How to research a company before you contact it

Research is for improving your judgement, not for decorating your message. Almost every mistake in pre-contact research follows from getting that backwards.

Get it backwards and you land in one of two places. You spend an afternoon on five companies and contact none of them, because the research became the work. Or you skip it and contact fifty blind, because it never seemed to end. The way out is to know what you are looking for: whether to contact this company at all, and what to ask if you do.

Three buckets

Everything you find belongs in one of three categories, and keeping them apart is most of the skill.

Verified facts

Things the company has published about itself in a form that is hard to misread. A contact address written as a mail link. A phone number written as a telephone link. A company page on a known professional network. A service on their services page.

These are strong because the structure carries the meaning. A number inside a telephone link is a number the company intends you to use. The same digits sitting loose in a block of text might be a fax line from 2019.

Reasonable hypotheses

Things you infer. They have four offices, so coordination between them is probably non-trivial. Their careers page lists three open consultant roles, so they are probably growing. Their enquiry form asks nine questions, so intake is probably slow.

Each is a decent guess and none is a fact. "Probably" is not padding. It is the accurate description of what you have.

Still a question

Things you cannot know from outside and should stop trying to. How they feel about their current software. Where the real bottleneck sits. Who decides. What they have already tried and abandoned.

These are not gaps in your research. They are the agenda for your first conversation.

A bounded checklist

Five things. When you have them, stop.

What they actually do. Not the industry label, the services. A recruitment agency doing contract logistics placement has different problems from one doing permanent executive search.

Rough size and operating context. Enough to know whether you are talking to one person who does everything or a team with defined roles. Team pages, office locations, review counts, and listed opening hours all give you a read.

How enquiries reach them. A form, a general inbox, a named person, a phone number. This is often the single most informative thing on a small company's website, because it tells you how inbound work is handled.

Visible signals of change. Recent hires, a new office, a new service line, a rebuilt site. Change creates the conditions where a conversation is welcome.

Contact relevance. Whether the person you are about to write to plausibly owns the work you want to discuss. Writing to the wrong person is not neutral. It costs you the company.

Record the source and your confidence

For each item, note where it came from. It takes seconds and it changes how you use the information later.

Compare these two notes:

• "Their contact page lists a single shared inbox." Source: their contact page. Confidence: high.
• "They probably manage candidates in spreadsheets." Source: my assumption from team size. Confidence: low.

Both are worth having. Only the first can appear as a fact in a message. The second belongs in your questions, phrased as a question.

Confidence is not a formality. A fact taken from a structured, intentional signal deserves more weight than one scraped out of marketing copy, which deserves more than something you inferred at the end of a long afternoon.

Writing a hypothesis honestly

The wording in your own notes shapes what you believe an hour later, so hedge deliberately.

You are looking at a small property firm with a busy enquiry form and no visible team page.

• Weak note: "Slow lead response is killing them."
• Honest note: "Property enquiries are usually time-sensitive. With a shared inbox and no visible team structure, response speed may be a bottleneck. Worth asking how enquiries get picked up, and by whom."

The second is longer and less exciting. It also survives contact with the actual person, which the first will not.

Where the pattern runs out

Pattern-based research has a hard edge, and it is worth knowing where yours is.

Any research method built on industry patterns, whether it is software or your own accumulated instinct, recognises the industries it was built for. Bring it a business that does not match one and it falls back to assumptions that fit almost any small firm: repetitive admin is probably absorbing someone's time, enquiries may not be triaged consistently, process knowledge may live in people's heads.

Those defaults are honest, and they are also exactly what you would have guessed without looking. That is the signal. When research produces only generic conclusions, it has told you nothing, and the correct response is to ask rather than to assert. A specific question beats a confident guess, and it is far easier to recover from.

The stopping rule

Stop when you can name one specific thing worth asking about.

Not one impressive thing to mention. One thing you genuinely want to know, arising from something you actually saw, that they would find reasonable to be asked.

If you have it after four minutes, stop at four minutes. If you have not got it after twenty, the honest conclusion is usually that the public information is thin, which is worth recording in itself.

What not to pretend to know

Public information does not reveal internal pain. A messy website is not proof of a messy process. Plenty of profitable firms have terrible websites, and plenty of struggling ones have excellent ones.

A score is not a prediction. Whatever scoring you use, a tool or a back-of-envelope rating, it sorts your own attention. It does not predict whether a company will buy. Treating a number as a forecast is how you talk yourself out of a good prospect because a heuristic disagreed.

Absence of evidence is not evidence. No visible CRM on a website tells you nothing about whether they have one.

Research is not a licence for familiarity. Finding a fact does not create a relationship. Reference something specific because it genuinely shapes your question, not to demonstrate that you did homework. That distinction is the whole difference between research that earns a reply and personalisation that reads as performance.

What this looks like in Scout

Scout reads a prospect's public website and extracts contact details, links, and service signals, attaching a source and a confidence level to each value. Nothing reaches the prospect record until you approve, edit, or reject it, which is the three-bucket discipline above expressed as a screen. It also produces a short briefing per company, with the evidence behind each suggested pain point visible rather than summarised away.

Two limits stated plainly. The pain-point view is rule-based rather than learned, so treat it as a structured starting point rather than an assessment of that business. And we have not tested extraction accuracy with a group of external customers, so the only accuracy claim we will make is the one you can check on the screen in front of you.

Explore Scout

The method above came out of building that product, and there is no customer study, survey, or performance data behind it. Use it the way you would use any working method someone handed you: try it on the next five companies, and keep the parts that survive.