Most personalized cold email is a first name, a company name, and an adjective about growth, and the people reading it can spot that pattern before the second sentence. What makes one land is a single true fact about the reader, found before anything is written, and a reason to write that follows from it.
That fact has to come from somewhere. Most outbound stacks find the lead and write the message in four tools that never share a list: a prospecting tool exports a CSV, an email finder fills in part of it, somebody reads company websites in a spare tab, and the export of enriched leads to a sequencer carries a file that knows less than the rep did an hour earlier.
This guide shows annotated examples first, then walks one list from search to send. If you only want the search half, the Find leads page covers filters and saved searches in more depth.
What makes a personalized cold email
A message has to pass three tests:
It could only go to this person. Swap in another company's name and it stops making sense.
The fact is checkable. The reader can tell within a sentence whether you looked, because the fact is theirs: a role they posted, an office they opened, a job they moved into.
The fact leads somewhere. It connects to the reason you are writing. A compliment about their podcast that goes nowhere is trivia.
Doing this for one prospect is easy. Doing it for four hundred is where teams fall back on templates, and the fix is to split the work. You write what you're asking for once. AI writes why you're reaching out to that person, fresh for each contact, from that contact's research.
What software can find leads and write personalized outreach?
Oneprofile does both from one list. Describe who you sell to in plain words and it sources matching leads, fills and verifies their work emails, and researches each company and person. The outbound agent writes each contact's email from that research, and email and LinkedIn sequences send it.
To walk your own list through it as you read, describe your buyer in a free workspace and watch the match count before you spend anything. Free comes with 200 credits a month, 3 tables, and 100 rows per table.
Personalized cold email examples, annotated
Each example pairs one research fact with the opening line it produced. The companies are invented, on reserved domains such as cedarcreative.example, so read them for the pattern rather than the wording.
Research fact, and where it came from | The line it drove | Why it lands |
|---|---|---|
Cedar Creative has three account manager roles open, from its careers page | "Three account manager roles open at once usually means the client list grew faster than the team." | They posted the fact themselves, and the inference is easy to confirm or correct |
Thistle Advisory opened a Leeds office in August, from its own news page | "Saw you opened the Leeds office in August, so the client handoff now runs across two cities." | Recent, specific, and tied to the problem you solve |
Pebble Studio's head of operations moved over from agency work in the spring, from her LinkedIn work history | "Six months into running operations after years on the agency side, you have probably rebuilt the handoff once already." | It is about her job, not her company's homepage |
Copper Studio's team tried your product last quarter and invited two colleagues, from your own product data | "Your team tried us last quarter and two colleagues joined, then it went quiet." | Only you can know this, so no other sender can write it |
Redwing Advisory was listed on a startup launch board last week, from a launch search | "Congratulations on the launch last week. The first month after one is usually when the onboarding emails pile up." | It reacts while the news is still news |
Merge fields only | "I saw [Company] is growing fast." | It fits every company, so the reader learns nothing and can tell |
Here is the Leeds example as a whole email. Say you sell project software to advisory firms:
Subject: Leeds and the client handoff
Hi Priya, saw you opened the Leeds office in August, and two project coordinator roles are open on your careers page. A second office usually means the client handoff gets rebuilt in two places at once.
Our project software keeps one client record across offices, so a coordinator in Leeds sees what the first office promised. Worth fifteen minutes next week?
Where each part came from:
The first two sentences come from research: the office from the company's news page, the roles from its careers page. They were written for Priya alone.
The offer and the fifteen minutes come from you. The step's prompt and the sequence's goal say what the email is for, and every contact gets the same request.
The subject is written per contact too, unless you set one subject for everyone.
Made-up facts get marked. The agent is told not to invent anything, and a check marks any figure, link, or address it could not find in the research or your business context.
The Copper Studio example works differently, because no outside source holds the fact:
Subject: Two colleagues, then quiet
Hi Sam, your team tried us last quarter and two colleagues joined, then it went quiet. That usually means the first project stalled before anyone saw a result.
We could set that first project up with you in twenty minutes, and you keep it. Would Thursday work?
The opening sentence came from your own product data, pulled in beside the research, so it is the one line in the email nobody else could send.
A complaint while we are here: readers learned to spot first-name personalization years ago, and outbound kept sending it anyway. Forty emails that mention something true beat four hundred that mention a first name.
Research that feeds a personalized cold email
Every example above shares one thing: the fact existed before the email did. Getting there for a whole list takes seven stages, and tools that cover one stage each tend to lose the fact in the handoff between them.
Stage | What has to happen | Where it usually breaks |
|---|---|---|
Describe | Turn "who we sell to" into a search | Filters nobody can read back |
Source | Pull the matching people in | Exported as a file, then forgotten |
Enrich | Fill work email and company details | One data source, half the list blank |
Verify | Confirm each mailbox exists | Skipped, then the domain pays for it |
Research | Learn something true about each company | One browser tab per company, by hand |
Draft and send | Write the line, run the follow-ups | Merge fields posing as research |
Sync back | Put what happened on the CRM record | Nobody updates the CRM |
Everything up to the draft happens before a word is sent, and the steps below take it in order. Sending and syncing back come after, under the next heading.
Describe who you sell to
Start with a sentence, not a filter panel. Something like "heads of operations at logistics companies in the UK with 50 to 200 employees" is enough. The AI Assistant turns it into a role, an industry, a location, and a headcount range, and every filter it set is marked so you can read it back and change it.
Counting the matches is free, so you see the number before a credit is spent. That number is worth pausing on. Twelve matches means the description is too narrow, and forty thousand means you described a market rather than a buyer.
Save the search when it looks right. Once your list is built from it, the list can run that search again on a schedule. Each run counts the new matches, prices them, and waits for your approval before adding anyone. That matters later, because the sequence enrolls from that same list.
Source leads and fill work emails with a waterfall
Add the matches to your list. You can also start from your CRM, a CSV, a Google Sheet, or another of the 144 integrations, and everything below works the same way. A list imported from a connected tool keeps a link back to it, so leads added to the sheet later arrive with the next refresh.
Work email is where most lists go thin. No single data source covers everyone, so the waterfall asks data sources one at a time, most accurate first, and stops at the first one that returns an answer:
Most accurate answer first. Later data sources run only for the contacts the first one missed.
A work email lookup that finds nothing costs nothing. You pay for the source that answered.
The price before the run. Every enrichment shows the most a lookup can cost before you start it.
Tick Enrich company in the same pass and each lead's company arrives with its industry and a description, beside the job title the search already found. The bulk email finder guide covers the ordering in depth.
Verify every address before the first send
Verification is its own step beside the finder, not a second tool afterwards. Add Verify email next to Find work email, and every address gets a Deliverability verdict:
valid. Send to it.
catch-all. The domain accepts everything, so the check cannot confirm this person. The second verifier you can switch on settles many; send the rest in a small, separate batch.
risky. Hold these with that batch.
unknown. No source reached a verdict, often because a mail server refused the check for a while. Verify it again later.
invalid or disposable. Leave it out.
This matters more for a founder than for anyone. A new sending domain has no reputation banked to absorb bounces, and Google's email sender guidelines set out the bar every domain sending to Gmail is held to. A bounce-heavy first send is hard to undo on a domain that young.
Research each company and draft the line
This is the step a merge field cannot fake, and it is where every example above came from. Four built-in research agents do the reading:
Company researcher reads the website and LinkedIn page: what they sell, who they sell to, and how they talk.
Person researcher reads the LinkedIn profile and recent activity, and returns career history and a hook for that one person.
Hiring signal finder counts open roles and says which departments are growing. The hiring signals guide shows how to build a list around them.
Use case finder names the one workflow your product would change, in their terms.
The person researcher and the hiring signal finder each return a hook: the most specific true thing anyone found about them. The first touch opens on it. If the company is already in your billing, support, or product systems, Look up your own data pulls that record in beside the research. That is where the Copper Studio line came from.
Then the outbound agent writes each contact's email from that research, the step's prompt, and your business context. It writes when a contact's first message falls due, drafting the follow-ups in the same pass so each one builds on the last. Before anything goes out, the AI writing tab writes one contact's whole sequence in order and sends nothing, so you read what Priya would get.
Send a personalized cold email as a sequence
The sequence enrolls from the list you just built, so there is no export and no second copy of anyone. How many follow-ups to plan, and how far apart, is its own question: the B2B cold email sequence guide lays one out day by day.
Run email and LinkedIn sequences from the same list
With Keep in sync on, new reachable contacts join on their own, and contacts who leave the audience stop. One sequence holds both channels:
Email steps send from your Google Workspace mailboxes. Each mailbox paces its daily budget across the day, volume spreads across as many mailboxes as you add, and reply and bounce rates are reported per mailbox.
LinkedIn steps cover profile views, connection requests, and messages. They run through a Chrome extension in the rep's own signed-in browser, not from a data center.
Waits and branches sit between steps. Branch on a reply, a bounce, or an accepted connection request, so someone who accepted on LinkedIn does not get the cold email version.
Whether a message goes out on its own is the sender's send mode. On Send when confident, it goes when the agent rated it high and had enough data, and the rest wait in Tasks for a person to approve.
The Outreach page covers mailboxes and the step types in detail. One thing setup does not cover: a brand-new mailbox still needs its daily limit raised a little at a time before it sends at full volume.
Write what happened back to your CRM
Replies from every mailbox land in one threaded inbox. A reply stops the sequence for that person, and a reply asking not to be contacted again adds them to your do-not-contact list and stops every sequence they are in. Out-of-office notices and bounces are filed apart from real replies.
Every email and LinkedIn message a sequence sends is kept with the contact and exports with the rest of the list. Map it to the CRM fields you want filled, and create any field that does not exist yet in the CRM first.
Limits of this outreach setup
A few things this setup does not do, stated plainly so you find them here rather than in week two:
Google Workspace only. Sending works from Google Workspace mailboxes, connected one at a time or provisioned in bulk. Microsoft 365 and generic SMTP or IMAP mailboxes cannot be connected. If your company runs on Outlook, outreach goes out from a separate Google Workspace domain.
No open or click tracking. This is deliberate. Oneprofile adds no tracking pixel and rewrites no link, and a sequence sends plain text unless you turn it off. Replies are the number that counts.
No A/B testing of sequence steps. If you want to compare two openers, run them as two sequences on comparable slices of the list and compare replies yourself.
Research is only as good as what is public. A six-person company with a one-page website gives the agents very little to work with. Without a real fact, the email argues from what you sell rather than from anything true about them.
What it costs to find leads and write personalized outreach
One credit meter covers adding leads, enrichment, research, writing, and sending. Each enrichment and agent shows its price before it runs, and an enrichment lookup that finds nothing costs nothing. An AI research step is different: it pays for the research it did, answer or not.
Plan | Price | Credits |
|---|---|---|
Free | $0 | 200 a month |
Team | From $20 a month | 1,000 a month |
Pro | $200 a month | 5,000 a month |
More credits are $20 per 1,000 on any paid plan, credits on paid plans never expire, and every paid plan has unlimited seats. Free has 1 seat, up to 3 integrations, 3 tables, and 100 rows per table. For the wider picture of running this every week, outbound sales in one tool follows the same list from first search to booked meeting.
The fastest way to judge the research step is to read it on companies you already know. Describe your buyer, pull twenty matches, and open the research. Get started free and read what it finds on your own market.

