Buyer personas and buying committees
Your ICP says which companies can buy. A buyer persona says who inside one of them you write to, and what changes in the message because of it. This is the person layer, built short enough to use and specific enough to be proven wrong.
By Kshitij Maheshwari, co-founder · Updated August 2026 · 14 min read
Written by operators who build these documents for live campaigns, not by a vendor selling persona software.
What a buyer persona is, and what it is for
The company layer decides whether an account can buy. The person layer decides what you say once it can, and it is the layer most outbound teams walk straight past.
A buyer persona is a short description of one seat at a purchase: the problem that seat owns, the words they use for it, and what changes in your message because of it.
The company layer is settled elsewhere: our ideal customer profile guide hands the person layer to this page by name. A persona has to earn a change in three things:
Which seat gets the first message, and which ones you map but leave alone.
The problem in their words, not yours. Same product, four different first lines.
A user and a budget holder cannot say yes to the same thing, so the ask moves too.
A persona is a seat at a purchase, not a person at a desk. If the document does not change a single word you send, you have written a biography.
The marketing persona and the outbound persona are different documents
Both are called a buyer persona. The difference is not how much detail each one holds. It is whether any line in it changes what you write.
| Dimension | The marketing persona | The outbound persona |
|---|---|---|
| Length | Forty fields, a name, a photo, a hobby. | Six lines, each one changing a sentence. |
| Built from | A template, a category report, a workshop. | Replies, call recordings, job posts. |
| The unit | A person, with an age and a commute. | A seat at a purchase. |
| What it changes | Brand tone and content themes. | The first line, the ask, the seat. |
| How you check it | You cannot. Nothing in it can be wrong. | One prediction, checked against replies. |
| Shelf life | Reviewed annually, edited cosmetically. | Rewritten when the organization changes shape. |
Why a long persona describes nobody
Personas get longer because length feels like rigor. It runs the other way, and you can do the arithmetic in your head.
Every attribute you add shrinks who the persona fits. Past about nine of them, it fits nobody at all.
That is measured, not a hunch. In a 2008 study that ran persona profiles against six real customer datasets, the more attributes a profile carried the fewer real people matched it, and past about nine it matched nobody.
So the useful edit is subtraction, never another line. Take one line out and see whether the email changes: that is the subtraction test below, and it is the only edit that makes a persona more accurate rather than longer.
Where the persona came from, and what B2B dropped
The persona came out of Alan Cooper's 1999 book The Inmates Are Running the Asylum, and it was built for software design, where one person uses the product. B2B borrowed the document and left the design purpose behind: in a purchase, nobody decides alone.
The buying committee is a set of seats, not a list of titles
The same seat is called five different things across five companies. Name the seat first and the titles second: the title is the lookup key, the seat is the reason.
Whose week is worse because of the problem. They answer most often and sign for nothing, which is why you open with them.
Where the budget for this line item sits, often one desk away from the title that sounds like it.
Whoever has to make the thing work after the signature. They care about their own Tuesday, not your category.
Security, finance, legal, or whoever owns the system you plug into. You map this seat so it cannot surprise you in week six.
The champion is not a fifth seat. It is a relationship somebody in one of these four decides to have with you. The two roles people confuse most get worked in full in champion vs economic buyer.
How big is a buying committee, really
Committee pages quote whichever number is largest. The published counts differ because they count different things, and the most useful answer is not a count.
6 to 10 people touch the decision
Gartner's 2019 buying-journey research puts 6 to 10 people in a complex buying group, each arriving with independently gathered information. That is everybody with a hand on it, not everybody you reach.
Three to five you can actually work
Our own multithreading play publishes the working number: most committees run three to five people who matter. That is a capacity choice, not a measurement.
Two counts, and a sentence has to say which it means. The group is everyone who touches the decision. The seats are the ones you can thread and still read. Six to ten touch it, you work three to five, and a first buy in a new category fills more chairs than a renewal.
What changes per seat, and what must not
One story, told from four vantage points. Most of the resistance in a deal is the seats disagreeing with each other rather than with you, so the job is agreement, not attention.
| The seat | The angle that lands | The common mistake |
|---|---|---|
| Feels it | Their week, described accurately enough to sting. | Return on investment, to somebody with no budget. |
| Funds it | The cost of carrying this another quarter. | Forwarding the workflow detail meant for their team. |
| Runs it | What their day looks like once this exists. | Selling strategy to whoever has to configure it. |
| Can veto it | Their risk, answered before they raise it. | Cold emailing them first, which spends the account. |
What must not change is the story. Four unrelated pitches inside one account read as a vendor working an org, and the seats compare notes. The personalization race aims at the wrong unit for the same reason: writing harder at one person does nothing for the room they still have to convince. The order to run them in is in multithreading outbound.
How to build a persona at seed, from evidence you already have
No template, no tool, no purchase. Every input below is already in your inbox, your call recordings, or on a careers page anyone can read.
-
1
Read a quarter of your replies
Which titles answered, which never did, and the words they used for the problem. Three real replies beat a forty-field template, because the replies are in their language.
-
2
Pull twenty job posts for the seat
A posting for the role lists the responsibilities, the tools, the metrics and often the reporting line. It is a persona published by the employer.
-
3
Take their phrasing, do not summarize it
Call recordings are for quotes, not themes. A persona in your own words is a document about your assumptions with a job title on top.
-
4
Write six lines, not forty
The seat, the problem in their words, the workaround they run, what a good week looks like, who else has to agree, and what makes them stop reading.
-
5
Write the falsifier underneath it
A persona is a hypothesis, so give it a test it can fail: if this is right, this seat replies and that one will not.
The first persona I wrote took a week and changed nothing about the emails. The one that worked came out of an afternoon reading replies we already had, in an inbox nobody had read properly.
The subtraction test
Delete a line from the persona. If the email you would write is identical, the line was decoration. Run it over every line once and the document comes out shorter and more useful in the same pass.
Describe a person
Marketing Mary, 34, Boston. Values authenticity, hikes at weekends.
- ✕Nothing in it can be checked
- ✕Each added detail shrinks who it fits
- ✕It survives because nothing can kill it
Describe a seat and its Monday
Owns the weekly pipeline number. Rebuilds it by hand in a sheet every Monday morning.
Falsifier: ops leads reply, finance does not.
- ✓Verifiable at an account in ninety seconds
- ✓Every line moves a sentence you send
- ✓It carries the test that would kill it
Want the committee mapped on your top accounts?
Book a Fit CheckHow many personas, and who to open with
Two decisions about the same scarce thing: attention. Everyone answers the first with a number and nobody says where the number came from.
The exception is a price question. When one person can approve the spend there is no committee, so you open with the money seat. Aiming high by default is the expensive habit: Gong Labs reported in January 2026 that C-level executives reply to cold email less often than everyone else.
Two people can carry two personas well and four badly. The constraint is never research time, it is message variants: each is a sequence to write, watch and read, and the fourth always gets the least attention.
A worked example: one persona, built from four replies
An illustrative walkthrough of the method, not a specific client result. We report real numbers only when they are real.
-
1Week 1 · The material
Four replies, one sentence
Four people in one segment answered, all describing the same Monday: rebuilding one report by hand before a leadership call.
-
2Week 1 · The seat
Name the seat, not the title
Their four job titles did not match. The seat did: whoever owns the weekly number and rebuilds it by hand.
-
3Week 2 · The document
Six lines, in their words
The seat, their phrasing, the sheet they already run, what a good Monday looks like, who signs, and what loses them.
-
4Week 2 · The test
The line that could kill it
Written underneath: if this seat is right, ops answers and finance does not. Two send windows later the split held.
When the persona is wrong, and when it has just decayed
A wrong persona and a stale one look identical from the dashboard and need opposite fixes. The field version fits in two lines.
Same titles reply and never convert: the persona is fine and the company layer is wrong. Nobody at any title replies: check the ICP first. The full diagnostic is in ICP vs buyer persona.
- !A new function head moves the seats beneath them
- !A first dedicated hire for the founder's old job
- !A reorg splits one seat, or merges two
- !A quarterly review reminder, which moves nothing
Where persona work goes wrong
Four failures we see repeatedly, and one that costs a quarter. None of them looks like a failure while it is happening.
Four seats, four unrelated messages, one Slack thread where they compare them. A persona changes the angle, never the story.
Knowing a blocker exists is not the same as answering them. The seat you never covered arrives late, as a question nobody prepared for.
If every line can be selected in a database, you have written a list definition and given it a first name.
In a new category the seat may be split, or still held by the founder. Say so in the document instead of inventing a title.
A persona nothing can disprove never gets fixed
Give the document a name, a face and an age and no evidence can contradict it. The quarter it costs you is the one where every message is aimed at somebody who does not exist.
Where the common advice is wrong
Most persona advice is published by companies that sell persona software, and all of it points the same way: add more detail.
"Build a detailed persona for every role on the committee, give each one a name and a face, and review the set once a year."
- ✕More detail makes it more accurate
- ✕One persona per committee role
- ✕A name and a face make it memorable
- ✕Review the whole set annually
"Two short personas, six lines each, built from replies you already have, each carrying the test that would prove it wrong."
- ✓Each attribute added shrinks who it can describe
- ✓Map four seats and message two of them
- ✓A name and a face make it unfalsifiable
- ✓Rewrite when the org changes shape
What works now, August 2026
A deep-research prompt returns a formatted persona in under a minute. Adding an unverified attribute now costs nothing, and nothing in the tool checks whether the attribute is true.
Three jobs, all of them summarizing evidence you already own: clustering call recordings so their phrasing is searchable, pulling the metrics out of job posts, and sorting replies by seat.
It cannot tell you which seat holds the budget at a company that has never bought your category, and it will write a confident persona for a role nobody has.
We give it one job: turning a quarter of replies and twenty job posts into the buyer's own phrasing, sorted by seat. The moment it starts inventing attributes nobody observed, close the tab.
What to keep from all of this
The persona is not the deliverable. The change in what you send is, and the document exists to hold it.
- 1 A persona is a seat at a purchase, not a person at a desk.
- 2 Every line changes a sentence you send, or it goes.
- 3 Map four seats, message two, tell one story.
- 4 Rewrite on org change, never on a date.
Questions founders ask
What is a B2B buyer persona?
Do I need personas at all if I have a tight ICP?
How many people are on a B2B buying committee?
Which roles are on a buying committee?
How many buyer personas should a startup have?
Which seat do I write the first persona for?
Do buyer personas actually work?
Co-founder of Real Good GTM. He has been the first business hire and Chief of Staff at seed-stage B2B startups, building outbound pipeline before any playbook existed. This guide is the persona he actually uses on live campaigns: six lines pulled out of a reply inbox, with the falsifier written underneath it.
Connect on LinkedInThe layers either side of this one
The company layer sits above the persona, two of the four seats have a page of their own, and exclusion is the other half of the list.
The ideal customer profile
The company layer above this one: what has to hold inside an account before any persona matters.
Read the guideChampion vs economic buyer
Two of the four seats in depth: how to tell them apart, and which to open with.
Read the postThe negative ICP
The roles that reply often and buy never, and the companies that can never be a good customer.
Read the postWant the seats mapped, and the messages written per seat?
Book a fit check. We'll look at who actually decides inside your accounts, which seat to open with at your price point, and whether outbound is the right motion for your stage. If it isn't, we'll tell you that too.
Book a Fit CheckNo hard sell. No fake numbers. Real good work speaks for itself.