Companion to a field report built with this method. Run it every three to six months.
Almost everything LinkedIn tells you about your own performance is misleading. Not maliciously - the dashboard is built to make posting feel good, and it does that well. But if you want to know whether any of this is working, the numbers on the screen will not tell you.
Three problems, all fixable in an afternoon:
Every export is capped at 50 posts. Ask for twelve months and you get your best fifty, with no warning that anything is missing. Ask month by month and you get everything. My one-year export overstated my median reach by 90% against the same data pulled monthly.
Impressions are the headline metric and the least useful one. In my data, reach and engagement move in opposite directions.
Nothing joins to anything. Post performance lives in one place, followers in another, website traffic in a third. The interesting findings are all in the joins.
Here is the whole method.
What you need
- LinkedIn analytics exports, one per month. Not the twelve-month one. Each file contains the post data and the follower data as separate tabs, so this is one download per month, not two.
- A list of your posts with categories, if you have one. A content calendar, a spreadsheet, anything with a category per post.
- Your website's server logs, if you want the traffic half. Analytics tools only see visitors who accept cookies; server logs see everyone.
No paid tools. A spreadsheet will do most of it; a short script does it faster.
Step 1: export month by month, not year by year
This is the step that matters most, and the one everybody gets wrong.
Go to your profile, then Analytics, then the content section. Set the date range to a single month and export. Repeat for every month you want to cover.
The cap is 50 posts per export, not per account. Over a year it silently removes most of your posts, and because the export returns your top performers, the ones it removes are the weakest.
Monthly is not automatically safe either - if you published more than 50 posts in a single month, that month is capped too. Check the row count against what you actually posted. If it comes back at exactly 50, split that month in half and export twice.
Same tool, same account, same period - the date range decides whether the data is complete.
I exported eight months separately and got 135 distinct posts. The twelve-month export of that identical period gave me 50, and the 85 it dropped were precisely the half that tells you the truth.
Each file has six tabs. Four of them matter:
- ENGAGEMENT - daily impressions and engagements
- TOP POSTS - per-post URLs, dates, impressions and engagements, in two side-by-side blocks
- FOLLOWERS - daily new followers, plus your total on the export date
- DISCOVERY - headline impressions and members reached for the period
The other two, AUDIENCE DEMOGRAPHICS and CONTENT DEMOGRAPHICS, describe who is seeing you rather than how you performed. Useful, but not part of this audit.
Step 2: read the follower tab
No second download needed - the FOLLOWERS tab is in the file you already have. One row per day with new followers, plus your total on the export date.
Two things to know before you use it.
The daily numbers are gross additions, not net. Mine summed to 148 against a stated total of 139 - nine people unfollowed, and the data does not say when. Build your audience curve forward from zero using the daily adds. That gives you the audience that actually existed when each post went out, which is what you need.
And it only covers the export window. Anyone who followed you before it started is not in the daily rows.
Step 3: get your posts with categories
If you keep a content calendar, export it. You want, at minimum: the post text, the publish date, and a category.
If your calendar has several category fields, check which one is actually filled in. Mine has two: one was blank on a third of posts, the other on none. Use the complete one; a category analysis with 40% missing data proves nothing.
If you have no calendar, this step is optional - you can still do everything except the category breakdown. But it is the analysis that tells you what to write more of, so it is worth starting one now for the next audit.
Step 4: join posts to categories
LinkedIn's post URLs contain a slug built from the opening words of the post. That is your join key.
linkedin.com/posts/yourname_my-kids-are-growing-up-in-three-languages-share-748...
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^Normalise both sides to lowercase words, strip the trailing activity ID, and match the slug against the first dozen words of each post body. Accept a match above about 60% word overlap.
Do not try to join on the post ID. In my calendar only 59 of 135 rows had a stored URL. Content matching got 129 of 135.
Spot-check ten matches by hand before trusting the rest. The slug comes from the post's opening line while your calendar title may be the article headline, so correct matches often look different at a glance.
Step 5: remove the outliers before computing anything
One post will dominate everything. Mine was 58% of my entire year's impressions and 254 times the median of the rest.
Compute your statistics twice: once with everything, once with the extremes removed. The trimmed set is your real baseline. The full set tells you how concentrated your results are, which is its own finding.
A reasonable cut: drop anything more than about ten times your median. Say which posts you dropped and why.
Step 6: compute the six things worth knowing
1. Your real distribution
Median, quartiles, minimum. Not the mean - one viral post makes the mean meaningless.
The number to hold onto is the median, and the 25th-to-75th percentile range. That is what a normal post does for you.
2. Concentration
What share of your total reach came from your single best post? Your top three? If one post is more than half your year, every average you have ever computed about yourself is wrong.
3. The reach-to-audience ratio, over time
For each post, divide its impressions by your follower count on the day it was published.
This is the most valuable number in the audit and nothing in LinkedIn shows it to you. Track it monthly. If it is falling while your audience grows, the platform's early-account distribution subsidy is being withdrawn - which means reach will no longer grow on its own.
4. Engagement rate by reach band
Group posts into reach bands and compute engagement rate for each.
I expected higher reach to mean more engagement. The opposite was true: my posts under 120 impressions engage at 2.58%, over 200 at 1.08%. If that holds for you, impressions are actively misleading you about which posts worked.
5. Performance by category
Median reach and engagement rate per category. Rank by engagement rate, not reach.
Watch for the category with high reach and low engagement. That is the one the platform likes and your readers do not, and it is probably occupying slots you could use better.
6. What a big post is actually worth
Take your best post. Count followers gained in the fortnight before and the fortnight after. The difference is roughly what that post bought you.
Then check your server logs for the same days. Impressions are not visitors, and the gap is usually enormous.
Step 7: the website half
If you want to know whether any of this produces traffic, use server logs, not an analytics tool.
Google Analytics only sees visitors who accept the cookie banner. Server logs see everyone. On my site the two disagree by roughly an order of magnitude, and the logs are right.
Two cautions from doing this badly first:
Filter crawlers properly. A naive "browser fetched the JavaScript bundle" test counts Googlebot, bingbot and headless browsers as people. When I first ran this, 29 of 38 "human visitors" in a day were HeadlessChrome. Check the user agents.
Your own visits are in there. If you browse your own site regularly - and you do - some share of every daily figure is you.
Compare your biggest LinkedIn day against site sessions that day. That single comparison is usually the most sobering number in the whole audit.
Step 8: write down what you will change
An audit that does not change what you do next was entertainment.
Pick a small number of specific changes, and state a prediction you are willing to be wrong about. Then re-run this in three to six months and check whether the prediction held.
That last part is what separates a measurement from a dashboard.
What this cannot tell you
It is one account, and your results will differ. It measures correlation, never causation - if two things changed in the same week, this method cannot separate them, and you should say so rather than pick the flattering explanation.
It also says nothing about whether the audience you are building is the right one. Engagement rate tells you people responded. It does not tell you they were people who might hire you.
That question needs a different instrument, and I do not have a good one yet.
I ran this on my own account and published everything it found, including the parts that make me look worse: I Rebuilt My Network From One Connection. Here Is the Data.