The single-signal sequence
What happened
29% of invites accepted, and 10% of messages answered1
Buyers can spot a sequence in the first line. And push the volume, and LinkedIn restricts the account.
You can’t win LinkedIn on volume (or silence).
Most teams either spam LinkedIn… or go on mute.
Neither works. Drumbeat works out what’s actually worth doing, for each person, every day.
Whoever told you to automate LinkedIn hasn’t done the maths.
Here’s what people tried. And here’s what happened.
Sales tried
The scraped list
What happened
1 in 10 messages answered, however many you send1
The same averages, just bigger numbers. And the more you send, the sooner LinkedIn restricts you.
Marketing tried
The silent team
What happened
86% of B2B professionals don’t post regularly2
Your buyers are reading. Your team isn’t there.
The AI post-a-day
What happened
Marked as AI slop.
People spot it in a line, and scroll on. It reads like everyone else’s, because it is.
- The LinkedIn average: a blend of published outreach benchmarks. How it’s derived.
- Drumbeat survey of 5,000 B2B professionals, manager to C-suite. The research.
It’s a maths problem. So we did the maths.
Take a team of ten, and 20,000 people in their market. Each person could invite, message, comment on or reply to any one of them today. That’s 800,000 things your team could do. They have time for about a hundred.
And that’s the simple version. Add every topic worth posting about, every moment to post it and every way to say it, and the options run towards infinity. Picking the right hundred by feel, every day, for every person, isn’t something anyone can do.
Mathematicians call this constrained optimisation: far more choices than capacity, and one best answer somewhere in between. So that’s how we treat LinkedIn. Drumbeat works out what every possible action is worth, and hands each person the few that matter most, as the day unfolds.
One signal isn’t a reason to act.
Signal-based tools run on a simple idea: someone changes jobs or likes a post, so you act. It’s a good start. But a signal on its own doesn’t tell you much.
What gives it meaning is context. Who you are, and who they are. Whether they fit your ICP, and whether they sit on a buying committee. How many people at their company your team already knows. What’s happening in their market this week.
Signals and context combine in ways that aren’t obvious: two weak signals in the right context can matter more than one strong one out of it. Drumbeat’s models weigh all of it together, for every person, every day.
What Drumbeat’s models predict.
These are a sample, not the full set. With all that context, for each person, every day, they predict:
- Which topics your buyers will respond to Matched to what each person actually knows, so the ideas get sharper every week.
- When each person’s post will travel furthest Their own best moment, learned from their own audience.
- Who’s likely to accept your invite And who’s most worth it if they do. Two different questions.
- Which words will land in your post Learned from what your own audience responds to, not from generic best practice.
- When to follow up, and when to stop Timed to the conversation, not to a sequence timer.
- Which wording will get your message a reply Measured against the replies real messages get, so the first one is worth answering.
The right tool for each job.
Language models are brilliant with language, so that’s what we use them for: reading what people write, and helping each person say what they mean, in their own voice.
But who’ll accept an invite, which post will travel and what a reply is worth are questions of probability, and a language model is a poor judge of probability. So Drumbeat pairs them with statistical models that predict outcomes, ranking models that weigh every option against every other, and business rules that keep everything safe and on brief. Each does the job it’s best at. Then a person makes the call.
-
Language models
Read the context, and write in each person’s own voice.
-
Statistical models
Predict what’s likely to happen: who accepts, who replies, what travels.
-
Ranking models
Weigh every option against every other, for the next ten minutes.
-
Rules and constraints
Pacing, eligibility, and whatever your team has ruled out.
-
Humans
Make the final call.
Under the hood
Each candidate action gets an estimate of what it is likely to be worth: classifiers score the parts that can be scored (will this person accept, will this post travel, will this message get a reply), and those estimates carry their uncertainty with them rather than hiding it behind a single number.
A ranking layer then orders the candidates against each other, because the question is never “is this a good action?” but “is this the best use of the next ten minutes?”. Rules and constraints sit on top: pacing inside what LinkedIn tolerates, eligibility, who has already been contacted, what the team has decided is off limits.
Models are chosen and kept on outcomes. If a method stops predicting well for your business, it stops being used for your business.
Human-first AI. Not AI-first humans.
“Everyone else is trying to automate LinkedIn. We automate the decision, then help you take that action as well as it can be taken. Think Iron Man, not the Terminator.”
Parry Malm CEO, Drumbeat Nobody wants a bot posting for them, and buyers can smell one. The answer isn’t less AI. It’s AI that starts from a real idea, and sounds like you.
-
It starts with your idea.
Slop is a bad idea, written well. Riff draws out what only you know, from your own work or what’s happening in your market, before a word is drafted. Once the idea is right, the writing is the easy part.
-
It writes like you.
AuthorDNA learns how you write from what you’ve already written, across 243 measures: rhythm, vocabulary, how you open, how you sign off. Every draft is scored against them before you see it.
-
It removes the slop.
Anti-Slop isn’t an AI marking its own homework. It learns what real readers actually call slop, strips those tells before you see the draft, and keeps adapting to your preferences, and everyone else’s.
Everything you do teaches Drumbeat something.
No one lives in a vacuum. Everything you do affects someone else, and what they do affects you. Drumbeat understands this, learns from it, and optimises your next best actions as a result.
Useful quickly. Smarter with use.
Drumbeat starts with your company, ICP, market and available signals. Then it adds your team’s own decisions and outcomes.
-
More people posting
6× more posts per week
Especially the subject-matter experts you wish would post more. From one post every six weeks to nearly one a week (0.16 → 0.99).
-
Networks that grow
+50% more likely to accept your invite
When a prospect has engaged with a colleague’s post: 58% acceptance against 39% cold. Deeper into your industry, prospects and customers.
-
More conversations
6 days to a rep’s first conversation
Median, from their first invite. 83% of reps have their first conversation inside 30 days.
-
More pipeline
10.6 conversations a month for daily users
Daily users average 10.6 conversations a month. What each one is worth is your number, not ours.
Less time on LinkedIn. Far more from it.
Drumbeat works out what’s worth doing, for each person, every day.