Most outbound prep is either too thin or too heavy. Too thin is a name, a title, and a script. Too heavy is a page of biography the rep never uses after “how have you been.” Neither produces a reason for the buyer to stay on the line.
An account hypothesis is the middle path. It is a one-page claim you can say in fifteen seconds, defend if the buyer pushes back, and abandon if they tell you it is wrong. This guide is the operating version: what to look for, what to ignore, how to write the first line, and how to keep the email on the same story.
If you want the weekly argument for why this matters now, read AI Made Prospect Research Cheap. It Did Not Make It Useful.. If you want the same workflow inside an AI tool, install the free account-hypothesis skill.
1. Define the Job Before You Open a Tab
Write three sentences before you research anyone:
- Who you help, in operating language rather than category language.
- The workflow that changes if they buy.
- The person who owns that workflow when it fails.
If you cannot write those three sentences, research will only generate decoration. “We sell to construction companies” is not a job. “We pull spreadsheet takeoff off preconstruction teams so estimators review risk instead of counting fixtures” is a job. The second sentence tells you which pages matter and which titles to dial.
“A useful ICP is a workflow plus an owner. Industry and employee count are filters, not the reason to call.”
2. Collect One Fact, Not a Dossier
You are looking for a fact that implies load on the workflow you change. Good sources are public and checkable:
- Careers pages: roles that match the bottleneck (estimators, SDRs, RevOps, implementation, support).
- Product, solutions, or customer pages: how they describe the current process.
- Leadership or newsroom pages: a new owner for the workflow, not generic “growth.”
- User-provided context: a CRM note, a lost-deal reason, an event question, a reply.
Ignore most of what models like to recite. Founding year, office locations, mission copy, and “innovative culture” almost never change the first fifteen seconds. Tech-stack guesses are worse: if you did not verify the tool, do not mention it.
Stop at three candidate facts. Pick the one that best implies your problem. Park the others. If none imply your problem, the account may still be a fit later, but it is not a fit for this call. Skip it.
3. Write the Four-Line Card
Force the research into this shape. If a line is missing, you are not ready to dial.
- 1Fact: the checkable observation, with the source in parentheses.
- 2Problem: the workflow that likely breaks.
- 3Consequence: slower, costlier, riskier, or less visible — named in buyer language.
- 4Why now: a real trigger, or the honest label “ranked cold.”
Example, labeled as an illustration, not a customer story: a GC is hiring two estimators and still describes spreadsheet takeoff on its services page. Fact is the hiring plus the described workflow. Problem is bid volume outrunning estimator capacity. Consequence is thinner risk review and overnight takeoff. Why now is that they are buying more people for an unchanged process.
That card can be wrong. The VP might say they already bought takeoff software, or that hiring is covering it. Good. A hypothesis that can be corrected is doing its job. A compliment cannot be corrected. It can only be deleted.
4. Map One Owner, Then Maybe One More
Call the person who owns the consequence. Adjacent titles are not a multithreading strategy if they receive the same paragraph. The estimator lives in the workflow. The VP of Preconstruction owns coverage and bid quality. The CFO owns overtime and the cost of bad coverage. Those are three messages, not one message sent three times.
For a first pass, pick one primary owner. Add a second only when the account is high value and the first owner is a known bottleneck. Log both so two reps do not collide. Dialfyne team campaigns lock a lead on claim for that reason; the same rule applies in any dialer you already use.
5. Write a Spoken Opener, Then the Email
Write the call first. Email written first tends to sound like email on the phone.
- Length: 12–15 seconds. One breath, then a question.
- Shape: fact, implied problem, question. No product name until they engage.
- Question: something they can answer without a meeting. “Is coverage the constraint, or is takeoff still the overnight work?”
- Stranger test: a competitor who only knows the industry should not be able to send it.
Voicemail, if you leave one, is the opener minus the question plus a reason to call back. One voicemail in a cadence is enough unless the next attempt has new evidence. Repeating the same recording is not persistence. It is noise. For a full call schedule, use the 10-day outbound calling cadence.
Email 1 is the same hypothesis in under 80 words, with one question and no fake familiarity. Email 2 is shorter, adds a smaller ask or new evidence, and does not restate Email 1. If platform email is part of your workspace, keep the thread on this story instead of launching a second sequence with a different angle.
6. Practice the Pushback Before You Burn the List
The hypothesis will meet four common responses: “that’s not our problem,” “we already have a tool,” “just send information,” and “now is a bad time.” Write one honest line for each. Concede when the objection is valid. “If hiring is covering capacity, I’ll stay out of the way” is stronger than a pivot into features.
Say the opener out loud once. Then run it against a skeptical buyer. Dialfyne AI Roleplay is built for that: a custom AI buyer, a scorecard, and no live pipeline at risk. Three free sessions are enough to find out whether the question survives interruption.
7. Run It as a Team Play, Not a Personal Hobby
If every rep invents a research ritual, you will get heroics and empty notes. Make the card the unit of work:
- 1Managers approve the three-sentence offer definition once per segment.
- 2Reps attach the four-line card to the lead before the first dial, or skip the account.
- 3After twenty live conversations, review which hypotheses were accepted, corrected, or rejected.
- 4Rewrite the segment hypothesis from those calls. Do not ask the team to “personalize more.”
This is the same loop as the objection-led micro-campaign: one claim, one audience, evidence from real calls, then a decision to scale, revise, or stop.
8. What Good Looks Like in a Week
- Monday: freeze the offer definition and one segment.
- Tuesday: each rep writes five cards, not fifty emails.
- Wednesday: roleplay the opener and the two most likely objections.
- Thursday–Friday: dial the cards. Keep email on the same hypothesis. Exit cleanly on no-fit and opt-out.
Measure conversations that confirm, correct, or reject the hypothesis — not custom-token count in the sequencer. A corrected hypothesis is a win. It gives the next touch a real fact. A deleted compliment teaches you nothing.
How Dialfyne Holds the Loop Together
The free prospecting skill produces the card. AI Roleplay pressure-tests the opener. The Dialfyne Power Dialer is the human-led calling path: local presence, voicemail drop, team campaigns with claim-and-lock, live coaching, and scorecards on the same dimensions you can practice. Platform email, where it is enabled for the workspace, stays on the same suppression and story instead of becoming a second outbound system.
Reps and browser softphones are free. Usage draws from one shared credit balance. The product is useful after the hypothesis exists. It will not invent a reason to call if the team skipped that step.