Original research · Human in the Loop Talent
Which roles precede the fastest growth at startups?
Not measured yet. This study will show which hire, from a first sales leader to a Head of AI, comes just before a startup's fastest headcount growth, compared with companies of the same size that did not make it. Until the first results are in, the category studies show what similar companies hired first.
What similar companies hired first: Seed-stage AI companies, Valor Ventures portfolio, Vista Equity Partners portfolio.
Who this is for. Founders deciding which seat to fill first, and the investors and talent leaders advising them. Lens. Hiring as a leading indicator: the seat a company opens just before it grows says what growth required.
What this study will show
For each type of hire, how much a company's headcount grew in the six months after it first opened that role, compared with companies of the same size that did not. It also shows how often those companies raised money in the same six months, and how long the role took to fill.
The roles tracked: Sales leader, Founding account executive, Account executive, SDR / BDR, Revenue operations, GTM engineering, Marketing leader, Demand generation / growth marketing, Product marketing, Customer success, Talent / recruiting, People operations / HR, Product management, Engineering leader, Software engineer, AI leader, AI / ML engineer, Data / analytics, Finance, Operations, Chief of staff.
Why it matters
Founders are told which seat to fill first by people whose evidence is their own experience. This study shows whether the companies that grew fastest had, in fact, just opened a particular seat, compared with companies of the same size that had not. Even as a correlation, that is a better guide than an anecdote.
What it will not show is that making the hire causes the growth: companies also hire because they expect to grow.
What to do next
Founder or CEO
- Until the first results are in, use the category studies: they show what similar companies hired first.
- A completed Revenue per Employee Planner session is one way to have your company included.
Investor
- Read any result here as an association: companies also hire because they expect to grow.
- The portfolio studies show which seats your companies opened this year, the observable half of this question.
Talent leader or recruiter
- The roles are the same ones the category studies use, so a role that precedes growth here can be looked up as current demand there.
Common questions
What does this research measure?
Which hire a startup made just before it grew fastest. For each type of role, from a founding account executive or a first sales leader to a Head of AI, it compares a company's headcount growth in the six months after opening that role with companies of the same size that did not. It also notes whether the company raised money in those six months and how long the role took to fill.
Does opening the role cause the growth?
No. The study shows what the fastest-growing companies had just hired, compared with companies of the same size. Companies open roles because they expect to grow, and some grow because of the hire; the data cannot separate the two. It is a guide to which seat to fill first, not a proof.
Methodology and data
Original research computed from the signal layer of the recruiting CRM behind Human in the Loop Talent: public job boards, company news, funding announcements and headcount readings, collected continuously for every tracked company and recomputed every Sunday. A figure is published only when the sample supports it; the data behind every page is published as JSON next to it.
Study design, fixed in advance
The design is stated before any result so that a finding cannot be fitted to the data after the fact. Each element below is also encoded in the computation (crm-architecture/src/research/growthRoles.ts) and repeated in the dataset's methodology field.
Are particular hiring sequences associated with faster subsequent headcount growth at Seed to Series B technology companies?
A company's first opening in a role family (21 families, classified from the title with the CRM's size-aware seat taxonomy), dated by the day the posting was first seen.
Percentage change in the company's recorded headcount (its LinkedIn employee count, read each time the company is enriched) from the reading at or before the opening to the first reading at least 180 days later.
180 days after the first opening. A company enters the measured set only when a reading exists on both sides of the window; openings whose window has not yet closed are counted as "opened", not "measured".
Companies in the same headcount band (1–19, 20–49, 50–99, 100–199, 200–499, 500–999, 1,000–2,000, 2,001–10,000, 10,001+) over the same horizon, measured from their first reading. Lift is the difference in percentage points between the medians.
A role family is reported only when at least 30 companies have both an opening and a measurable window. Below that the table shows the count so far and no figure. Medians, not means, so one hyper-growth company cannot move a row.
Whether the company raised funding inside the window (any funding signal from news, LinkedIn posts or ColdIQ) and how long the posting stayed open (first seen to disappearance from the board).
Causation. Companies open roles because they expect to grow, and some grow because of the hire; the design cannot separate the two. Nor can it see companies without a public job board or a LinkedIn page, hires made without a posting, or headcount changes LinkedIn does not reflect. No predictive validity is claimed until the panel has at least two full windows of longitudinal readings and the out-of-sample check described under Methodology has been run.
How the panel is built
- Companies enter the panel when the CRM starts tracking them: a hiring signal, a funding announcement, a leadership change, a completed planner session, or a category census.
- Every weekday the CRM reads the public job boards (Greenhouse, Lever, Ashby, Workable) and LinkedIn jobs for each company and classifies each opening into a role family (sales leader, RevOps, GTM engineering, founding AE, Head of AI and so on).
- A headcount reading is taken from the company’s LinkedIn page each time the company is enriched; funding and leadership signals come from company news and press.
- The report is recomputed every Sunday in code from the panel; no number is written by hand or by a language model.
- A role family is reported only once 30 companies have both an opening and a measurable 180-day headcount window; below that the row shows the count so far and no claim.
- Results describe correlation: what fast-growing companies had just hired, against a same-size baseline. They do not claim the hire caused the growth.
- Panel of B2B technology companies tracked by the GTM Architecture CRM: public job boards (Greenhouse, Lever, Ashby, Workable) and LinkedIn jobs are read every weekday; headcount is recorded from the company's LinkedIn page each time the company is enriched.
- An observation is the first time a company opens a role in a family (21 families, classified from the title with the same size-aware seat taxonomy the CRM uses). Growth is the change in recorded headcount from the reading at or before the opening to the first reading at least 180 days later.
- The baseline for each observation is the median growth of companies in the same size band (1–19, 20–49, 50–99, 100–199, 200–499, 500–999, 1000–2000, 2001–10000, 10001+) over the same horizon, measured from their first reading. Lift is the difference in percentage points.
- Funding within the horizon counts any funding_or_financing signal (news, LinkedIn posts, ColdIQ) observed in the 180 days after the opening. Days to fill is the time from first seen to the posting disappearing from the board.
- A family is reported only when at least 30 companies have both an opening and a measurable growth window; below that the row shows the count and no claim. Figures are medians, not means, so a single hyper-growth company cannot move them.
- Planning-canvas figures come from founders who completed the Human in the Loop planning canvas (self-reported team, stage, bottlenecks); they describe intent, not outcome, and are shown separately.
What would confirm or disprove a finding
- Confirm an association. A role family reaching 30 measured companies with a median lift against its size-band baseline that persists across two consecutive weekly recomputations and across at least two size bands.
- Disprove it. The lift shrinking toward zero as the measured set grows, reversing between size bands, or disappearing when companies that raised funding inside the window are held out.
- Before any claim of predictive validity. A held-out check: families ranked on the first year of the panel, then scored on openings observed in the following year. Until that has been run and published here, every figure is descriptive.
Sample so far
Research in progress; panel computed 2026-10-10. 0 companies tracked, 0 with two or more headcount readings, 0 openings classified, 0 funding events, 0 planner sessions. 0 of 21 role families have reached the 30-company threshold.
Limitations
- Headcount is read from LinkedIn, which lags real headcount and is self-reported by employees; a 180-day change of a few people at a small company is within its noise.
- The panel sees companies with a public job board or LinkedIn jobs; companies that hire without posting, or through agencies, are under-represented.
- Companies enter the panel when the CRM starts tracking them, which is itself triggered by a signal; the panel is not a random sample of the stage.
- Selection and reverse causality are not separable in this design; a lift is an association.
- As of 2026-10-10 the research store holds no company snapshots, classified openings or funding events: the weekly refresh has not yet delivered its first readings. That is a pipeline status, not a result.
Questions about the method
Where does the data come from?
From the signal layer of the recruiting CRM behind Human in the Loop Talent: public job boards (Greenhouse, Lever, Ashby, Workable) and LinkedIn jobs read every weekday, and headcount readings taken from the company’s LinkedIn page each time the company is enriched. Planner figures come from founders who completed the Revenue per Employee Planner and are shown separately as stated intent.
Why are there no figures yet?
A role family is reported only once at least 30 companies have both an opening and a measurable 180-day window. Below that the study publishes no figure: a number founders will cite has to survive the question “how many companies?”.
How often is it updated, and how do I cite it?
Every Sunday the panel is recomputed from the CRM and this page is rebuilt from the result. Cite it as Human in the Loop Talent, “Roles that precede growth”, with the computed date shown under Methodology and data and the URL https://humaninthelooptalent.com/research/roles-that-precede-growth/. Reuse is permitted with attribution.
Dataset and versions
- The report: /research/roles-that-precede-growth.json (version research@1.0.0, computed 2026-10-10, status collecting).
- Revision log: 1 revision since 2026-10-10, read from the git history of the published JSON.
- Licence: CC BY 4.0. Reuse with attribution.
Cite this
Related reading: Seed-stage AI companies, Valor Ventures portfolio, Vista Equity Partners portfolio, post-funding hiring, workforce architecture, startup recruiting.
Deciding which seat to fill first?
The free Revenue per Employee Planner works through whether the constraint needs a hire at all, and which one. Or compare notes on how similar-stage companies sequenced theirs.