Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Meta Ads › Creative & Formats

Reporting on Meta Creative

How to build a creative report for Meta: name ads by concept, hook and format, export ad-level data, read each stage, track fatigue and log what to make next.
Tutorialintermediate10 min readUpdated

What a creative report answers that a test cannot

The creative testing guide is about one question at a time: two or three variants, isolated from the delivery system's preferences, read on the metric that matches the objective. That is the only way to decide between concepts with confidence, and it is slow and costly by design. Most of the ads an account runs are never in a test. They rotate inside working ad sets, where delivery backs the one it predicts will perform best, and they leave behind a long record of impressions, plays, clicks and results that nobody reads as a whole.

The creative report reads that record. It does not crown winners, because rotation is not a test and the ads never had comparable exposure. What it can do is find patterns across many ads that no single test would show: a concept that converts whichever hook opens it, a hook style that stops the scroll and then loses people, a format that keeps underperforming in the same placement mix, a type of ad that wears out in weeks while another lasts for months. Those patterns become hypotheses for the next tests and directions for the next briefs.

It also differs from its neighbours. The frequency and fatigue guide reads one ad's decline against its own first week; the report draws that curve for every ad and compares the curves by concept. The SOP for pausing underperforming ads decides whether one ad should stop now; the report decides what the account should make next. And the performance reporting guide sets out the cadence and structure of reporting across channels; this is the creative section of that pack, built for Meta.

QuestionWhere it is answered
Which of these two concepts should we run?A controlled test, as the creative testing guide sets out
Is this ad wearing out?The ad's curve against its own first week, as the fatigue guide sets out
Should this ad be paused today?The pausing SOP's agreed thresholds
Which ideas keep working, and where do the others fail?This report
What should the next briefs ask for?This report, read with the test log

Before you start

  • Reporting access to the ad account in Ads Manager, at ad level, including retired ads.
  • A naming convention for ads, or the willingness to rename live ads to one. The UTM and naming conventions guide sets out the rules this tutorial follows.
  • The account's attribution setting, and the compare attribution settings columns, which the attribution settings guide explains.
  • The test log from the creative testing guide, so that the report can sit beside what controlled tests have already settled.
  • A spreadsheet or reporting tool that can split ad names into fields and group by them.
  • A fixed day each month and an owner, agreed with whoever runs the paid social account management playbook, whose monthly review is where the report is presented.

The steps

  1. Name every ad by concept, hook and format

    A report can only group what the names carry. Give every ad a structured Naming Convention with fixed fields in a fixed order: the concept, the hook, the format and a variant letter, for example fresh-roast_question_video_a. Follow the naming guide's rules: lower case only, one separator inside a field and a different one between fields, and letters, digits, underscores and hyphens only. Keep a controlled list of values for each field, so that "origin story" is never also "origin-stories" or "farm". Rename live ads where the convention is new; editing an ad's name does not appear on the learning phase guide's list of significant edits. Retired ads can be mapped to the fields in the spreadsheet instead.

  2. Fix one attribution setting and date range

    Choose the attribution setting the report will use and write it at the top. Meta's documentation on attribution settings describes the window as the period in which conversions are credited to an ad, and every result, cost per result and conversion rate in the report inherits it. Use the account's own setting, or a click-only view from the compare attribution settings columns if the team wants a stricter count, but one setting for every row and every month. Choose a date range that ended long enough ago for the window to have filled: the attribution guide notes that conversions are reported against the date of the ad interaction, so the last days of any period keep rising for a while, and that modelled figures and fine slices deserve care.

  3. Export ad-level results with spend and video plays

    Pull a report at ad level for the period, including ads that have been paused, with the columns the site's Meta guides use: amount spent, impressions, Frequency, link clicks, landing page views, results and cost per result for the ad set's optimisation event, plays of the opening seconds and plays at a later point or ThruPlays for video, and purchases and purchase value where the account sends them. Add the ad's first delivery date and the campaign and ad set names, because the same ad can sit in prospecting and retargeting and should be read separately in each. Keep the export as it came out and work on a copy.

  4. Compute the stage rates for each ad

    The creative testing guide reads an ad link by link, because a variant can win at one stage and lose at the next. Compute the same rates for every ad.

    StageRateWhat it says
    StopPlays of the opening seconds divided by impressions (video)Whether the hook stops the scroll
    HoldLater plays or ThruPlays divided by opening plays (video)Whether the body keeps the viewer
    ClickLink clicks divided by impressionsWhether the message earns an action
    ArrivalLanding page views divided by link clicksWhether the click became a visit
    ConversionResults divided by landing page viewsWhether the page kept the ad's promise
    CostSpend divided by resultsWhether the whole chain is affordable
    ValuePurchase value divided by spend, where sentWhether the buyers it found are worth having

    Use link click-through rather than all clicks, as the low click-through diagnosis explains, and keep each rate's count beside it. The CTR calculator and the CPA calculator check the arithmetic for any row that looks odd.

  5. Group ads by concept and by hook

    Split the ad names into their fields and total the counts for each concept, then for each hook within a concept, then for each format. Recompute the rates from the totals, never by averaging the ads' rates, because a rate averaged across ads with very different spend overweights the small ones. The grouping is where the report earns its keep: one ad with a good click-through is an anecdote, but a concept whose ads all hold viewers and convert well, across several hooks and formats, is a pattern. Note where a concept or hook received so little delivery that it was never really read; under rotation that is common, and it means "untested", not "lost".

  6. Draw each ad's curve against its own first week

    For every ad that has run for more than a few weeks, chart its click-through, cost per thousand and cost per result week by week beside its frequency over a trailing window, and compare each week with the ad's own first week, as the fatigue guide recommends. A falling click-through with rising frequency while reach flattens points to saturation; falling click-through while reach still grows points to wear-out. Then compare the curves by concept. Some concepts last; some burn out quickly however they are opened. That lifespan is one of the most useful things the report can tell the people who plan production, and it is the evidence behind any statement in the report about Creative Fatigue.

  7. Separate patterns from noise

    Before writing a conclusion, check the counts behind each rate. The guide to reading small numbers explains why a rate built on a handful of conversions jumps around, how to put a range around it and when to pool data until it can be read. Pool months for slow-moving concepts rather than ranking them on one. Check that a pattern holds in more than one ad set before believing it, since a concept that only ever ran in retargeting will look better than one that only ran in prospecting. And remember the rotation problem: an ad that took most of the spend may have done so because delivery predicted it early, not because it was better. Label each finding as a pattern to act on, a hypothesis to test or noise to ignore.

  8. Decide what to scale, iterate and retire

    Turn the findings into three lists. Scale: concepts that converted across hooks and formats get more variants, new hooks, new formats or new creators on the same idea, since the testing guide notes that an iteration of a proven concept has a higher hit rate than a new one. Iterate: concepts that held viewers or converted well but lost people at one stage get a change aimed at that stage, such as a new hook for a concept that held but did not stop, or a new landing page for one that earned clicks but did not convert. Retire: ads whose curves have bent and concepts that failed at the same stage repeatedly are paused through the pausing SOP, with a note of why, so the idea is not remade by someone who has not read the report. Hypotheses that matter go to the test backlog.

  9. Log the lessons and write the next briefs

    Write the commentary with the what, so what, now what framework: what happened by concept and stage, why it probably happened, labelled as a judgement, and what will be made or tested next. Add the lessons to the same log as the controlled tests, so that one record holds both. Then write the briefs. Each should name the concept, the stage the brief is meant to fix, the hooks to try and the metric that will judge it, which is the performance brief the creative hub describes. A report that ends without briefs has described the past without changing the next month.

  10. Repeat on a fixed day each month

    Run the report on the same day each month, on a period that has closed, with the same attribution setting, columns and groupings. Keep each month's version rather than overwriting it, because the concept lifespans and the stage patterns only become clear across several months. Review the naming convention at the same time: new concepts need values added to the controlled list before they launch, not after. Present the report in the monthly review, and record the decisions alongside the performance reporting guide's decision log.

A coffee roaster's monthly creative report (illustrative)

Illustrative example — not a real result

Hazeldrum Coffee Roasters reads a month of creative

Hazeldrum Coffee Roasters, an invented roaster selling a coffee subscription, runs one prospecting ad set and one retargeting ad set and rotates several ads in each. Its October report covers every ad live in the month, read under the account's attribution setting, with figures for prospecting only. Ads are named by concept, hook and format, so the team groups them in minutes.

ConceptHooks runSpendStop rateHold rateLink CTRConversion from page viewsCost per first box
Freshness: roasted this weekQuestion, close-up£3,10031%22%1.4%6.8%£24
Origin: the farm storyFounder to camera, landscape£2,40038%12%1.1%4.1%£41
Convenience: never run outProblem first£1,50019%25%1.2%7.4%£27
Price per cup against cafésNumber first (static)£400n/an/a0.6%3.9%£58

The origin concept stopped the most people and then lost them: its hold rate was the lowest and its conversion rate weak, so its curve looked fine at the top and failed further down. Freshness converted under both hooks. Convenience held and converted well but stopped fewer people, a hook problem rather than a concept problem. The price static received too little spend to read and was marked untested. The freshness close-up ad's click-through had fallen from 1.9% in its first week to 1.2% in its fifth, with trailing frequency up from 1.3 to 2.6, so it moved to the retire list once a successor was live. Three briefs followed: two new freshness hooks in vertical video, a stronger opening for the convenience concept, and a controlled test of the price angle against freshness, since the report could only suggest that comparison, not settle it.

Common mistakes

  • Unstructured ad names. Without concept, hook and format in the name, the report cannot group, and the patterns stay invisible.
  • Mixing attribution settings. One table under the account setting and another under click-only is two reports pretending to be one.
  • Reading an open period. The last days of the month are still filling in; the report will flatter the previous month.
  • Calling winners from rotation. Delivery chose early; the report finds hypotheses for tests, not verdicts.
  • Averaging ads' rates. Recompute rates from totals, or small ads distort every group.
  • Ranking on tiny counts. A concept with a handful of conversions is unread, not best or worst.
  • Comparing an ad with other ads to judge fatigue. Compare it with its own first week.
  • Ending with a league table. The report is finished when the briefs are written.

Frequently asked questions

Can a monthly creative report tell me which ad is best?

Not on its own. Most ads in a report ran side by side in the same ad sets, and the creative testing guide explains that delivery favours the ad it predicts will perform best, so the others never had comparable exposure. The report is good at finding patterns, such as a concept that keeps converting or a hook that keeps losing viewers early, and at turning them into hypotheses. Deciding between concepts needs a controlled test, which the report should feed rather than replace.

Which attribution setting should a creative report use?

One setting for the whole report, written at the top of it. Results, cost per result and conversion rate all inherit the window they were counted under, and Meta's documentation on attribution settings describes how a wider window credits more conversions to the same ads. Use the setting the account optimises on, or a click-only view through the compare columns if the team prefers a stricter count, and never mix the two in one table or between months.

How do I tell creative fatigue from a bad month in the auction?

Look at the shape of the change. If cost per thousand rose across every ad at once while click-through held, the auction moved. If one ad's click-through fell against its own first week while its frequency over a trailing window rose, and newer ads held up, that ad is wearing out. The frequency and fatigue guide sets out the signs; the report's job is to draw each ad's curve so the pattern is visible.

Terms used on this page

Tools to try

Frameworks

Also related

Sources

  1. About attribution settings — Meta Business Help Centre
  2. Meta Business Help Centre — Meta Business Help Centre

Related