1.060 | 0.845 | 0.999 |
| 200-999 | 1,720 | 1.208 | 1.111 | 1.106 | 0.982 | 0.838 | 1.003 |
| 1,000-4,999 | 517 | 1.130 | 0.964 | 1.052 | 1.004 | 0.883 | 1.062 |
| 5,000+ | 219 | 1.408 | 1.084 | 0.939 | 1.028 | 0.731 | 1.131 |

Regrowth variants:

| Variant (software and IT, back within 8 quarters, %) | Recruiting | Marketing | Sales | Customer success | Support | Engineering | Whole company |
|---|---|---|---|---|---|---|---|
| Main (censored at 2026 Q1, 5+ people at peak) | 20.8 | 36.3 | 41.2 | 31.7 | 29.2 | 38.9 | 36.1 |
| (g) censored at 2025 Q3 | 20.8 | 46.4 | 52.0 | 39.7 | 34.8 | 47.9 | 46.9 |
| (g) 10+ people at the peak | 18.4 | 35.1 | 38.5 | 29.0 | 26.8 | 38.0 | 36.1 |

Recruiting has the highest net cut index of the six in five of the six variants; with every episode weighted equally, marketing (1.129) edges it (1.125). Support has the lowest net index in all six. Recruiting's net index runs from 1.125 to 1.664 across the variants, and is highest for episodes that bottomed out by 2024 Q1, mostly the 2022-2023 wave of cuts. In those early episodes engineering was cut a little less than its size (0.958), so the slightly-above-1 engineering index of the main sample comes from falls that ended in 2024. With size taken from the company record, recruiting is highest in every band. Censoring regrowth at 2025 Q3 or requiring 10 people at the peak keeps recruiting the least often back.

## 7. Hypotheses

The hypotheses were pre-registered in the study plan before any data was read. They are scored on the pre-registered net index as written, and on the headline net index for comparison.

| Hypothesis | Scored on | Verdict | Detail |
|---|---|---|---|
| 1 | headline net index | wrong | highest raw: recruiting (1.862); highest net: recruiting (1.262); recruiting net 1.262 |
| 1 | pre-registered net index | wrong | highest raw: recruiting (1.862); highest net: customer_success (1.149); recruiting net 1.016 |
| 2 | headline net index | right | recruiting net 1.262 vs raw 1.862 |
| 2 | pre-registered net index | right | recruiting net 1.016 vs raw 1.862 |
| 3 | headline net index | right | marketing net 1.139 |
| 3 | pre-registered net index | right | marketing net 1.146 |
| 4 | headline net index | right | outside 0.8-1.25: none (sales 1.068, customer_success 1.084, support 0.805) |
| 4 | pre-registered net index | wrong | outside 0.8-1.25: support 0.750 (sales 1.084, customer_success 1.149, support 0.750) |
| 5 | headline net index | wrong | lowest: support (0.805); engineering 1.056 |
| 5 | pre-registered net index | wrong | lowest: support (0.750); engineering 1.081 |
| 6 | regrowth (share back within 8 quarters) | right | order: sales 41.2, engineering 38.9, marketing 36.3, customer_success 31.7, support 29.2, recruiting 20.8 |
| 7 | headline net index | wrong | size bands failing: 50-199 (top marketing, bottom support); 200-999 (top recruiting, bottom support); 1,000-4,999 (top recruiting, bottom support); 5,000+ (top recruiting, bottom support) / industry groups failing: tech_media (top recruiting, bottom support); financial (top recruiting, bottom support); business_services (top customer_success, bottom support); industrial (top customer_success, bottom support); life_sciences (top customer_success, bottom support) |
| 7 | pre-registered net index | wrong | size bands failing: 50-199 (top marketing, bottom support); 200-999 (top recruiting, bottom support); 1,000-4,999 (top engineering, bottom support); 5,000+ (top engineering, bottom support) / industry groups failing: tech_media (top customer_success, bottom support); financial (top customer_success, bottom support); business_services (top customer_success, bottom support); industrial (top customer_success, bottom support); life_sciences (top customer_success, bottom support) |

In short: recruiting was cut hardest of the six and recovered last, as expected, but by less than expected once normal churn is netted out (1.262 on the headline measure, 1.016 on the pre-registered one, against the 1.5 we expected). Marketing was cut harder than its size, as expected. Engineering was not protected; support was the least-cut function. Outside software, customer success, not recruiting, often had the highest index.

## 8. Verification

An independent re-analysis, written without seeing any of this study's code or outputs, recomputed the main numbers from scratch. It was given only the study plan (question, brief, hypothesis, plan and logged decisions), the shared headcount history, and the industry-group map as definitions only (label and group, no counts). It wrote and ran its own code, with a synthetic test of its index code, and reported 134 numbers. Its map holds 17 software and IT labels rather than the 18 of the SDR study, because one of the 18 has no company in the panel; no panel company was left unmapped. Its numbers are compared with the study's at the precision the re-analysis reported, both for the first run and for the current one. Every number is listed in the appendix at the end of this document.

| Block | Numbers | Agree | Within bootstrap noise | Not computed by the study | Differ | Differed in the first run |
|---|---|---|---|---|---|---|
| Sample counts | 8 | 8 | 0 | 0 | 0 | 0 |
| Software and IT, by function | 69 | 65 | 4 | 0 | 0 | 8 |
| Software and IT, by size band | 12 | 12 | 0 | 0 | 0 | 3 |
| All companies, by function | 6 | 6 | 0 | 0 | 0 | 2 |
| Regrowth, software and IT | 18 | 18 | 0 | 0 | 0 | 4 |
| Timing, software and IT | 4 | 4 | 0 | 0 | 0 | 1 |
| Check rows | 12 | 6 | 0 | 6 | 0 | 0 |
| Other checks | 5 | 4 | 0 | 1 | 0 | 0 |
| All | 134 | 123 | 4 | 7 | 0 | 18 |

**First comparison.** Every sample count, every number for marketing, sales, customer success, support, engineering, the sales subteams and the whole company, the matching checks and the timing totals agreed. The 18 numbers that differed were all recruiting numbers. In software and IT the re-analysis had recruiters at 2.19% of headcount at the peak and 4.07% of the people lost (first run: 2.16% and 4.05%), a raw index of 1.862 against 1.876, a net index of 1.262 against 1.265, and 36,551 recruiters at the peak against 36,060. Recruiting's net index in three of the four size bands, its raw and net indices across all companies, its regrowth counts and shares, and its timing differed by similarly small amounts.

**Cause.** The study plan defines each function as a group of the shared title labels "summed over levels", and the shared build counts a person once per function and title level. The study's recruiter pass instead counted each recruiter once per company and quarter, so a person holding recruiter titles at two levels at once (a recruiter who is also head of talent, say) counted once in recruiting, where the study plan, like every other function, counts them twice. The re-analysis followed the study plan. Its own each-person-once variant gave 1.876, 1.265 and 1.012 (raw, net and pre-registered) in software and IT: the first run's recruiting numbers exactly.

**Fix.** Recruiting and the HR remainder are now counted the study plan's way, once per person, company, quarter and title level, and the each-person-once count is kept as the check row "Recruiting, each person once" in the tables. The recruiter pass and every analysis step after it were re-run. Recruiters now add up to 8,954,550 person-quarters, against 8,799,994 when each counts once. After the fix all 123 comparable point numbers agree, and the 4 bootstrap interval ends are within sampling noise of each other (recruiting 1.108 to 1.416 here, 1.133 to 1.407 for the re-analysis), as expected from different random draws. The 7 numbers the study does not compute are extra check rows from the re-analysis, for recruiting and HR together, and its smallest comparison cell (52 companies).

**What changed in the report.** Recruiting's quoted numbers moved in the last digit: the headline net index from 1.27 to 1.26, the raw index from 1.88 to 1.86 (1.9 at one decimal either way), the fall at comparison companies from 10% to 9%, the 5,000+ band from 1.48 to 1.46, and recruiting at comparison companies by 2026 Q1 from 12% to 11% below early 2022. No ranking, hypothesis score or conclusion changed: recruiting is still the most-cut of the six functions in software and IT and the least often back at its peak.

**Rounding.** The re-run also fixed a smaller issue found while comparing: the outputs stored indices at 3 decimals and the text rounded them again to 2, so support's net index (0.804927) was printed as 0.81. The outputs now store 6 decimals and every number shown is rounded once, half-up. This moved a few quoted numbers by one in the last digit, among them support's net index (0.81 to 0.80) and customer success in manufacturing, industrial and energy (1.67 to 1.66).

## 9. Alternative explanations and limitations

- **Resignations, not layoffs.** Rebuilt headcount shows people leaving and not being replaced; it cannot tell a layoff from a resignation or a hiring freeze. The comparison companies remove the churn that happens anyway. What is left is linked with the contraction, but part of a function's excess loss may be people who left and were not replaced rather than people who were laid off. For recruiting that distinction matters little: with hiring frozen, recruiters are not replaced either way.
- **Normal drift is not the same everywhere.** The comparison companies did not shrink, and companies that grow may hire recruiters and SDRs for reasons that do not apply to shrinking companies. The matching (same industry group, size band and quarters) narrows the gap but cannot close it. That is the main reason to read the raw and net indices together: they bound recruiting between 1.262 and 1.862 in software and IT.
- **Six countries.** Ken internal data holds people who live in the US, UK, Canada, Australia, Ireland and New Zealand. A company that moved engineering or support jobs to other countries during a contraction shows a cut in these six countries. Engineering is often offshored, so the engineering index may overstate cuts to the engineering function as a whole.
- **Profiles are self-reported and late.** People add jobs late, leave short jobs off and delete profiles, and only active profiles exist. Late additions make the most recent quarters undercount hiring, so the regrowth shares are likely lower bounds, and some falls ending in 2024 Q4 may be partly artificial (check h). Deleted profiles remove people from every quarter, not only from the trough.
- **Titles define functions.** A recruiter who moved into an HR or operations title inside the company counts as a recruiting loss and a gain elsewhere. The HR remainder was cut about in line with the companies (raw 1.142, net 0.938 in software and IT), so a large shift of recruiters into HR titles is not visible. Some people edit an old title instead of adding a new role when promoted, which the data cannot show.
- **Double counting.** A person holding two titles at once counts in both functions (3.5% of person-function counts). This changes shares at the peak and at the trough alike.
- **Shutdowns and acquisitions.** The still-operating rule removes most of them (1,156 software and IT companies with a fall had no joiner afterwards). Check (a) adds them back; recruiting's net index falls to 1.217 and the order is unchanged.
- **Industry labels.** Industries come from the current company record. A company that changed its label, or a label that mixes B2B and consumer companies, puts some companies in the wrong group. The software and IT definition is the list of 18 labels from the SDR study.
- **Small cells.** The 5,000+ band holds 35 software and IT episodes, and its cut indices are dominated by a few very large companies.
- **Correlation, not causation.** The study describes which functions shrank when companies shrank, matched with companies that did not. It does not show that cutting a function caused any outcome.

## 10. Sources

- Ken internal data: Ken's dataset of public professional profiles, work histories and company records, people in the US, UK, Canada, Australia, Ireland and New Zealand; profiles updated up to 2026-04-29.
- Company headcount history, rebuilt from Ken internal data for the SDR study.
- Title taxonomy: the shared sales title taxonomy, built for the GTM engineer study.
- Software and IT industry labels and the reporting-lag measurement: the SDR study.

## 11. Appendix: the re-analysis numbers

Every number the independent re-analysis reported, with the study's value rounded to the same decimals. A value agrees when the two differ by no more than rounding; bootstrap interval ends agree when within 0.03. "Not computed by the study" marks check rows only the re-analysis computed.

| Number | Study | Re-analysis | Match |
|---|---|---|---|
| Software and IT: companies with 50+ people | 12203 | 12203 | yes |
| Software and IT: falls of 10% or more from a 50+ peak | 6638 | 6638 | yes |
| Software and IT: contraction episodes (still operating) | 5482 | 5482 | yes |
| Software and IT: comparison companies | 5565 | 5565 | yes |
| All companies: companies with 50+ people | 151784 | 151784 | yes |
| All companies: falls of 10% or more from a 50+ peak | 58223 | 58223 | yes |
| All companies: contraction episodes (still operating) | 51470 | 51470 | yes |
| All companies: comparison companies | 93561 | 93561 | yes |
| Software and IT, recruiting: share of headcount at the peak (%) | 2.19 | 2.19 | yes |
| Software and IT, recruiting: share of the people lost (%) | 4.07 | 4.07 | yes |
| Software and IT, recruiting: change, peak to trough (%) | -48.84 | -48.84 | yes |
| Software and IT, recruiting: change at comparison companies (%) | -9.23 | -9.23 | yes |
| Software and IT, recruiting: raw cut index | 1.862 | 1.862 | yes |
| Software and IT, recruiting: net cut index | 1.262 | 1.262 | yes |
| Software and IT, recruiting: pre-registered net index | 1.016 | 1.016 | yes |
| Software and IT, recruiting: people at the peak | 36551 | 36551 | yes |
| Software and IT, marketing: share of headcount at the peak (%) | 4.25 | 4.25 | yes |
| Software and IT, marketing: share of the people lost (%) | 5.06 | 5.06 | yes |
| Software and IT, marketing: change, peak to trough (%) | -31.22 | -31.22 | yes |
| Software and IT, marketing: change at comparison companies (%) | 13.46 | 13.46 | yes |
| Software and IT, marketing: raw cut index | 1.190 | 1.190 | yes |
| Software and IT, marketing: net cut index | 1.139 | 1.139 | yes |
| Software and IT, marketing: pre-registered net index | 1.146 | 1.146 | yes |
| Software and IT, marketing: people at the peak | 70949 | 70949 | yes |
| Software and IT, sales: share of headcount at the peak (%) | 15.15 | 15.15 | yes |
| Software and IT, sales: share of the people lost (%) | 16.08 | 16.08 | yes |
| Software and IT, sales: change, peak to trough (%) | -27.84 | -27.84 | yes |
| Software and IT, sales: change at comparison companies (%) | 14.45 | 14.45 | yes |
| Software and IT, sales: raw cut index | 1.061 | 1.061 | yes |
| Software and IT, sales: net cut index | 1.068 | 1.068 | yes |
| Software and IT, sales: pre-registered net index | 1.084 | 1.084 | yes |
| Software and IT, sales: people at the peak | 253087 | 253087 | yes |
| Software and IT, customer success: share of headcount at the peak (%) | 2.12 | 2.12 | yes |
| Software and IT, customer success: share of the people lost (%) | 2.04 | 2.04 | yes |
| Software and IT, customer success: change, peak to trough (%) | -25.24 | -25.24 | yes |
| Software and IT, customer success: change at comparison companies (%) | 19.59 | 19.59 | yes |
| Software and IT, customer success: raw cut index | 0.962 | 0.962 | yes |
| Software and IT, customer success: net cut index | 1.084 | 1.084 | yes |
| Software and IT, customer success: pre-registered net index | 1.149 | 1.149 | yes |
| Software and IT, customer success: people at the peak | 35411 | 35411 | yes |
| Software and IT, support: share of headcount at the peak (%) | 4.64 | 4.64 | yes |
| Software and IT, support: share of the people lost (%) | 4.28 | 4.28 | yes |
| Software and IT, support: change, peak to trough (%) | -24.18 | -24.18 | yes |
| Software and IT, support: change at comparison companies (%) | 5.08 | 5.08 | yes |
| Software and IT, support: raw cut index | 0.922 | 0.922 | yes |
| Software and IT, support: net cut index | 0.805 | 0.805 | yes |
| Software and IT, support: pre-registered net index | 0.750 | 0.750 | yes |
| Software and IT, support: people at the peak | 77554 | 77554 | yes |
| Software and IT, engineering: share of headcount at the peak (%) | 21.01 | 21.01 | yes |
| Software and IT, engineering: share of the people lost (%) | 21.43 | 21.43 | yes |
| Software and IT, engineering: change, peak to trough (%) | -26.76 | -26.76 | yes |
| Software and IT, engineering: change at comparison companies (%) | 15.39 | 15.39 | yes |
| Software and IT, engineering: raw cut index | 1.020 | 1.020 | yes |
| Software and IT, engineering: net cut index | 1.056 | 1.056 | yes |
| Software and IT, engineering: pre-registered net index | 1.081 | 1.081 | yes |
| Software and IT, engineering: people at the peak | 350904 | 350904 | yes |
| Software and IT, SDR/BDR: share of headcount at the peak (%) | 1.43 | 1.43 | yes |
| Software and IT, SDR/BDR: share of the people lost (%) | 2.50 | 2.50 | yes |
| Software and IT, SDR/BDR: change, peak to trough (%) | -45.98 | -45.98 | yes |
| Software and IT, SDR/BDR: change at comparison companies (%) | 24.04 | 24.04 | yes |
| Software and IT, SDR/BDR: raw cut index | 1.753 | 1.753 | yes |
| Software and IT, SDR/BDR: net cut index | 1.632 | 1.632 | yes |
| Software and IT, SDR/BDR: pre-registered net index | 1.795 | 1.795 | yes |
| Software and IT, SDR/BDR: people at the peak | 23817 | 23817 | yes |
| Software and IT, whole company: share of headcount at the peak (%) | 100.00 | 100.00 | yes |
| Software and IT, whole company: share of the people lost (%) | 100.00 | 100.00 | yes |
| Software and IT, whole company: change, peak to trough (%) | -26.23 | -26.23 | yes |
| Software and IT, whole company: change at comparison companies (%) | 12.77 | 12.77 | yes |
| Software and IT, whole company: raw cut index | 1.000 | 1.000 | yes |
| Software and IT, whole company: net cut index | 1.000 | 1.000 | yes |
| Software and IT, whole company: pre-registered net index | 1.000 | 1.000 | yes |
| Software and IT, whole company: people at the peak | 1670162 | 1670162 | yes |
| Software and IT: episodes | 5482 | 5482 | yes |
| Software and IT, recruiting: net cut index, 95% interval, low end | 1.108 | 1.133 | within bootstrap noise |
| Software and IT, recruiting: net cut index, 95% interval, high end | 1.416 | 1.407 | within bootstrap noise |
| Software and IT, support: net cut index, 95% interval, low end | 0.719 | 0.710 | within bootstrap noise |
| Software and IT, support: net cut index, 95% interval, high end | 0.902 | 0.894 | within bootstrap noise |
| Software and IT, 50-199 people: episodes | 4174 | 4174 | yes |
| Software and IT, 50-199 people, recruiting: net cut index | 1.022 | 1.022 | yes |
| Software and IT, 50-199 people, support: net cut index | 0.848 | 0.848 | yes |
| Software and IT, 200-999 people: episodes | 1097 | 1097 | yes |
| Software and IT, 200-999 people, recruiting: net cut index | 1.298 | 1.298 | yes |
| Software and IT, 200-999 people, support: net cut index | 0.833 | 0.833 | yes |
| Software and IT, 1,000-4,999 people: episodes | 176 | 176 | yes |
| Software and IT, 1,000-4,999 people, recruiting: net cut index | 1.287 | 1.287 | yes |
| Software and IT, 1,000-4,999 people, support: net cut index | 0.815 | 0.815 | yes |
| Software and IT, 5,000+ people: episodes | 35 | 35 | yes |
| Software and IT, 5,000+ people, recruiting: net cut index | 1.463 | 1.463 | yes |
| Software and IT, 5,000+ people, support: net cut index | 0.620 | 0.620 | yes |
| All companies, recruiting: raw cut index | 1.702 | 1.702 | yes |
| All companies, recruiting: net cut index | 1.352 | 1.352 | yes |
| All companies, support: raw cut index | 1.102 | 1.102 | yes |
| All companies, support: net cut index | 0.954 | 0.954 | yes |
| All companies, SDR/BDR: raw cut index | 1.874 | 1.874 | yes |
| All companies, SDR/BDR: net cut index | 1.825 | 1.825 | yes |
| Regrowth, recruiting: episodes eligible, trough by 2024 Q1 | 259 | 259 | yes |
| Regrowth, recruiting: back at peak within 8 quarters (%) | 20.85 | 20.85 | yes |
| Regrowth, marketing: episodes eligible, trough by 2024 Q1 | 559 | 559 | yes |
| Regrowth, marketing: back at peak within 8 quarters (%) | 36.31 | 36.31 | yes |
| Regrowth, sales: episodes eligible, trough by 2024 Q1 | 933 | 933 | yes |
| Regrowth, sales: back at peak within 8 quarters (%) | 41.16 | 41.16 | yes |
| Regrowth, customer success: episodes eligible, trough by 2024 Q1 | 322 | 322 | yes |
| Regrowth, customer success: back at peak within 8 quarters (%) | 31.68 | 31.68 | yes |
| Regrowth, support: episodes eligible, trough by 2024 Q1 | 411 | 411 | yes |
| Regrowth, support: back at peak within 8 quarters (%) | 29.20 | 29.20 | yes |
| Regrowth, engineering: episodes eligible, trough by 2024 Q1 | 1153 | 1153 | yes |
| Regrowth, engineering: back at peak within 8 quarters (%) | 38.94 | 38.94 | yes |
| Regrowth, whole company: episodes eligible, trough by 2024 Q1 | 1550 | 1550 | yes |
| Regrowth, whole company: back at peak within 8 quarters (%) | 36.06 | 36.06 | yes |
| Regrowth, recruiting: episodes eligible, any trough | 1000 | 1000 | yes |
| Regrowth, recruiting: back at peak within 4 quarters (%) | 6.10 | 6.10 | yes |
| Regrowth, whole company: episodes eligible, any trough | 5482 | 5482 | yes |
| Regrowth, whole company: back at peak within 4 quarters (%) | 7.94 | 7.94 | yes |
| Timing, recruiting: % of peak level 2 quarters after the peak | 77.25 | 77.25 | yes |
| Timing, recruiting: episodes | 5261 | 5261 | yes |
| Timing, whole company: % of peak level 2 quarters after the peak | 91.17 | 91.17 | yes |
| Timing, whole company: episodes | 5261 | 5261 | yes |
| Software and IT, recruiting, each person once: raw cut index | 1.876 | 1.876 | yes |
| Software and IT, recruiting, each person once: net cut index | 1.265 | 1.265 | yes |
| Software and IT, recruiting, each person once: pre-registered net index | 1.012 | 1.012 | yes |
| All companies, recruiting, each person once: raw cut index | 1.715 | 1.715 | yes |
| All companies, recruiting, each person once: net cut index | 1.355 | 1.355 | yes |
| All companies, recruiting, each person once: pre-registered net index | 1.267 | 1.267 | yes |
| Software and IT, recruiting and HR together: raw cut index | not computed | 1.721 | not computed by the study |
| Software and IT, recruiting and HR together: net cut index | not computed | 1.216 | not computed by the study |
| Software and IT, recruiting and HR together: pre-registered net index | not computed | 1.021 | not computed by the study |
| All companies, recruiting and HR together: raw cut index | not computed | 1.433 | not computed by the study |
| All companies, recruiting and HR together: net cut index | not computed | 1.209 | not computed by the study |
| All companies, recruiting and HR together: pre-registered net index | not computed | 1.161 | not computed by the study |
| Recruiter split: company-quarter cells that differ from the shared build | 0 | 0 | yes |
| Software and IT: episodes matched on size band only | 0 | 0 | yes |
| All companies: episodes matched on size band only | 1 | 1 | yes |
| All companies: episodes whose cell has no SDR/BDR at the peak | 6 | 6 | yes |
| Software and IT: smallest comparison cell (companies) | not computed | 52.0 | not computed by the study |