How to research an account before outreach

Research is a verdict, not a dossier

Ask most sellers to research an account and they’ll return with facts: revenue grew, a facility opened, the CEO mentioned AI. All true, none useful — twenty tabs of trivia and still no reason to write the first line of an email.

That’s because research done well isn’t collection, it’s a verdict in four questions, asked in order: what are they trying to change, who owns it, how do we fit, and what do we believe. Each question narrows the next; the last one ends in a sentence you can act on — or an honest no. An hour is enough, if the hour is spent on the questions instead of the tabs.

---

Step 1 — What is this account trying to change?

Where do you find live problems, in their words?

Half the hour goes here, because everything else keys off it. You’re not building a company profile — you’re hunting for change in motion, in the account’s own words:

• What they announce: initiative language in press releases, earnings remarks, the CEO’s letter. Companies narrate their own problems when they think customers and investors are listening.
• Who they’re hiring: job postings are confessions. Two integration-engineer roles is a systems consolidation saying its name.
• What leaders say out loud: interviews, podcasts, conference talks — where the polished version cracks and priorities show.

Record each finding as a claim with a receipt: what, source, date. Skip firmographics entirely — size and industry got the account onto the list; they have nothing left to tell you. (How the list itself gets built.)

Step 2 — Who owns the change?

Whose job is the problem?

A live problem with no owner is a headline, not an opportunity. Attach every change you found to a person: the leader who announced it, the VP hired to run it, the function whose week it ruins. The org’s own materials usually name them — and where they don’t, the gap itself is intelligence, because finding the owner becomes your first conversation’s job.

This step is deliberately shallow here: full buying-group work comes later, seat by seat. Before outreach you need the two or three people the change actually touches — the ones with a reason to answer.

Step 3 — How do we fit?

Does what they’re changing match what you fix?

Now the honest gate. Put their live problems next to the problems your best customers hired you to fix. A real match is specific — their stall is your stall, their consolidation is your consolidation — not “they’re digital-transforming and we’re software.”

If nothing matches, stop researching. Write the no, date it, set a re-entry signal, and give the rest of the hour to an account that deserves it. A fast, evidenced disqualification is the second-best outcome research can produce, and treating it that way is what keeps the working cohort honest.

Step 4 — Write the hypothesis

What do you believe — and what would prove it wrong?

Everything converges into one falsifiable sentence: this account is trying to change X, owned by Y, which creates a problem we solve for Z — and here’s what would prove us wrong. That sentence is the research. It tells you who to approach, what to say, and what the first conversation must test.

Then write it down where the team can see it — with the sources and dates attached — or accept that the hour just purchased a perishable good. Research that lives in a rep’s head gets bought again every quarter and invoiced as ramp time. The one-page format that holds it is its own post: the one-page account plan.

Illustrative hour

The hour, budgeted

Where the sixty minutes go — and what gets none of them.

1. 1 · What are they changing?Announcements · hiring · leaders' own words
2. 2 · Who owns it?Names attached to each change
3. 3 · Do we fit?Their changes vs. your best customers' problems
4. 4 · The hypothesisOne falsifiable sentence, sourced and dated
5. — · Firmographics, tech stack, triviaAlready did their job getting the account listed

Share of the hour (illustrative)

The trap, and the tooling question

Two ways this goes wrong. The first is the dossier trap — twenty true facts and no verdict, which feels like diligence and produces nothing a buyer would notice. The four-question order exists to prevent it: every minute serves the sentence.

The second is the re-purchase trap: the hour works, the knowledge lands in one head, and the company buys the same understanding again at every handoff. Writing the one-pager is the manual fix, and it holds — until keeping dozens of them current becomes a job nobody was hired for.

That’s the point to systematize the gathering, never the verdict. Narrative AI runs this method continuously — the account’s changes, owners, and fit evidence collected, sourced, and dated, the hypothesis kept current as facts move — so the seller’s hour becomes the judgment at the end instead of the tabs at the start; every message stays rep-initiated, and every AI-suggested CRM change waits for the rep’s approval.

Four questions, one sentence, one page. That’s research a buyer can feel.