The social media audit guide.

Pull the last 90 days of per-post data before you judge anything: views or unique viewers, average view time, view rate, follower movement, shares and saves, and revenue where it applies. Put it in a sheet you own, not a dashboard the platform can change or delete.

Mark the anomalies before you look for patterns. Filter out your top and bottom 10% of posts, then read the average that remains. One viral outlier or one flop will bend every conclusion if you let it.

Note what was changing while the data was collected: a rebrand, a platform transition, a posting gap, a team change. A metric collected during a disruption is a baseline to beat, not a verdict on what works.

Ask the five-second question first: could a stranger say who this account is for and what they get, from the name, bio and last three posts alone? If the answer needs a paragraph, the positioning is doing no work.

Then look for the two silent killers: drift and dependence. Drift is content that has wandered from the promise the name makes. Dependence is numbers resting on one creator, one format or one trend that could vanish tomorrow.

This is a ten-minute sweep, but do it on every platform, on a phone, logged out. You are checking what a first-time visitor actually sees, not what the brand guidelines say they should see.

Pick the three to five accounts genuinely competing for your audience's attention, not the ones in your industry directory. Write down what each does well, then list the themes they all share. The shared themes are table stakes, not differentiators.

The output you want is the gap: an angle, format or lane the field has left open that fits what you can consistently produce. While the audit runs, scroll the discovery feed daily and log which competitors keep appearing. The algorithm is telling you who is winning.

The thumbnail or tile has one job: match what the first seconds actually deliver. Measure it by pairing click-through with early retention. A high click-through rate with a terrible average view time is not a win, it is evidence the packaging wrote a cheque the video does not cash.

Where a platform gives you multiple thumbnail slots, use them as a real test: variations of the same moment, not four unrelated clips. Compare two titles over the same footage, or two visuals behind the same hook, and keep whichever the data picks.

Watch the first five seconds of your last 20 posts in one sitting and count how long each takes to reach its first genuinely interesting moment: the fact, the reveal, the conflict. Anything that opens with a fade, a logo, a title card or a slow build-up is donating its audience to the next video in the feed.

Then put a number on it. Early drop-off, the share of viewers gone in the first few seconds, is the single most diagnostic metric in this guide. The benchmarks further down are the exact tiers we grade client work against.

Give every surface its own job and audit it against that job. Short-form discovery formats exist to reach strangers, so they must be self-contained with the payoff in the first seconds. Feed and story formats exist to keep the audience you already own, so they can carry sequence and stakes. Evergreen collections exist to give a new visitor somewhere to go next. The same cut posted to all three is failing at least two of them.

Score reused content before it ships: can a stranger understand the premise in three seconds? Is there a clear payoff? Does it survive without context from somewhere else? If not, re-edit it for the surface or do not post it.

The cadence rule we apply everywhere: when performance stalls, increase volume and test more variables. When something pops, slow down and protect its reach rather than burying it under the next upload. A fixed calendar that ignores the data is just a habit with a spreadsheet.

Audit speed too. For news-led content, measure the time from moment to post. Past 24 hours, your audience has already seen it somewhere else and you are paying full production cost for stale reach.

Read the last 50 comments and DMs and check two things: how long replies take, and whether anyone from the brand ever starts a conversation rather than just closing one. Reach is rented from an algorithm that changes; the community is the only part of the audience you actually own.

Read the retention curve at three points. The start tells you whether the packaging delivered: if viewers leave immediately, they did not get what they clicked for. The middle tells you whether the structure holds: a sharp mid-video slide means the edit drags or wanders off-theme. The end tells you whether the payoff landed, and on monetised video the end is where the money lives.

Judge videos by finishes, not starts. Views tell you who showed up; finishes tell you who was satisfied. Then put your top 10% and bottom 10% side by side and write down the pattern: topic, opening, length, format. That pattern is your next month of content decisions.