How to Run a Social Media Audit.

A social media audit is a structured test of the system that produces your social output, run to find the one or two things actually limiting growth.

It is not an inventory of what exists.

Most published audit advice gets this backwards. It hands you a template, walks you through follower counts, posting frequency and whether the bio has a link, and produces a tidy document full of observations. Everything in it is true. None of it is a decision.

A real audit works upstream of the posts. It tests positioning, packaging, cadence, workflow and measurement, in that order, and it finishes with a ranked list of constraints, because a team can fix one or two things well and twenty things not at all.

What a social media audit actually is

An audit is a diagnostic. Its job is to explain the gap between the effort going in and the results coming out, then name the specific thing that closes it.

Which means an audit has not finished until it produces four things:

• Evidence behind every claim: a number, a curve, a date or a count.
• A ranked list of constraints, not a list of everything that could be better.
• A named owner for each item at the top of that list.
• A date to re-test, so the audit becomes a measurement rather than an opinion.

A competitor analysis is not an audit. A brand review is not an audit. Both are inputs to one.

Why most audits change nothing

Because they audit the output. The posts are the only part of a social operation that is public, so the posts are what gets reviewed, and the review comes back with notes about the posts.

The posts are the symptom. A caption is weak because nobody agreed what the account is for. A launch missed its moment because approvals took nine days. A format got dropped because the report never showed it working.

Fix the caption and the same caption comes back next month. Growth stalls behind the scenes, which is exactly where the standard audit does not look.

The second failure is length. An audit that names twenty problems has named none, because the team will pick the three that are easiest rather than the ones that matter.

Observation, finding, constraint

Three different things get called an audit finding. Only one of them is.

• An observation is something you can see. You post four times a week. Your bio has no link.
• A finding is an observation plus evidence plus a consequence. Three of your four weekly posts are product announcements, and product announcements hold the lowest average watch time of anything you publish.
• A constraint is the finding that, once removed, moves the number most, and that nothing downstream can compensate for.

The test is a single sentence. If we fixed this, that would move, and here is the evidence. If you cannot write it, you have an observation, and observations are what fill a forty-slide deck.

The five layers, and the order you test them

Behind every post sit five layers: positioning, packaging, cadence, workflow and measurement. Audit them in that order.

The order is not a preference. Each layer’s data is only readable if the layer above it is sound. If positioning is unclear, your packaging data tells you which of the wrong things people clicked. If workflow is broken, your cadence findings are a picture of the queue rather than of the audience.

Measurement comes last because you cannot judge whether the right things are being measured until you know what the account is for.

Auditing in the wrong order is how a team spends a year rewriting hooks while the real fault sits two layers above them.

Layer 1: Positioning

Positioning is who the account is for and what they reliably get.

What to look at:

• The last 30 posts. Can you group them into three or four repeating subjects, or is it a list of unrelated things?
• The share of your reach coming from people who do not already follow you.
• Your single best post of the quarter. Could you make another one on purpose next week, or was it a one-off?

Signs positioning is the constraint: reach comes almost entirely from existing followers, the best performer is unrepeatable, and nobody on the team can describe the audience without naming three of them.

Layer 2: Packaging

Packaging is the title, the thumbnail, the first frame, the first line and the first three seconds. It decides whether the work inside gets seen at all, which is why it holds the cheapest wins in the audit.

What to look at:

• Opening drop-off on video. NBK treats under 25% as strong and 35% or more as broken, and broken means you stop everything else and fix the opening.
• YouTube’s audience retention report, where the intro figure tells you what share of viewers were still watching after the first 30 seconds.
• Your five best and five worst posts by views. Compare the openings, not the topics.
• Whether the account is even eligible to be recommended. Instagram’s Account Status exists to show you why your content may not be eligible to be recommended, and Instagram has said it reduces visibility for reels that are low resolution or watermarked, muted, bordered, majority text, or already posted on Instagram.

Layer 3: Cadence

Cadence is not how often you should post. It is whether the rhythm you have chosen is one the team can hold, and whether volume is being used to paper over a packaging problem.

What to look at:

• Output per week plotted against average views per post, across the last 90 days, on one chart.
• The gap between your best and worst post each month. A widening gap usually means the process has stopped filtering.
• How many planned posts actually shipped, and how many were swapped at the last minute.

Sign cadence is the constraint: output went up and views per post fell by more than the extra output earned back. That is a system publishing to fill slots.

Layer 4: Workflow

Workflow is the layer teams least expect to be audited, and the one that most often turns out to be the problem.

What to look at:

• Time from idea to live, measured on real posts rather than on the plan.
• Number of approval hops, and the hop where work waits longest.
• Whether the team could ship something today if the moment demanded it.
• What happened to the ideas that were killed. Read them.

Sign workflow is the constraint: the calendar is full of things that were easy to approve rather than things that work, and the best ideas are the ones nobody remembers approving.

Layer 5: Measurement

Measurement is judged on one question. What did the last report change?

If the answer is nothing, the reporting is decoration, however good the charts look.

What to look at:

• Whether the metrics on the report are the ones the platform actually optimises for. Instagram announced in August 2024 that views would become the primary metric across every format, and Adam Mosseri has pointed at views and sends per reach as the pair worth watching. YouTube says its recommendations run on viewing behaviour, likes and dislikes, subscriptions and satisfaction surveys.
• Whether anyone can name a decision that came out of the last three reports.
• Whether goals are written as numbers a person can act on this month.

When NBK runs a social audit, measurement is the layer that most often turns out to be hiding the others. A report that only shows totals will never show you a format quietly dying.

What to benchmark against

In order: yourself, the platform, then everyone else.

Your own baseline is the only comparison where you know the inputs. Same account, same format, previous 90 days. Everything else is a guess about someone else’s conditions.

External benchmarks are the weakest yardstick, and cross-platform ones are close to useless. Socialinsider’s benchmark study of 70 million posts put engagement per post, measured against followers, at 3.73% on TikTok, 0.48% on Instagram, 0.15% on Facebook and 0.12% on X for 2025. No single number for what counts as good engagement survives a spread that wide.

A competitor’s follower count is the worst yardstick of all. You can see their output and none of their inputs.

Check that your history is still comparable

Before you report a trend, confirm the metric meant the same thing at both ends of the line.

Meta’s own developer changelog introduced a views metric for Instagram and deprecated impressions and plays, applied across all API versions from 21 April 2025. Instagram had announced views as its primary metric across formats back in August 2024.

So a chart showing impressions falling off a cliff that spring may be recording a definition change rather than an audience change, and any tool reading the Instagram API inherited the same break.

This is not a footnote. A false trend in an audit is worse than a missing one, because a team will act on it.

How to rank what you find

Score every finding out of five on three things, in this order:

1. Size. How much of the gap between effort and result does this explain?
2. Confidence. How good is the evidence? A curve beats a hunch.
3. Control. Can your team fix it themselves, this quarter, without waiting on anyone?

Sort by the total and take the top two. Everything else goes on a dated list you do not open until those two have been re-tested.

Two tie-breaks are worth knowing. Prefer the constraint furthest upstream, because fixing a symptom while its cause stands just moves the bottleneck somewhere less visible. And change one thing at a time, or you will never know which change did the work.

What good looks like

A finished audit is one page, not forty slides.

Ranked constraints, three at most. For each one: the evidence, what should move if it is fixed, who owns it, and the date you re-test.

Underneath, a short list of what you checked and cleared. That cleared list is more valuable than it looks, because it stops the team relitigating the bio, the posting times and the hashtags for another quarter.

The item-level sweep sits alongside this rather than inside it. NBK’s free social media audit checklist walks the nine areas item by item, from positioning and packaging through to analytics and the competitor gap. Use it to gather the raw material, then use the method above to decide what any of it means.

The mistakes that make an audit useless

• Auditing only the output, so every finding is a note about a post.
• Benchmarking against a competitor’s follower count.
• Reporting twenty problems, which a team reads as none.
• Fixing three things at once, so no change can be attributed to anything.
• Blaming the algorithm. Instagram’s own ranking explainer says there is no singular algorithm, and that feed, Stories, Explore and Reels each rank on their own signals.
• Running it once. An audit is a repeatable test, not an event.
• Running it without the people who do the work, who already know where the queue jams.

How NBK thinks about audits

NBK is a social operations partner, so an audit is the front door to the system rather than a standalone document.

The team’s background is high-volume publishing, with senior experience across UNILAD, LADbible Group, SPORF, Social Chain and Supercar Blondie, and operations moving 600M+ views a month and 46K+ posts shipped. At that volume you learn quickly that output is almost never the constraint.

The constraint is upstream. An unclear promise. A hook nobody tested. An approval chain that eats a day. A report that hid the format that was working.

So an NBK audit is written to be actioned by the people who publish, and it ends where a plan can start: two constraints, the evidence behind them, an owner each and a date to re-test.

Next step

Run the five layers in order on your own account before you commission anything. Most teams find their constraint sits higher up than they expected, and finding that out costs a morning.

If your social output feels busy but not effective, an audit is the fastest way to find out why, before you commit another quarter to posting through it.