How to research a prospect before a cold email: a 5 to 10 minute checklist

Prospect research checklist: five lines in 5 to 10 minutes

To research a prospect before a cold email, spend 5 to 10 minutes finding five things and write down a URL and a date for each: what they sell and to whom, one recent change, one hiring fact, who owns the problem, and one reason to write now. Then put only what those notes support into the email.

Set a timer. If a line takes more than two minutes, write "not found" and move on. An empty line is better than a guessed one.

1. What they sell, and to whom

Read the product or pricing page, not the tagline. Write one plain sentence ("dispatch software for trucking fleets of 10 to 100 trucks") and the URL, then check that the account fits your offer at all. Many product pages show no date, so record the day you read it.

2. One recent change

Look for something dated in roughly the last 90 days (our rule of thumb, not a standard): a changelog release, a press release, a partnership or a funding round. Record the date it was published, not the date you found it.

Funding needs care, because articles often repeat one another. Prefer the company's own announcement.

For a US private company, a Form D filed with the SEC can be a primary record: the SEC says a company must file the notice within 15 days after the first sale of securities in the offering, and a filed Form D is publicly available on EDGAR. It is a notice of an exempt offering, not an announcement. Only offerings that rely on certain exemptions, mainly Regulation D, file one, so a round done another way will not appear and a missing filing proves nothing.

3. Hiring as a signal

A posting hints at what a company is about to build or buy: an onboarding hire suggests onboarding is getting attention. The buying signals guide covers other kinds. Record the role, URL and posted date.

Two cautions. A posting shows intent, not a proven problem. Postings also go stale: Google's job posting guidelines say jobs that are no longer open must be expired or removed, but that governs what Google lists, and a careers page can still show a role that has been filled. Greenhouse's job board API documentation shows updated_at and first_published fields without saying what they mean, so read a timestamp as "seen on this date", not "opened on this date".

4. Who owns the problem

Find the one person who would feel the problem your offer solves, using the company's own team page, a byline or a speaker page. Record name, title, URL and date, and only professional facts.

Titles go stale. The US Bureau of Labor Statistics reports that in January 2026 wage and salary workers had been with their current employer for a median of 4.1 years, and 20.6% had a year or less. That figure covers all wage and salary workers, not decision-makers in your market, but it shows why a team page or data snapshot a year old is likely to be out of date in places. Re-open it before you send.

5. One specific reason to write now

Combine one fact from line 2 or 3 with your offer in one sentence, and label it as your inference: "API v2 plus an onboarding hire: docs may be about to matter." In the email, present it as a guess or a question. If no dated fact connects to your offer, you have no reason now: send a plainer email or leave the account for next week.

Dates: record when it was published and when you read it

Two dates matter. Published is when the thing happened or was posted. Read is the day you saw it. Keep both when they differ.

• Use the date the page shows. Google says it does not rely on a single date signal because any one can be prone to issues, and it advises sites to show a visible date. So prefer a date printed on the page to one inferred from a search snippet.
• No date is not "recent". Write the read date and "no date on page", and keep timing claims out of the email.
• Save a copy. A screenshot or PDF works. The Internet Archive's Wayback Machine can also capture a page as it appears now, and its availability API returns the capture closest to a date you give. A saved copy gives you a dated record of what you read.

The one rule of sales prospect research: cite it, ask it or cut it

An email may state a fact about the prospect only if your notes hold a URL and a date for it. Everything else must be one of three things:

• Your own offer. What you sell needs no source.
• A question. "Is documentation for the new API on someone's plate?" asserts nothing.
• A marked guess. "My guess, and it is only a guess, is..." is honest because it says so.

If a sentence is none of those and has no source, cut it. A wrong claim ("congrats on the Series B" when there was none) can cost more than a plain email, because it tells the reader you did not look, or looked badly. For AI-written personalization, see personalization without hallucinations.

Account research template: a worked example with a fictional company

Copy these columns and fill one row per line. Everything below is illustrative: Harborline Dispatch is a made-up company on a reserved example domain, and you sell API documentation and onboarding guides to logistics software teams. After eight minutes, your notes read:

One line was cut: a forum comment said Harborline had "just closed a Series A", with no filing, announcement or date. The email the notes allow (footer with your postal address and opt-out left out for space):

Subject: Docs for Harborline's API v2

Hi Dana,

Harborline's changelog shows API v2 with load-status webhooks shipped on September 22, and a Customer Onboarding Manager role was posted on October 6.

My guess, and it is only a guess, is that documentation for the new API is about to land on someone's plate. We write API guides and onboarding checklists for logistics software teams.

Would it help if I sent a one-page outline for the webhooks section? No call needed.

Check it sentence by sentence:

Before sending, re-open the careers page. If the role is gone, that sentence goes, and so does most of the reason.

How long research stays usable

Notes are perishable, and we know of no fixed rule for how fast. Our rule of thumb: treat news, funding and hiring notes as stale 90 days after you read them, and everything else after about a year. If a note is more than a few days old, re-open the source before you send, because pages change. FoxKnock's cards use the same windows, counted from when a card was retrieved.

Does researched personalization pay?

Published numbers come from vendors that sell outbound, and they disagree:

• Woodpecker reports average reply rates of about 17% to 18% for emails with advanced personalization against 7% to 9% for basic or none. The page (updated June 23, 2026) defines advanced personalization as custom snippets beyond a first or company name, but describes no sample size, date range or method for the comparison.
• Sales.co, a done-for-you cold email agency, published on September 27, 2026 an analysis of 1,279,153 contacts first-emailed between April 2025 and September 2026. Contacts whose first email opened with an AI-written line (360,004) had a human reply rate of 0.53%, against 1.00% for a plain template (28,896). It compared groups sorted by how the first email was written, does not describe random assignment, and says the data covers only its own tool and clients. Among 19 clients that ran both on at least 300 contacts each, the AI line had the higher human reply rate in 10.

Both are vendor data, and neither says whether the personalized lines were checked against a source. The Sales.co test is about AI-written opening lines, not researched and cited facts. So compare your own batches: same segment, same offer, a cited opener against a plain one.

Notes about people are personal data: GDPR and CAN-SPAM

Under the GDPR, personal data means any information relating to an identified or identifiable natural person, so a note that names someone and says where they work qualifies. If the GDPR applies to your outreach, the URL-and-date habit does more than keep you honest:

• Article 14(2)(f) requires you to tell a person where their data came from, when it did not come from them, including whether the source was publicly accessible. Notes with URLs let you answer. Article 14 also lists other items to tell them.
• Article 14(3) makes that information due, when you use the data to contact the person, at the latest at your first message, and in any case within one month of obtaining the data. Article 14(5) lists exceptions, for example when the person already has the information.
• Article 5(1)(c) and (d) ask for data that is limited to what is necessary, and accurate and, where necessary, kept up to date: professional facts only, with dates.

In the US, the FTC's CAN-SPAM guide says the law covers all commercial email and makes no exception for business-to-business email. It requires accurate header information and bars deceptive subject lines: the subject line must accurately reflect the content of the message. A fake "Re: our call" for a call that never happened would be hard to defend under that test. More in our GDPR cold email guide.

This is not legal advice: whether legitimate interest covers your outreach depends on your facts and your recipients' countries, so ask a lawyer who knows them.

Doing this in FoxKnock

FoxKnock's prospect research covers the same ground and turns what it finds into evidence cards. Each card holds a one-sentence claim, the source URL and title, a short quote when one is available, a confidence score from the AI, observed and retrieved dates, and an expiry date. A fact is dropped unless it cites a page the agent actually retrieved. That checks the URL, not that the AI read the page correctly, so open the source before an email repeats a claim.

News, funding, hiring and event cards expire 90 days after retrieval, and the rest after 365 days. Expired cards are not used to write emails.

There are three depths, at default credit prices that can change (see pricing):

• Quick (100 credits): the homepage plus up to three key pages, and a news search over the last 365 days. It has no live web-search agent.
• Standard (225): a Claude web-research agent with up to six searches and six page fetches: news and funding over 12 months, hiring, likely pains and a "why now".
• Deep (550): up to twelve searches and fetches, second-source checks on the strongest claims and, if you pick a key contact, a research pass on that person.

Reports are charged only when they complete. At 1 credit = $0.01 list price, a standard report is $2.25 at default prices.

When FoxKnock drafts an email, each specific claim comes back with the id of a card, and a second AI pass flags claims that no card or knowledge item supports. An unsupported claim blocks the draft until a person fixes or removes it. The check is AI-based, not infallible, and templates you write yourself skip it. You still judge whether each source supports its line.

Sequences start at Review each, so every email waits for your approval, and automatic sending is opt-in and has to be earned. AI features send content to Anthropic, which makes Claude.

To compare, run the checklist by hand on a few accounts, then create a free account and run a report on one of the same companies. The Free plan has 500 one-time credits (granted once you verify your email) and needs no card; at default prices that covers two standard reports.