This is a step-by-step guide to the give-first cold email: one sentence and a link to a free list you built for one company you want as a client.
We’ll build one list from start to finish for Ladlepay, a payroll tool for independent restaurant groups in the US.
In the example, a pool of 9,000 restaurant groups becomes a list of 150. Every group on it is opening a new location, and every row names the person who runs payroll there, with a checked email address.
This guide is for founders, agencies and anyone whose service gives a sales team leads or lists. If you already have a checked list of companies, skip to step 7.
How a give-first email works
Most cold emails describe work you could do for the reader. A give-first email does the work once, for one company, and sends the result for free.
The list is the proof. The reader opens it, checks a few rows and knows within minutes whether your work is good. If they use it and never reply, your best work still reached the one company you wanted most.
You make the offer after they reply. When they answer, you offer to keep the list fresh every month, or to build a second list around a different event, like groups hiring a new general manager. That’s the paid part.
In my view it fits best when your service produces something the reader’s sales team would otherwise build themselves, like lead lists, account research or tracking which companies are hiring. It fits worse when your work only shows results after months.
One careful list takes a few days. So you send it to one company at a time, never as a campaign.
Steps 1 to 3 decide what goes on the list. Steps 4 to 6 find the companies and the people and check both. Steps 7 to 9 turn the list into a sheet and an email and check it one last time.
1. Pick one company you want as a client
Pick one company you’d love to work with that clearly pays people to find new customers. This is your dream company. Four things make a good pick:
- You can find their buyers on the web. If you can’t, you can’t build the list.
- They do outbound today. Outbound means contacting people who haven’t asked to hear from you, by cold email, cold calls or LinkedIn. Open job ads for SDRs (sales development reps) or account executives show it. Your list then replaces work they already pay for.
- One person owns new business. A head of sales, a VP of growth or, at a small company, the founder.
- A deal would matter to you. You’re about to spend days on one email.
For this guide, the dream company is Ladlepay, a payroll tool for independent restaurant groups in the US. Its homepage talks to groups with 2 to 15 locations. Its case studies are full-service and fast-casual groups, and its sales job ads talk about selling to owners and operators.
2. Turn their ICP into tests a stranger can check
The ICP (ideal customer profile) describes the companies that fit a product best. Your dream company has already published its ICP. It’s spread across three places:
- Website: who the homepage speaks to, the company sizes it mentions, and the integrations page, for example which POS (point of sale) systems it connects to.
- Customer logos and case studies: the kinds of companies that already buy. These are also the companies you leave off your list.
- Job ads: ads for sales roles say who the team sells to, at what size, and often which job titles they call.
Turn every rule you find into a test that a single row passes or fails. For Ladlepay:
ICP tests: Ladlepay
1. Independent restaurant group, not a franchise of a national chain
2. 2 to 15 locations in the US, counted on the group's own website
3. Full-service or fast-casual concept
4. Not already a Ladlepay customer (checked against logos and case studies)
Each test needs a source someone else could check. “Looks like a good fit” isn’t a test, because nobody can check it.
3. Pick one signal you can prove
A signal, short for buying signal, is a recent change at a company that makes the dream company’s product more useful right now. Pick exactly one. The email has to say in one line what every row has in common, and that only works with one signal.
To be usable, a signal passes four rules:
- Provable: there’s a page a stranger can open to see it.
- Dated: you can say when it happened.
- Relevant: one clause explains why it makes the product more useful.
- Expiring: it’s only true for a while, which is what makes the list urgent.
Here’s how I’d compare the options for Ladlepay. How long each one stays true is my estimate, so adjust it for your market.
| Signal | Where the proof is | Stays true for | Fit for payroll |
|---|---|---|---|
| Opening a new location | Location page or news post on the group’s own site, job ads naming the new address | From the announcement until two months after the opening | Strong: new staff, new payroll setup |
| Hiring many hourly staff at once | Job ads | Two to four weeks | Medium: restaurants hire all year |
| New general manager or head of operations | LinkedIn post, team page | A few months | Medium: many small operators aren’t on LinkedIn |
| Raised money | Press release | Several months | Weak: rare for independent groups |
I’d pick the new location. The proof sits on the group’s own website with a date, and a new location puts a group of new employees on payroll within weeks.
Count before you build. Test the signal on one metro area first and count how many groups show it. If the count is close to zero, the signal won’t fill a list. Pick a different signal for the same need instead of widening the region until the number grows.
4. Find the companies that show the signal now
Start with a wide pool and narrow it down. These sources build the pool and find the signal:
- Google Maps lists almost every restaurant with its address and website. You get the data through a Maps API on RapidAPI or a Maps scraper on Apify. Both are marketplaces for small tools that pull data from sites like Google Maps into a spreadsheet.
- Job ads show hiring for a new address. BlitzAPI and LeadMagic sell job ad data through an API, a way for a script to ask a database for data. Many groups also post jobs on their own careers page.
- The group’s own website is where you prove the signal: a “coming soon” or “now open” location page, or a news post.
One group shows up on Maps as several restaurants that share one website. Merge the listings by website domain, so you have one row per group.
Say Ladlepay’s customers sit in ten metro areas, and a Maps search there returns 130,000 restaurants. Merged by website, they include 9,000 groups with two or more locations. That’s your pool. For scale, the Census counts 528,714 restaurant locations in the whole US (2023).
I’d run the rest in Claude Code, an AI assistant that writes and runs scripts on your computer. It writes a script that opens each group’s website, counts the locations and checks the four tests. A group that fails a test drops out.
For the signal, the script sends each group’s locations and news pages to an AI model with one narrow question: “Is a location opening or newly opened? Quote the sentence and the date.”
Then the script checks the answer. The quote has to exist on the page, and the date has to fall inside your time window. For Ladlepay that’s an opening in the next 90 days or in the last 60. The model only reads, and the script makes the decision.
Open the Terminal in a new folder, type claude to start Claude Code and paste:
Read groups.csv (one row per restaurant group: name, website, city,
state) and icp_tests.txt. Build tools/check_groups.py.
For each group: download the homepage, the locations page and the
news page. Count the locations. Ask the model each ICP test and
"Is a location opening or newly opened? Quote the sentence and the
date." Save every answer with the quote and the proof URL.
Then decide in code: the quote must exist on the page and the date
must fall between 60 days ago and 90 days from today.
Write passed.csv and excluded.csv with the test that failed.
Run it on test_groups.csv first and show me where it disagrees.
Test the prompt on groups you already know. Before the full run, label 30 groups by hand in test_groups.csv: some that open a location soon, some that don’t, a franchise, a single restaurant. Fix the questions until the model agrees with you on all of them.
Here’s how the numbers could look for Ladlepay:
- 9,000 groups in the pool
- 2,600 pass the ICP tests
- 260 show the signal, according to the model
- 190 pass the script check: the quote is on the page and the date is in the window
- 150 have a checked person and email after steps 5 and 6
Check the signal again the day before you send. Every row keeps a proof URL, the link to the page that shows the signal, and the date you saw it. Openings move and pages change, so re-open every proof URL before you send. A row drops out if its page is gone, or if 60 days have passed since the opening.
5. Find the person who runs payroll at each group
The right person is whoever the dream company’s sales team would call. For Ladlepay that’s whoever owns payroll at a restaurant group, and that changes with the size of the group. Write it down as a contact cascade, an ordered list of job titles to try at each size:
Contact cascade: Ladlepay
2 to 4 locations: owner or managing partner
5 to 9 locations: director of operations, then owner
10 to 15 locations: head of people or HR, then controller, then director of operations
Never the main contact: a single location's general manager, chefs, info@ addresses
Find people at the larger groups in a people database. Start with GetLeads and BlitzAPI. Both are flat rate: one fixed price a month instead of a price per contact. Use Prospeo last, because you pay per contact there.
Owners of small groups are often not on LinkedIn. People databases are built mostly from LinkedIn, so they find few of these owners. These sources work better:
- a web search for “who owns [group] in [city]”, run by a script through a search API, with an AI model reading the results
- state business registers, which list who runs each company, where the state publishes them
- the group’s About page and local press
Only keep names with a source. If there’s no page behind a name, it doesn’t go on the list.
People databases can be months behind when someone changes jobs. So confirm the current role on a live source before the row goes in, like the person’s own profile or the group’s team page. Read the exact title too, because a search for “owner” can return “assistant to the owner”.
Keep everyone whose title is in the cascade, not only the first match. The best match goes into the contact columns of the sheet, and the others go into an “other contacts” column. That keeps one row per group, and the reader still sees everyone their team could call.
6. Check every email before it goes on the list
Find the email first. Then verify it with a different tool.
- Find: Build the usual formats, like first.last@group.com, first@group.com and flast@group.com, and verify them. Pay an email finder like Prospeo or LeadMagic only for the people none of them works for.
- Verify: Check every address with an email verifier like MillionVerifier. A verifier asks the mail server whether the mailbox exists, without sending an email. Keep “valid” and drop every “unknown”, because the verifier couldn’t confirm the mailbox.
- Catch-all domains: Some mail servers accept mail for any address, so a normal check can’t confirm them. Keep these only after a tool built for it, like LeadMagic or BounceBan, confirms them.
- Cost: Most finders charge per email found and most verifiers per email checked. For 150 rows the total is small.
Then check every row against its person:
- The person still works at the group.
- The title matches the cascade.
- The email is on the group’s own domain. An address at the person’s old employer or side business doesn’t count.
If a row has no checked email, leave the cell empty. The reader forgives an empty cell. A wrong address in a row they check makes them doubt the rest.
Give the sheet an email status column with two values, “valid” and “catch-all, confirmed”. The reader sees at a glance which rows are safe to email.
7. Put the list in a shared Google Sheet
Use a Google Sheet that anyone with the link can view. It gets two tabs with these columns:
Tab 1: List
company | website | city, state | locations | signal (one line) | proof URL |
date seen | contact name | title | email | email status | LinkedIn URL |
other contacts | why this row fits
Tab 2: How this was built
ICP tests | signal and time window | sources | date of last check |
what was excluded and why
Four rules make it feel finished:
- Title it for them, with the date of the last check: “Ladlepay: 150 restaurant groups opening a location, checked Oct 26”.
- Sort by the date of the signal, newest first. The reader starts at the top.
- Keep it plain. Freeze the header row. No formulas, no colors that need a legend.
- Describe the list, never their business. Your counts say what you found. They say nothing about what their sales team is missing, so leave that sentence out.
The second tab shows the reader exactly how you work, and that’s what they’d be buying.
8. Write a one-sentence email
This email has one job: getting them to open the sheet. One sentence, the link, your name.
Subject: 150 restaurant groups opening a location
Hi {first_name},
I built Ladlepay a list of 150 independent restaurant groups opening a new location before {month}, each with the person who runs payroll and a checked email: {sheet_link}
{your_name}
There’s no request for a call in it. The sheet does the selling, and asking for a meeting turns a gift into a trade.
How to send it:
- To one person who owns new business: the head of sales, the VP of growth, or the founder at a small company.
- From your normal inbox, written and sent by hand. One email to one company.
- Without a calendar link. I’d also turn off open tracking for this one.
- With one follow-up at most, a week later, and only with something new: “Added 12 groups that announced openings since.”
If they reply, that’s where the offer goes. Ask if they want the list refreshed every month, or built for a second signal like groups hiring a new general manager.
9. Read ten rows by hand before you send
The reader judges the whole list by the first few rows they check, so check them first. These rules catch the problems that ruin a list:
- Every row has proof. Say a group is listed as “opening in Austin” and its website says nothing about Austin. The reader stops trusting the other 149 rows.
- The signal is still true. An opening from eight months ago is old news. Check the time window again the day before you send.
- The person is the right one. The general manager of one location doesn’t run payroll for the group. Someone who left in spring is worse than an empty cell.
- No current customers. Listing the dream company’s own customers tells them you didn’t look at their website.
- One signal per list. Mixing opening rows with hiring rows to reach 150 makes the list look like a database export. Send 110 good rows instead.
- One row per company. Three locations of one group as three rows looks like padding.
- A keyword match needs reading. A job ad that mentions “new location” in a standard footer proves nothing. Read the full text before it counts.
Then pick ten rows and check every column against its source. If a row is wrong, fix the rule that let it through and pick ten new rows. The list is ready when three rounds of ten in a row come back clean.
A good score from your email verifier doesn’t replace this. Say it marks 98% of the emails valid. The list can still contain the row where the “owner” is the pastry chef, and that’s the row the reader will click.
Where to start today
You can build your first give-first list in three steps:
- Pick one company you want as a client and turn its ICP into four tests a stranger can check.
- Pick one signal you can prove and find every company that shows it now, with the right people and verified emails.
- Put the list in a shared Google Sheet and send the one-sentence email.


