Start from the account list, not from a topic. For each target account, find the handful of people who post, watch them and the conversations they join, and record which roles on the buying committee you have no visibility into. The gaps matter as much as the signals.
Disclosure: Signl is ours. It groups collected people by company, which is the mechanic this article describes. The method works with a spreadsheet and a browser at fifty accounts; what tooling buys you is not missing things.
Key facts
- Account-first inverts the usual listening setup, and the inversion is the whole method.
- Only a minority of any company's staff post publicly. Plan for the silence.
- Watch people, not company pages. Company pages are the weakest surface available.
- A coverage map of who you cannot see is more actionable than another list of signals.
- This caps out around a couple of hundred accounts, and saying so is part of the method.
Why account-first is a different setup
Ordinary social listening starts with a source and sees who turns up: watch a competitor, watch a creator, watch a topic, and whoever comments is who you get. The output is a stream of people, and the account they work for is an attribute.
Account-based listening runs the other way. You already know the companies. What you do not know is which humans inside them are reachable, what they care about, and whether you are watching the people who will actually decide.
That inversion changes three things at once. The unit of measurement becomes the account rather than the person. Silence becomes meaningful, because an account producing nothing is a gap you can see rather than an absence you never noticed. And coverage becomes a thing you can audit, which is the part that makes this worth doing at all.
Step 1: size the list honestly
Before anything else, decide how many accounts this is for.
Each account needs two or three named people watched, and someone reading what those people say. At fifty accounts that is manageable by one person. At two hundred it needs tooling and a daily habit. At a thousand it is not account-based listening, it is topic listening with an account filter on top, and you should set it up that way instead.
The usual failure is a list built for a different purpose. A target list of 800 logos assembled for ABM advertising does not convert into a watch list, because the attention required per account does not scale the way impressions do.
Step 2: find the people who post
For each account, the question is not who the buyers are. It is which of them say anything in public.
Three places to look, in order:
- The function you sell into. Not the whole team, the two or three who have written anything in the last quarter.
- One level up. The person who signs it off is often quieter, but their occasional post tells you the priorities the whole function is working to.
- Adjacent functions who complain about your problem. Operations people complaining about a finance process is one of the most reliable ways into a finance deal.
Expect most names to produce nothing. A minority of professionals post at all, and that fraction is lower in regulated and traditional industries than in software. Work with the ones who do, and record the rest.
Do not watch the company page. Corporate accounts post announcements, the comments are employees and partners, and the signal density is the lowest of any available surface. The founder's or a department head's profile is worth ten of them.
Step 3: the four surfaces per account
| Surface | What it gives you | Effort |
|---|---|---|
| Named individuals' profiles | Their own posts, and everything they comment on elsewhere | Low. The core of this method |
| Mentions of the account's name | Other people discussing them, including departures and problems | Low, and underused |
| Their competitors' threads | Your target's staff evaluating in public | Moderate |
| Their staff's posts in aggregate | The whole company's voice, useful for larger targets | Moderate |
The second row is the one teams forget. Watching for mentions of a target account's own name surfaces conversations about them that they are not in, which is where you hear about a reorganisation or a tooling change before anyone announces it.
In Signl these map to source types: a profile, a mentions search and a company staff search. A search counts as 2 sources against the plan's 25, so a serious account programme is a custom arrangement rather than the standard plan, and it is worth working out the arithmetic before you start.
Step 4: the coverage map
This is the artefact that makes the programme auditable, and the reason to run it account-first.
For each account, grid the roles you need against what you can actually see:
| Role on the committee | Posts themselves | Engages with others | Not visible |
|---|---|---|---|
| Economic buyer | |||
| Day-to-day owner | |||
| Technical gatekeeper | |||
| Blocker or incumbent owner |
Fill it in once per account. Three things fall out of it immediately.
Where you are blind. An account where only the day-to-day owner is visible is an account where you will hear about the problem and never hear the objection. That is a reason to work the account another way, not a reason to watch harder.
Who to ask for. A row with nothing in it is the introduction to request from the person who does engage.
Which accounts to drop. An account with four empty rows is not an account-based listening target. Move it to the list motion and stop pretending.
Signl's company view does a related thing from the other direction: it groups the people it has collected by company and names the roles in your ICP that nobody at that company holds yet, which is the same gap seen from the data rather than from the plan.
Step 5: what to do with a signal
The rules are the ordinary ones, with one addition specific to account work.
Where the person engaged decides what you may say: on your own post you may reference it, on a creator's you may mention the topic, on a competitor's you say neither. That is covered in how to message someone who commented.
The addition: treat several people from one account as one event, not several leads. Three people from Acme engaging with the same topic in a month is an internal conversation you can see the edges of. Messaging all three makes you look automated. The better move is often to write to the person who said nothing, using what the others said as the reason you knew to ask.
Measuring it
Per account, not per person, and four numbers:
- Accounts with at least one person visible, against accounts on the list. This is coverage, and it is the number that tells you whether the programme is real.
- Accounts that produced any signal in 90 days.
- Accounts where you started a conversation.
- Accounts that progressed.
The first number is the one to watch in month one. A programme covering 20 of 80 accounts is not failing to convert, it is failing to see, and those need different fixes. The full attribution model is in how to measure social listening ROI, and the records model that makes per-account reporting possible is in LinkedIn comment tracking.
Limits
- Coverage will be partial and that is structural. Most people at most companies post nothing. This programme supplements the others; it does not replace them.
- Signals cluster in the wrong seniority. Individual contributors post more than executives, so you will often see the user and not the buyer.
- Big accounts generate noise. At enterprise size, watching staff posts in aggregate returns a lot of recruiting and brand content.
- It is slow to show results, because you are waiting for a fixed set of people to say something rather than casting wider.
- The compliance position is unchanged by the account being a target. See GDPR and B2B prospecting, and note that LinkedIn's User Agreement section 8.2 restricts automated collection (read 25 October 2026). This is not legal advice.
