The DIY Social Video Audit: 90 Minutes, Honest Answers

Agencies charge $2,500 for a social video audit, and I've read enough of them to tell you what's inside: your own analytics, rearranged into a slide deck, with three findings you could have spotted yourself in an afternoon. The data isn't hidden. It's sitting in your platform dashboards right now, unexamined, because staring at your own numbers honestly is uncomfortable and there's always another post to make.

So block 90 minutes and do the uncomfortable thing. This is the exact audit I run on brand accounts — a timed, five-part process that ends with three specific changes, not a vibes-based "post more consistently." You need a spreadsheet, your last 30 videos, and a willingness to learn that your favorite video was probably not your best one.

Minutes 0–20: build the sheet

Open your platform analytics (TikTok Studio, Instagram professional dashboard, YouTube Studio — whichever is your primary) and log your last 30 videos, one row each. Columns:

• Date, hook (the literal first line spoken or shown), format (talking head / demo / trend / skit / educational), length
• Views, average watch time or retention %, likes, comments, saves, sends/shares, follows attributed (where available)

Twenty minutes is enough if you don't editorialize while logging. Resist the urge to explain any number ("that one flopped because of the long weekend"). Excuses during data entry are how audits die.

If 30 videos spans more than 90 days, note that too — your first finding is a cadence problem, and everything below will be noisier because of it.

Minutes 20–40: score what actually matters

Now compute five numbers. These are the ones that diagnose; vanity metrics stay out of the room.

Metric	How to get it	Healthy signal	What a bad number means
Median views	Middle value of all 30	Stable or rising over time	Falling = format fatigue or niche drift
Hook rate	Retention at 3s (or watch time ÷ length)	60%+ holding at 3s	Your first two seconds are the leak
Send rate	Sends ÷ views	Above 0.5% is strong	Content isn't worth a DM — no share trigger
Save rate	Saves ÷ views	Above 1% for educational	Content consumed, not valued
Follow conversion	New follows ÷ reach on best videos	0.5–1%+	Profile/promise mismatch

Use the median, never the average — one lucky video will otherwise convince you everything is fine. And compute each metric per format, not just overall. The whole audit hinges on comparing your formats against each other, because your account's real strategy question is "which 20% of what I make earns 80% of the results?"

Flag your top 5 and bottom 5 videos by views, and separately your top 5 by send rate. They will not be the same list, and the difference is informative: views measure what the algorithm pushed; sends measure what humans valued enough to forward (more on why that gap matters in dark social video sharing).

Minutes 40–60: interrogate the winners and losers

With the lists in front of you, answer these in writing — one sentence each. Writing forces honesty.

1. What do the top 5 have in common that the bottom 5 lack? Look at hook style, subject, format, and emotional register before you look at anything technical.
2. Read the ten hooks aloud — top 5 and bottom 5. Nine times out of ten, the winners open with a specific claim, question, or visual surprise, and the losers open with context-setting ("Hey guys, today I want to talk about..."). Hook quality explains more variance than any other single factor I've measured.
3. Where does retention die? Open your three best videos' retention graphs. Find the timestamp where each cliff happens and note what's on screen. Common culprits: the pivot to a sales pitch, a second idea starting, a static shot outstaying its welcome.
4. Which format is your quiet workhorse? Often a format with modest views but outsized saves or follows. That's a keeper even if it never trends.
5. What are you making that the data says nobody wants? Every account has a pet format kept alive by internal enthusiasm. Name it. You don't have to kill it today, but you have to know it's a hobby, not a strategy.

Minutes 60–75: audit the account surface

Ten videos of context, then zoom out. Visit your own profile logged out, as a stranger would:

• Does the bio promise what the top-performing videos deliver? Misalignment here is the most common follower-conversion leak — people enjoyed one video, hit the profile, and found a different show.
• Do your first nine grid tiles read as one coherent account? Mixed visual identity suppresses the "I know what following gets me" decision.
• Are your captions and covers doing any work? Text on covers is searchable surface area on TikTok and YouTube; blank covers waste it.
• Cadence check: count gaps longer than 5 days in the last 90. Every gap resets algorithmic momentum, and irregular accounts underperform identical content posted steadily.

This section usually produces the cheapest fixes of the whole audit — a bio rewrite and a cover template take twenty minutes and lift conversion permanently.

Minutes 75–90: write the three changes

An audit that ends in a document is decoration. End with exactly three commitments — one kill, one double-down, one fix:

• Kill: the format with the worst median across views and saves. Free that production time.
• Double-down: the winning format, at 2x its current share of your calendar. If demos win, make more demos; the production side stops being the constraint once you batch with AI generation — a week of the winning format is one sitting.
• Fix: the single biggest leak the numbers exposed. Usually it's hooks (rewrite your first two seconds against your own winning patterns), sometimes retention (cut videos at the cliff timestamp), sometimes profile mismatch (rewrite the bio to match the winners).

Put a date eight weeks out to re-run the audit and compare medians. One audit is a snapshot; two audits are a trend; quarterly audits are a strategy system. If the re-run shows your changes moved nothing, the problem is upstream of execution — likely positioning, which is a platform and audience choice question, not a content question.

Two closing honesty notes. First, 30 videos is a small sample; treat any pattern that hinges on two data points as a hypothesis to test, not a law. Second, if your last 30 videos span six months, your finding is already made: volume is your bottleneck, and the fastest fix is batching production — our video-only growth plan for your first 1,000 followers lays out the weekly system.

FAQ

How often should I audit my social media video performance?

Quarterly for a full 90-minute audit, with a 15-minute weekly glance at median views and send rate. More often than quarterly and you're reacting to noise; less often and you'll spend months scaling a format that stopped working.

What is the most important metric in a video audit?

If forced to pick one: retention at 3 seconds (hook rate). It gates everything else — no metric downstream of the hook can be good if viewers leave at second two. After that, send rate is the best single proxy for content people genuinely value.

How many videos do I need for a meaningful audit?

Thirty is the practical minimum for per-format comparisons; fifty is better. Below twenty, individual outliers dominate and you'll draw confident conclusions from luck. If you don't have thirty videos yet, your audit finding is simply "post more, audit later."

Should I delete my worst-performing videos?

Generally no — old flops don't harm current distribution on major platforms, and they're useful baseline data. The exception is content that actively misrepresents your current positioning. Archive those; leave the rest as your record of what didn't work.

Can AI tools help with a social video audit?

Yes, in two places: trend analysis tools can benchmark your numbers against what's currently working in your niche, and AI production tools make the "double-down" step feasible by removing the cost of making more of your winning format. Versely's trend analysis and batch generation cover both ends.

Found your winning format? Scale it the same week — batch it with the AI video generator or build it into a repeatable workflow that runs on schedule. Free credits daily.