How we measure LinkedIn market engagement
Our market reports measure public LinkedIn engagement within a defined sample and time window. This page explains the audience selection, calculations and limits so you can interpret each table.
How to read these reports
Our market reports measure public LinkedIn engagement within a defined sample and time window. This page explains the audience selection, calculations and limits so you can interpret each table.
The reports cover different measures. Revenue tooling and ecommerce customer-service tables show reactions per post. The AI agent platform report compares role groups.
What a market is here
For these reports, a market is a sample of people engaging with selected vendors' public LinkedIn posts during the stated window. We filter that pool by role and, for some reports, select its most active members.
Role filters are a proxy for relevance. They can include people with no purchasing responsibility and exclude people who influence a purchase. The reports do not represent every buyer in a category.
What we count
Market reactions per post = reactions and comments from the sampled people on a vendor's public posts, divided by the vendor's post count in the same window.
The numerator counts engagement actions, not unique people. One person can contribute more than one action. The total therefore does not measure reach, and the average is not an engagement percentage.
Zero means no qualifying reactions or comments were recorded from the sample in the window. It does not mean the posts had no impressions. The post count is needed to interpret the average and should accompany it.
The AI agent platform report uses a different measure: builders and practitioners per service owner or lead, based on profile titles. Its ratios should not be compared with reactions per post.
What we do not count
- Impressions. We cannot see them on somebody else's posts and neither can you.
- Follower counts. A follower is not a reaction.
- Engagement from outside the sampled market. A vendor with a large US audience scores only on reactions from the market we defined.
- Private profiles, private groups, and anything that is not public.
- Attendance. We know which events sit inside a market's most-reacted posts at company level. We do not know who is in the room and we do not claim to.
Why the 170 most active
The revenue-tooling report uses the 170 most active people from a role-filtered pool of 642. The ecommerce customer-service report uses the 170 most active from 264. Activity is measured within the category and time window.
This selection favours people who use LinkedIn frequently. Quiet readers and people outside the selected vendor accounts may be missing. The results describe the visible sample, not a random sample or population estimate.
A result of 29 of 170 describes 17.1% of that sample. It cannot be extrapolated to the full category without a different sampling design.
Unique people engaging with an account
Creator overlap counts the unique people in the stated sample who engaged with a public creator or institution during the window. Repeat engagement with the same creator does not create additional people.
Overlap does not measure followers or impressions. The same person can engage with multiple creators, so adding two creator counts can double-count people. A combined total requires deduplication.
Public accounts are named with their roles as recorded for the report date.
Event and community research
We identify events and communities mentioned in relevant public posts and check the category calendar. These findings help assess category and company relevance.
They do not establish individual attendance, exclusive community membership or the number of people an event campaign will reach. If no event stands out in a sample, the report says so.
Naming rules
- Vendors: named. They are public companies and every number comes from their public posts.
- Voices: named, with their role. Public creators and institutions, measured on public posts.
- Commissioning companies: Not identified without explicit permission. Report regions describe the research sample, not the identity or location of a customer.
- Private people: never named. The people inside a sampled market reacted to a post. We do not name them and we do not publish the list.
Postelix is operated by Unfair Advantage Ltd in Cyprus. See /privacy for the company's data practices.
Where the method is weak
90 days is short
A category with a slow publishing rhythm looks quieter than it is.
One pull is one pull
Every table is a snapshot of a window, not a trend. Two pulls a quarter apart can disagree.
Title filters miss people
Every table includes people who cannot buy and excludes people who can.
Reactions are not intent
Somebody reacting to a competitor's post is paying attention, and attention is not pipeline.
Public only
What happens in private communities and at events is outside anything we can count.
Regional pulls do not generalise
A Nordic revenue tooling table says nothing certain about the same category in the US.
Tell us if a number is wrong.
If a number here is wrong, tell us and we will fix it and say what changed. That includes a vendor who thinks their post count is off and a voice who wants their role corrected.
Write to help@postelix.com. When we change a published number we say what it was before and when it changed.
Questions
We start with the people who reacted to or commented on the posts of the vendors in a category during the window. That pool gets filtered by title and ranked by how often each person engages. The tables use the most active slice, and when we run this for a company, that company strikes out the fits that are wrong.
Total reactions rewards posting volume. A vendor publishing four times a week will out-total a vendor publishing twice a month even when each individual post lands worse. Dividing by posts in the window makes two vendors comparable, and we print the totals as well so you can read both.
The vendors in these tables are public companies and every number comes from their public posts. Nothing here comes from a private profile or a private list. We never name a person who reacted, and the companies whose readouts produced these numbers stay anonymous unless they ask to be named.
Yes. We need the category and the vendors you compete with, then we build the named market and run the same count over 90 days. The output is the table, plus the voices the market follows and the rooms it gathers in.
Book a demo. Then run it on your own market for 7 days.
The demo is with the founder and shows the loop in action. A work email is enough, no card.