How do I compare Facebook, Instagram and LinkedIn performance fairly?

Compute a baseline for each metric on each platform, using the median across the competitor set you track, then divide every account value by its platform baseline. The resulting index has one unit, so 1.20 on LinkedIn and 0.90 on Instagram are directly comparable. Keep the raw counts visible alongside the index, because the index describes relative standing and says nothing about scale.

Can I add up followers or engagement across all channels?

No. Summed followers double-count the people who follow you in two places and imply an audience you do not have. Summed interactions are worse, because they are unlike units: the total is dominated by whichever platform produces the biggest raw counts. Report each channel separately, and use an index or a share-of-set figure when a cross-channel view is needed.

Is engagement rate a normalised metric?

Only within a single platform. Dividing by followers removes the audience-size difference between accounts, which is why it works for comparing competitors on one channel. It does not remove the difference in what an interaction costs a reader, so an Instagram rate and a LinkedIn rate remain incomparable even though both are percentages.

Should I use the mean or the median for the platform baseline?

The median, in almost every case. Interaction counts are heavily skewed: a handful of posts collect a large share of the total, and one of them landing inside your baseline period will lift a mean enough to make every account look like it underperformed. Our own benchmark is the demonstration. Recomputing an average of these counts month after month moves it by more than an order of magnitude, tracking whichever large account is in the sample, while the median over the same accounts holds a narrow band. That is why every interaction figure we publish is a median.

How often should the baseline be recomputed?

Quarterly is a workable default, and the important part is doing it on a schedule rather than when a number looks wrong. When you rebuild it, restate the whole historical series on the new baseline and record the date you switched. A baseline that changes silently makes every trend in the chart uninterpretable.

Does normalisation apply to ad activity as well?

Yes, and it matters more there, because ad libraries expose counts rather than performance. Active ad counts on different networks reflect different account structures, so a competitor running more ads on one network is not necessarily spending more. Index the count against the same network across your tracked set before comparing anything.

Updated July 29, 2026

10 min read

Analytics

How to Compare Metrics Fairly Across Facebook, Instagram, X and LinkedIn

A Units Problem Dressed Up as a Performance Gap

Short answer

Compare platforms by indexing each metric against a baseline from that same platform, never by putting raw counts side by side. In the Oppira Benchmark for June 2026, the median tracked account drew 58.6 likes per Instagram post and 19.2 per LinkedIn post. That is a units difference, not a performance gap.

Written byGabor Barta— Co-founder, Oppira

Gabor leads product and content at Oppira. He has spent over a decade building tools and writing about competitive intelligence, social media analytics, and growth marketing for B2B SaaS companies.

Published July 29, 2026

1. Why can I not compare likes across Facebook, Instagram and LinkedIn?

Because a like costs the reader a different amount of effort on each platform, and audience sizes differ. In the Oppira Benchmark for June 2026, the median tracked account drew 58.6 likes per Instagram post against 19.2 on LinkedIn.

58.6likes/post

Median Instagram like count for a tracked competitor account

Oppira Benchmark, median of per-account values across 51 tracked Instagram accounts, as of June 30, 2026.

19.2likes/post

Median LinkedIn like count for a tracked competitor account

Oppira Benchmark, median of per-account values across 33 tracked LinkedIn accounts, as of June 30, 2026.

Median per-post interactions across tracked competitor accounts, June 2026
Platform	Likes per post	Second interaction measured	Accounts measured
Instagram	58.6	0.8 comments	51
Facebook	13.1	1.49 shares	54
LinkedIn	19.2	1.54 shares	33

Median per-post interactions across tracked competitor accounts, June 2026 Oppira Benchmark, median of per-account values, as of June 30, 2026.Read down a column and the spread looks like a ranking of channels. It is not. X is absent from the table because the cell held too few posts from too few accounts to describe the platform, which is a collection gap rather than a finding.

0.36comments/post

Median Facebook comment count for a tracked competitor account

Oppira Benchmark, median of per-account values across 54 tracked Facebook accounts, as of June 30, 2026.

Notice that the ordering changes depending on which interaction you pick. Instagram leads on likes by a wide margin, but on shares LinkedIn and Facebook are level with each other at 1.54 and 1.49, and on likes LinkedIn sits above Facebook rather than below it. Any single raw metric produces a different ranking, which is the clearest sign that none of them is measuring the same thing.

Metric normalisation: Converting a raw measurement into a comparable form, usually by dividing it by a baseline drawn from the same context, so that values from different contexts can sit in one chart.; Also known as: Indexing, Rebasing

2. What exactly differs between the platforms?

Four things: what an interaction costs the reader, how large a typical audience is, which interactions the platform exposes publicly, and how much of the feed is paid. Any one of the four is enough to break a raw comparison.

The four differences that make raw cross-platform counts meaningless
Difference	What it means in practice	What it does to a raw comparison
Cost of an interaction	A double-tap on a phone is close to free. A public reaction on a professional network is attached to your employer.	Inflates the low-friction platform and suppresses the high-friction one, regardless of content quality
Typical audience size	The same company usually has very different follower counts on each channel, built over different periods.	Turns an audience-size difference into an apparent content difference
Which interactions are exposed	Saves, plays and shares are visible on some surfaces and hidden on others, and the same word means different things per platform.	Changes the numerator without anyone deciding to change it
Share of the feed that is paid	An account that boosts regularly is showing bought distribution in the same number as earned distribution.	Makes budget look like performance, and the effect differs per platform

The four differences that make raw cross-platform counts meaningless

None of these are fixable by choosing a better metric, because they are properties of the platforms rather than of the measurement. They are only fixable by comparing each platform against itself first, then comparing the results of that comparison.

3. Which metrics can be compared across platforms, and which cannot?

Cadence, direction of change and share of a tracked set travel across platforms. Absolute interaction counts, follower counts and engagement rates do not, because each is scaled by something platform-specific.

Cross-platform comparability by metric
Metric	Comparable as-is?	What to report instead
Posts per week	Yes	Report it directly. Cadence is measured in the same unit everywhere, so no adjustment is needed
Direction and size of change	Yes	Report percentage change against the same channel in an earlier period
Share of your tracked set	Yes	Report each account share of the set total, computed within one platform, then compare the shares
Likes or comments per post	No	Index the value against the median for that platform across your tracked set
Follower count	No	Report growth rate on each channel separately, and never sum followers across channels
Engagement rate	No	Compare within one platform only, and state the denominator every time
Format mix, such as video share	Within a platform only	Report each platform mix against its own history, since the available formats differ

Cross-platform comparability by metricThe pattern is that ratios computed inside one platform survive the trip and absolute counts do not. Anything you can express as "compared to the same channel last quarter" or "compared to the other accounts on this channel" is already normalised.

4. How do I normalise a metric so platforms can sit in one chart?

Divide each account value by the median for that metric on that platform across your tracked set. The result is an index where 1.00 means average for the field, and index values are comparable between platforms.

Five steps, repeatable in a spreadsheet, roughly an hour the first time.

Fix the baseline period. Choose one period, typically a quarter, and use it for every platform. A baseline built from different months per channel bakes seasonality into the comparison and cannot be undone later.
Compute a median per platform per metric. Take every account in your tracked set on one platform and find the median value for the metric. Use the median rather than the mean, because one viral post moves a mean enough to reset the whole baseline.Include your own account in the set, so your index is measured against the same field.
Divide every value by its platform baseline. An account with 45 likes per post on a platform whose baseline is 30 scores 1.50. The same arithmetic runs on every platform, and the resulting numbers share one unit.
Chart the index, label with the raw value. Plot the index so the channels are comparable, and print the underlying count in the tooltip or the row label. A reader who sees only the index cannot judge scale.
Rebuild the baseline on a fixed schedule. Recompute quarterly, restate the whole series when you do, and write down the date. A baseline that shifts silently turns a normalised chart back into an uninterpretable one.

Illustrative example: one company indexed against a per-platform baseline
Platform	Our likes per post	Platform baseline	Index
Instagram	320	400	0.80
Facebook	60	80	0.75
LinkedIn	45	30	1.50

Illustrative example: one company indexed against a per-platform baselineBaseline and account values here are invented for illustration. The raw column ranks Instagram far ahead. The index ranks LinkedIn first, because it is the only channel beating the field it competes in. Those two readings would drive opposite budget decisions.

5. Should I combine platforms into a single score?

Only if the weights are printed next to the score. An unweighted blend quietly hands the whole number to whichever platform produces the largest raw counts, so the score tracks one channel and claims to track four.

Take the illustrative numbers above. Summing 320, 60 and 45 gives 425 total interactions, of which Instagram is 320, or about 75 percent. A blended score built that way moves when Instagram moves and barely notices the other two channels, including the one that was actually beating its field.

If a single number is genuinely required, build it from indices rather than counts, and state the weight for each platform. An equal-weighted mean of 0.80, 0.75 and 1.50 gives 1.02, which says the portfolio is roughly average overall and is a defensible summary. It also throws away the finding that one channel is doing the work.

6. How do I handle a platform where I barely have any data?

Suppress the cell rather than publish it. A thin cell describes your collection gap rather than the platform, and once it appears next to well-covered channels it will be read as a performance finding.

We follow this rule on our own published aggregates. X is missing from every table on this page because the number of contributing accounts and posts sat far below our publication floor for the month. Publishing it would have produced a figure that looks like a platform benchmark and is really a description of what we managed to collect.

The practical version for an internal dashboard is to show the tile with an explicit "not enough data" state instead of a number, and to record how many accounts and posts each cell rests on. A reader can then tell the difference between a channel that is quiet and a channel you are not measuring properly, which are two findings that call for completely different responses.

Set the floor before you look at the numbers. A threshold chosen after seeing the result is a threshold chosen to include the result.

7. How do I present a normalised comparison without hiding the raw numbers?

Show both, always in the same row. The index answers whether a channel is beating its field, and the raw count answers how much attention the work actually bought.

Four rules that keep a normalised chart readable:

Print the raw value next to every index, in the label or the tooltip. An index of 1.50 on 45 likes and an index of 1.50 on 4,500 likes are not the same news.
Name the baseline in the chart caption, including the period it was computed over and how many accounts fed it.
Keep one chart per platform for the raw series, and use the normalised chart only for the cross-platform view.
Never mix an index and a percentage change on the same axis. They look alike and answer different questions.

Cadence is the exception worth reporting raw and unindexed, because posts per week means the same thing everywhere. In the Oppira Benchmark for June 2026, tracked accounts averaged 1.98 posts per week on Instagram, 1.68 on Facebook and 1.04 on LinkedIn, and those three numbers can sit in one chart with no adjustment at all.

Oppira collects the same fields daily across every channel you track, which is what makes a per-platform baseline computable from your own set rather than borrowed from a published average.

Key Takeaways

The cross-platform spread is a units problem

The median tracked account drew 58.6 likes per Instagram post and 19.2 on LinkedIn in June 2026, a gap no amount of content work would close.

Ratios travel, absolute counts do not

Cadence, percentage change and share of a tracked set are comparable as-is. Likes, followers and engagement rates are not.

Index against the platform median, not the mean

One viral post moves a mean enough to reset the baseline. The median across your tracked set is stable enough to divide by.

A blended score follows the loudest platform

In the illustrative example, Instagram was 75 percent of total interactions, so an unweighted blend tracked Instagram and ignored the channel that was outperforming.

Suppress thin cells instead of publishing them

A platform with too few contributing accounts reports your collection gap. Shown next to full channels, it will be read as performance.

Always ship the raw number with the index

The index says whether a channel beats its field. The raw count says how much attention it bought. Both are needed for a budget decision.

Frequently Asked Questions

Explore More

Related analyses, benchmarks, and industry insights

Related Guides

Followers, Reach or Impressions: Which Engagement Rate to Use

In-depth guide on this topic

What Competitor Data Is Public on Each Channel

In-depth guide on this topic

How to Measure Share of Voice Without Enterprise Tools

In-depth guide on this topic

Social Media Analytics for Business

In-depth guide on this topic