How to find warm introductions in your own network

To find warm introductions, download your LinkedIn connections and your message history, match the people in them against your target list by the company each one works at now, and rank every target by its strongest tie rather than by how many people you know there. Then add the networks of people willing to lend you theirs, take everyone you know off the cold list, and refresh the file twice a year.

The asking is a separate job, covered in how to ask for a warm introduction someone will make. This page is the step before it: working out who can introduce you, and to which of your accounts.

Why the search starts with people you know

A cold list answers a failed attempt with more volume: a bigger list, more mailboxes, another follow-up. When the account is one named company, you get one attempt, and a path through somebody who knows them is the route that still works at a list of one.

One founder's plan for their first forty conversations with buyers put people one hop away at the top of the channel order, at twelve to fifteen of the forty, as the highest yield per hour by an order of magnitude. The plan wrote its working rates down as assumptions, before anything was measured: about 4 to 8 cold emails in every 100 becoming a conversation, against 40 to 60 introduction requests in every 100. At those rates, forty conversations take 500 to 1,000 cold emails, or 70 to 100 introduction requests.

People who run outbound for a living report the same order. The staff of one outbound tool told its user community which of its plays drew the most replies. Second was messaging people who had reacted to or commented on a post, at an average reply rate of 11%. First was warm leads: the user's own exported network, their followers and the people engaging with their posts. One member asked for a way to filter by second and third degree connections, because of "a much higher connection rate when there's a mutual connection". Another, selling into a trust-based, high-ticket niche, put their outbound reply rate at 0.5%.

Download the two files

LinkedIn will give you your own network as files. Go to its data export page, request the full archive, wait for the email with the download link, and unzip what arrives. Two files in it matter:

• Connections.csv holds the people you are connected to, with their names, companies and roles. It says who you know.
• messages.csv holds your message history. It says how well you know them.

Use the export rather than an automation tool that reads your account. One warm-introduction product with enterprise customers builds its whole LinkedIn half from the same export: no API, no scraping, no browser extension. In one outbound tool's community, the members who automate LinkedIn write about an invite cooldown still active after two weeks, connection requests nobody accepts and the fear of a banned account.

Your mailbox holds the other half of your network. To judge a relationship you need only the sender, the recipients, the date and the subject of each email. The bodies add nothing to the count.

Keep the tally, then delete the messages

The messages file carries the full text of every conversation: who sent it, who received it, the date, the subject and the content. It is the most sensitive file you will handle in this whole exercise, and almost none of it is needed.

For each person, keep four things and nothing else:

Keep	What it tells you
Messages you sent them	whether you write to them, and how often
Messages they sent you	whether they write back, and how often
The date of the first	how long you have been in touch
The date of the last	whether the tie is recent

Then delete the content, the subjects and the conversation titles. The four columns are enough to apply every rule of tie strength: history in both directions is the signal, traffic in one direction counts for nothing, and a connection with no messages behind it is weak until something else says otherwise. Count the same four numbers from your mailbox and add them to the same row.

Match the people to your target list

Put your target list beside the connections file and, for each company, write down everyone who works there now, with their tally next to their name.

Rank each company by its best door, never by its headcount of contacts. The mistake to avoid is the one a count makes: forty connections at one company you have never exchanged a message with outrank the one person elsewhere who replies to you, and the second company is the one you can reach. Read each account as a count line first: how many people you know there, and how many of those ties are strong. Then read the names.

Run the match in both directions:

• From a target to your network. You have an account you want, and you look for who can open it.
• From your network to your targets. You list every company on your list that somebody you know can reach, best door first. This second list tells you which accounts to work this week.

Count only ties you can point to: a message thread, or a page that names you both. Having worked at the same company at the same time is a guess about a relationship, and an introduction request built on a guess asks a stranger for a favour. The one correction to make by hand runs the other way: the friend you see every month and never message has no tally at all, so move them up yourself.

Borrow the networks around you

Your network stops at you. Advisors, investors, board members, customers who champion you and founders backed by the same investors each know people you do not, and the method above works on their files as well as yours. One member of a prospecting tool's community asked for exactly this: a way to ask for warm introductions based on who their friends are connected with.

Make it easy to say yes. Ask them to run the same match against your list, or to share their connections file with the people or companies they want to keep private taken out first. Each lender keeps control of who you can see.

Keep the file alive

An export is a snapshot, and people change jobs. Download it again every six months. When you do, merge the new file into the old one rather than replacing it, so that a connection the platform has since dropped stays on your list until you remove them.

Take the people you know off the cold list

Everyone you know at a target account comes off the cold sequence the day the match finds them. Members of one outbound tool's community describe what happens otherwise: automation messaging people they already knew, and the apology that follows. Put them on the do-not-contact list, and ask through the door instead.

Then track every path you open by its stage: asked, said yes, said no, meeting booked. Within a few weeks the share of introduction requests that became conversations is your own number, and you can replace the plan's assumption with it.

Five things

1. Download the full LinkedIn archive this week, for the connections file and the messages file.
2. Turn the messages into four numbers per person, then delete the text.
3. Match your connections against your target list, and rank each account by its strongest tie.
4. Ask three people to lend you their networks, and let each of them hide whoever they want to keep private.
5. Take everyone you know off the cold list, and download the archive again in six months.