How to build a hypothesis for every account on your list

A hypothesis is a claim you can be wrong about

Open ten reps’ notes on their top accounts and you will find ten versions of the same sentence. Good fit. Enterprise, growing, using a competitor, headcount trending up. That is not a hypothesis. It is a description wearing an optimistic tone. Nothing in it can turn out to be false, which means nothing in it can teach you anything.

A hypothesis is one sentence drawing a line between what this specific account is already trying to do and what you sell — written so that a reply can prove it wrong.

That last clause is the entire idea. If your sentence cannot be disproven, it is not a hypothesis, it is a pitch with the serial numbers filed off. Run the test before you write a single message: what reply would make me delete this claim? If you cannot name one, you have written something unfalsifiable, and unfalsifiable claims cannot be tested — only repeated, louder, until the account goes quiet and you file it under “no budget.”

The reason this matters is not intellectual hygiene. It is that a falsifiable claim converts every reply into information. An unfalsifiable one converts every reply into noise, and a quarter of noise is indistinguishable from a quarter of bad luck.

The anatomy: motion, friction, claim, falsifier

A hypothesis you can actually test has four parts, and the fourth is the one almost nobody writes down.

1. The motion. Something the account is already doing, with a name and ideally a date. Not “they care about efficiency” — a funded initiative, a new hire, a facility, a mandate, an integration.
2. The friction. What makes that motion hard, expensive or slow with what they have today.
3. The claim. Why what you sell is worth buying because of that friction, and why now rather than next year.
4. The falsifier. What would have to be true for this to be wrong, and how a reply would tell you.

Here is a weak one for a regional logistics firm:

Meridian Freight is growing and opening new facilities, so they need better workforce management.

Growth is not a deadline. “Better” is not a claim. And crucially, no answer contradicts it — if they say yes, the sentence survives; if they say no, the sentence survives. It cannot lose, so it cannot win.

The same account, made falsifiable:

Meridian is opening a second distribution center outside Columbus this fall and has been posting warehouse roles for it since April. Their onboarding today is a binder and a buddy — that worked with one site and one shift lead, and it does not survive two. So the VP of Operations funds a training platform before the site goes live, not after the first peak season goes badly.

Now name how it dies. The site slipped to next year. They are staffing the new center through a 3PL that runs its own onboarding. The VP already signed something in Q1. Each of those is a real answer, each arrives in a two-line reply, and each is worth more than another follow-up sent into the dark.

A second one, for a manufacturer under a margin mandate:

Weak: Now that they are PE-owned they are under cost pressure, and we help reduce costs.

Everyone with a P&L is under cost pressure. Rewrite it so the account could contradict you:

Their new owner put a margin target on the record at the last earnings call, and the plant’s clearest path to it is the unplanned downtime on the two oldest lines — the ones the maintenance lead has been quoted on twice. So the operations leader would rather spend on condition monitoring this budget cycle than defend another quarter of overtime. If that margin plan turns out to be a headcount plan, we are wrong, and the first reply will say so.

A third, compressed, for a software company post-acquisition: they merged two support organizations in January and are still running two ticketing systems, so the new VP of CX is buying consolidation before the renewal of the smaller contract in Q4 — unless the acquired stack is the one they are standardizing on, in which case we are calling the wrong side of the house.

Notice what the strong versions have that the weak ones do not. A date. A named role. A specific mechanism. And an explicit way to lose.

Without research, it is a guess with better grammar

You cannot write any of the sentences above from a firmographic filter. They come out of step 2 — mapping and researching the account — and the quality of the hypothesis is capped by the quality of that work. Three sources do most of the lifting:

What they have said in public. Earnings calls, funding announcements, press releases, the roles they are hiring for right now. Job postings in particular are budget that already cleared committee — nobody posts a req for a strategy they abandoned.

What your own system already knows. The closed-lost from eighteen months ago with the note “revisit when they finish the ERP migration.” The two contacts who opened everything and never replied. Most reps write hypotheses as if their company met this account for the first time this morning.

What the people are saying. Someone taking a new title, a leader posting about a priority, an operator complaining about the exact problem you solve.

And one more input that only exists if your team has been keeping score: the deal shapes that have already closed for you. If the last four wins all started with an operations leader inheriting a second site, that is not a coincidence, that is a pattern, and it should be the first hypothesis you try on the fifth one.

Everything downstream is written against it

Once the hypothesis exists, it stops being a note and starts being the specification for the work.

Every 1:1 message on that account gets written against it — not “personalized” with a line about their podcast, but actually arguing the claim to the person best placed to confirm or deny it. The CFO gets the margin version, the plant manager gets the downtime version, and both are the same hypothesis aimed at different desks. That is what multi-threading is for: you are not collecting contacts, you are collecting evidence from the people who would know.

Then the reply resolves it, and there are only three outcomes. A booked meeting — the claim held, go deeper. A disqualification — the claim failed, and the account comes off the list with a written reason instead of drifting through your pipeline for two quarters. An iteration — part of it held, part did not, so you revise the sentence and test again.

The middle one is the outcome most teams refuse to take. A clean no in week one is more valuable than a warm maybe in month three, because it returns your most finite asset — the hours you have to work accounts — to an account that might actually convert.

This is the part of the motion Narrative holds directly: one account hypothesis per account, drafted from that account’s research, carrying its own confidence score that rises when a reply confirms it and falls when the account goes silent or contradicts it, with every draft written against the current version and stamped with the version it was written from. The rep still edits every line and clicks send themselves — Narrative never sends for you, and nothing writes back to your CRM without a human approving it.

Written down, a hypothesis becomes coachable

Here is the team argument, and it is the one that changes how a manager spends their Monday.

When hypotheses live in reps’ heads, a manager can only coach activity: touches, dials, sequences enrolled, accounts worked. All of it is a proxy for the thing that actually matters, which is whether the rep has a defensible reason to believe this account would buy. So pipeline review becomes an accounting exercise. Everyone reports effort and nobody reports thinking.

When the hypothesis is written down, the review changes shape. You can read a rep’s sentence and see immediately that it has a motion but no friction, or a claim with no way to lose. You can coach that — in ten minutes, on one account — and the rep applies it to the next thirty. You are correcting the reasoning that generates the behavior instead of the behavior itself.

It also makes the account survivable. When a rep leaves, goes on leave, or hands an account to an AE, “why we thought they would buy” does not walk out with them. And across a team, hypotheses accumulate into something a playbook cannot fake: the shapes that keep converting, in the words of the people who tested them.

Start with five

Do not try to hypothesize your whole list this week. Take the five accounts you would be most annoyed to lose, and for each one write the sentence — motion, friction, claim, falsifier — and then write the reply that would kill it. If you cannot write the killing reply, you have not done enough research to have an opinion yet, and that is useful to learn on a Tuesday rather than in a QBR.

Then send against them, and let the answers come back. The rest of the motion — building the list, doing the research, threading the account, testing with 1:1 messages — is laid out step by step on the prospecting page. Step three is the one that turns the other four from activity into a method.