How to benchmark social media performance defensibly

The slide says your client is above average on engagement. Average of what, computed how, against whom, over which period. The client will not ask on Thursday. Their new head of marketing will ask in March, and by then nobody remembers who picked the four competitors or why the rate was divided by followers rather than reach.

This is how to benchmark social media performance so the number is still defensible in March. It needs a spreadsheet and about two hours, and the two decisions that matter are both made by a person.

TL;DR

Step one: choose the set, and write down why

The comparison set decides every conclusion for a year and it is usually written once, in a kickoff, by whoever was in the room. It is the most consequential and least audited input in the whole exercise.

1. Not aspirational. A brand ten times the size makes every chart depressing and every conclusion useless.
2. Not stale. A company that pivoted or stopped publishing drags the average down and looks like good news.
3. Include the one taking your deals. Often missing, because nobody in the kickoff had lost to them yet.
4. Cap it at about six. Past that the chart stops being readable and nobody opens it.

Write one line per competitor saying why they are on the list. Any entry with no reason is the first one to delete at the quarterly review, and the review should happen with the person who loses deals rather than the person who builds the report.

The set is chosen by a person and defended by a person. No product will tell you that the rival taking your customers is missing from it.

Step two: pick metrics that exist on both sides

This is where most benchmarks quietly break. A competitor has connected nothing to you, so anything the platforms reserve for the account owner is unavailable for them at any price.

Benchmark only on rows that are available on both sides. The moment a metric enters the table that only your own accounts have, the comparison flatters you by construction, and somebody will eventually notice.

Which figures across the whole tool category are measured and which are modelled is audited in best competitor analysis tools for social media.

Step three: fix the denominator once

Engagement rate is the benchmark metric everybody uses and the one two tools will disagree about for the same profile. The disagreement is always the denominator.

Divided by followers it is public on both sides and checkable by anybody. Divided by reach or impressions it relies on a number a competitor never published, so that half has been estimated. For a benchmark, use followers, and write the formula in the report footer. The full reasoning is in how to calculate engagement rate.

Step four: choose a period and never move it

Ninety days is the usual right answer. Long enough that one viral post does not decide it, short enough that a change registers within a quarter.

Whatever you choose, keep it. A benchmark computed over thirty days one month and ninety the next is a trend line describing your own method rather than the market, and that is the kind of error that survives for a year because it looks like a finding.

Use the median rather than the average across the set. One competitor with a viral quarter drags an average and tells you nothing about the typical account you are competing with.

Step five: capture with dates, because counts move

Likes, comments and views keep climbing after publication. A post read in week one and the same post in week eight are two different numbers, so a benchmark assembled by scrolling today cannot be reproduced next month.

1. Record the date of every reading. Without it the number is not checkable, only quotable.
2. Capture on the same weekday. Removes a surprising amount of noise for free.
3. Record follower counts alongside engagement. Absolute likes rising while the audience doubles is a decline.
4. Never backfill from memory. A gap is honest. An invented row is not.

This is the step that decides whether a benchmark becomes a series or stays a one-off slide, and it is the part worth automating first because it is pure counting.

Step six: end with a sentence, not a chart

A benchmark tells you where you sit, and sitting somewhere is not an action. The failure mode of this whole exercise is a beautiful monthly report that changes nothing.

So turn each finding into a question a person answers by looking. Why did their rate double in April. What did they stop doing in June. A dashboard can show the shape of a change and it will never tell you what the change was, because that is in the content rather than in the numbers. Budget ten minutes a month for actually reading what the top rival published.

Three patterns worth more than the averages

A benchmark produces a position, and positions are not actions. These three patterns are what a person spots in a series and no summary statistic shows.

1. A cadence break. A rival who published twice a week for a year and has posted twice in six weeks has usually lost a person rather than changed strategy. It is the best moment to move on their audience.
2. A format switch. A sustained change of format is a decision somebody defended internally. Understand it before copying it.
3. Followers rising while engagement falls. Almost always paid acquisition, or a viral moment that brought the wrong audience.

All three are visible from public data alone and none of them needs a reach figure, which is the argument of this page compressed: the questions worth asking about a rival are answerable, and the ones that are not answerable are mostly not worth asking.

The benchmark that is always available

Before comparing to anybody else, compare to yourself. It needs no competitor set, no denominator argument and no estimates, and it is the comparison a client actually acts on.

The third row is the one most teams never build and the one that explains the others. A quarter of weak engagement with a held cadence is a content question. A quarter with a missed cadence is a different problem entirely, and averaging them into one benchmark hides which you have.

A closing note on how often to do this. Monthly benchmarking sounds diligent and mostly produces noise, because ninety day windows overlap and the position rarely moves in four weeks. Quarterly is the honest rhythm for the comparison, with the underlying readings captured continuously so the quarter has something to compute from. Diligence is in the capture, not in the meeting.

One last note on scope. Benchmarking answers where you sit and it does not answer what a rival is doing week to week, which is a different exercise with a different rhythm and a different output. The method for that, including what to record and how often, is in track competitor social media.

How Groowth handles this

It does step five, which is the only step software should own. Paste the public handles of your set on TikTok, Instagram, YouTube or LinkedIn, and it reads each profile once a day and keeps what it read, with the date: posts and their timing, follower count, and public likes and comments.

Because nothing is connected, a competitor sits on the same board as your own accounts with identical columns, which is what makes the comparison like for like. It shows no reach, no share of voice and no industry average, because all three would have to be modelled for anyone but you. Three accounts free with no card.

Frequently asked questions