How to audit your own paid social account

You can audit the creative half of a paid social account with the exports Ads Manager already gives you, and these ten moves are how I would do it sitting beside you. Nothing needs a benchmark or a lift study. Every number is read against the account's own median and spread, because circulating benchmark rates mostly cite each other and your account is the only sample that describes your buyers.

1. One platform, one threshold

Export the quarter at ad level from one platform, never both into one sheet. On Meta pull impressions, reach, frequency, spend, 3-second video plays and ThruPlays, on TikTok the same with 2-second views and focused views. Write the threshold at the top, ThruPlay on Meta, completion or fifteen seconds, focused view on TikTok, six seconds, completion or an interaction inside six, and never compare one with the other. Note asset length beside every ad, because length inflates every rate. There is no good or poor here, only a usable sheet.

2. Stop, hold and watched rate per ad

Stop rate is 3-second plays divided by impressions and reads the opening. Hold rate is ThruPlays divided by 3-second plays and reads the body. Watched rate is ThruPlays divided by impressions and reads the whole ad. On TikTok use focused views, and treat the 2-second view as a lower bar than Meta's three seconds. The median of each column across ads with meaningful spend is your baseline. Good is a wide gap between your top ads and the median, with the spend in the top. Poor is a stop rate that barely moves whatever the ad, or a hold rate that collapses on every ad.

3. The spread

Rank the ads by watched rate and divide the top ad's rate by the bottom ad's. That ratio is the spread, the medium's capacity in your account, and the share of impressions served by the bottom half is where the money sat. Good is a spread that narrows across the quarter because the bottom was retired and the top extended. Poor is a wide spread holding for twelve weeks with most impressions in the long tail.

4. Served frequency against watched frequency

Pull frequency per campaign, impressions divided by reach, and multiply it by that campaign's watched rate. That is watched frequency, how often the average person reached actually watched, and a served frequency of six against a watched rate of one in ten is 0.6, so most people reached never watched once. Good is a watched frequency that reaches or passes one. Poor is a watched frequency well under one, however high the served number. Skips can refresh a memory that exists if the assets were on screen, they cannot build one, and Amplified Intelligence, which sells attention measurement, puts the threshold for new memory near 2.5 seconds.

5. CPM against cost per watched impression

CPM is spend divided by impressions, times a thousand. Cost per watched impression is spend divided by ThruPlays, the bridge between the tiers, because the ad decides what share of impressions become watched. Rank both ways. Good is the lists agreeing, or spend following the second where they differ. Poor is the account's best CPM ad in the bottom half on cost per watched impression, cheap impressions nobody watched. Write the objective beside both, because a falling CPM under reach and a rising CPM under conversion can both be the mechanism working.

6. Distinct concepts live

Group every ad that ran in the quarter by concept, a distinct idea, situation, cast and opening, a variant differing only in cut, length, format or copy, and count the concepts. Fifty ads that reduce to three concepts are three ads with fifty invoices, and the delivery system, which Meta says now uses creative diversification as its best lever for finding audiences, consolidates spend behind similar ones. Good is twenty to thirty distinct concepts across the quarter, each built for different buyers in a different situation. Poor is under ten, or variants counted as concepts.

7. Goal against objective

Pull the objective setting of every campaign and write the brand's goal beside it. The goal is what the brand wants, the objective is only the platform setting. Good is brand-building campaigns set to reach. Poor is a brand campaign set to ThruPlay or video views because the watched rate looks better that way, which it will, because the system then finds people who watch anything and reach narrows. We read attention, we do not buy it.

8. Distinctive assets in the first second

Open the ten highest-spend ads on a phone, sound off, and watch the first second. Count the ads where a distinctive asset is on screen inside that second, colour, character, pack, shape, a cue that reads without a logo, and the ads where the brand first appears after second three. A skip can refresh an existing memory if the assets were in frame, so the first second is where the reminding job is done or missed on most of the money. Good is the brand identifiable in the first second across most of the spend. Poor is the brand arriving at the end, so the unwatched majority got no cue.

9. Retirements in the last four weeks

Pull every ad paused in the last four weeks with its watched rate, cost per watched impression and frequency over its final three weeks, and the recorded reason. Fatigue has a signature, watched rate sliding, cost per watched impression climbing, frequency creeping, and TikTok's own definition is an audience that has seen the same ad too many times. Good is retirements carrying that signature and a reason, and ads still holding their watched rate left running whatever the calendar says. Poor is zero retirements, no loop running, or ads retired while still being watched. An ad is retired because it stopped being watched, never because it is old.

10. Who owns the weekly creative decision

No export needed. Write down the name of the person who, this week, decided which ads to pause, remix and replace, and the date of the last such decision. Good is a name and a date inside the last seven days, with the decision recorded against the hypothesis each ad was testing. Poor is a role rather than a name, the agency, the media team, the next planning cycle, and this move usually explains the other nine. Media has an owner for its weekly decisions, correctly. Creative usually has one at the start of a campaign and nobody between.

The worksheet

The worksheet is available as a download.

What to do with this

Run the ten moves on one platform this week and read the worksheet in the order How to read paid social results sets, effectiveness first if a lift study exists, then performance, then efficiency, then the operating model, moves six, seven, nine and ten. Then leave the media plan alone, the media team's remit is unchanged, because this audit reads the tier between the media report and the brand tracker, which most accounts have nobody reading. If the moves came back holding, creative is being managed in your account. If several came back poor, the explanation is structural, not personal. Media creates the opportunity to see your ads. Creative determines whether they work. Why those ten hold together, and what an account looks like when they do, is the subject of What is a creative system?

Questions

How many impressions does an ad need before its watched rate means anything?

Enough that the next thousand would not visibly move it, and the practical answer is to exclude ads with trivial spend from the spread and the medians rather than to set a universal floor, because the right floor depends on the account's volume. Read the quarter, not the day, and read ads of similar length together.

If the account is small, run the moves at concept level, pooling variants, which also makes move six do double duty.

Can we run the audit across Meta and TikTok together?

Run it twice. ThruPlay and focused view are different thresholds on different behaviour, Meta's 3-second play and TikTok's 2-second view are different bars, and a blended watched rate describes neither platform. One worksheet per platform, each with its threshold at the top, and compare the shape of the reading rather than the numbers.

Our media report says the account is efficient. Why would this show anything different?

Both can be true, because they read different tiers. The media report reads CPM, reach and frequency, which is media efficiency, and a well-run media team will have those in good order. The audit reads what happened to those impressions once they arrived, whether they were watched, by how many of the people reached, with the brand on screen, and at what cost per watched impression, which is the tier between the media report and the brand tracker.

An account can be efficient at buying impressions and poor at getting them watched, and the signals that show it are in the same export the media report came from.

Sources

• How to read paid social results, the measurement standard published in this section, September 2026. Stop, hold and watched rate definitions, cost per watched impression, and the rules on platform, threshold, baseline, objective and asset length. Our own standard, we sell the creative it measures.
• Meta Business Help Centre, video metric definitions (3-second video plays, ThruPlay) and "About ad auctions". Platform documentation, Meta sells the media. Verify wording against the live Help Centre.
• Meta for Business, "Demystifying Creative Diversification", April 2025, as quoted by Jon Loomer, August 2025. Creative diversification as the best lever to find relevant audiences. Platform, Meta sells the media.
• TikTok Ads Manager Help Centre, "About the video views objective", the focused view definition, and TikTok for Business creative fatigue documentation. Platform documentation, TikTok sells the media.
• Amplified Intelligence, "Why does the attention-memory threshold matter?", 27 June 2022. The 2.5-second attention-memory threshold. Amplified sells attention measurement.
• Jenni Romaniuk, Building Distinctive Brand Assets, 2018. Distinctive assets judged on fame and uniqueness, consistent assets with varied executions. Ehrenberg-Bass Institute, sponsored by large advertisers.