Scope: what this audit covers and what it does not
The question this audit answers is whether a TikTok Ads account is owned, measured and structured well enough to trust its results and add spend, and which weakness should be fixed first. It covers the paid side of an account that has been running for a while: who owns Business Center and the assets inside it, whether the events delivery learns from are sound, how campaigns, ad groups and budgets are arranged, how audiences and exclusions are built, how much genuinely different creative each ad group has, how Spark Ads are authorised, and whether reported results are compared with real orders or leads.
It does not re-teach the work. Building the ownership layer belongs to the Business Center set-up tutorial, installing and verifying events to the pixel and Events API tutorial, and the arrangement of campaigns and ad groups to the campaign structure guide. A campaign about to go live is checked with the TikTok Ads launch QA checklist, and a first campaign is built with the first campaign tutorial. The week-to-week running of paid social is the paid social account management playbook.
Organic posting is out of scope; it is scored by the social media presence audit. The pages ads send people to are scored by the landing page audit, and a business that also advertises on Meta can score that account with the Meta Ads account audit on the same scale. Every scorecard sits in the audit library.
Before you score: access and evidence
Ask for admin or analyst access to Business Center, Ads Manager and Events Manager. Without Events Manager you cannot see how events arrive, so the measurement rows stay provisional until you have it. Fix an evidence window long enough to cover several rounds of creative, because fatigue and learning only show over time, and keep the same window for every rescore.
Collect:
- the Business Center member and partner list with each role, and who holds admin;
- Events Manager diagnostics and test events, with each event's browser and server sources;
- campaign, ad group and ad exports for the window, with objective, optimisation event, bid strategy and the level the budget is set at;
- the audience list with sources, the exclusions in use and the seeds behind each lookalike;
- every Spark Ads authorisation with its expiry date, and the creator agreements behind them;
- the business's own orders or leads for the same window, and the last few monthly reports.
Write down the business target for each campaign before you look at results. Where a campaign is judged on cost per acquisition (CPACost of each conversion.), the CPA calculator helps check that the target is consistent with the margin the business has told you about.
Criteria and scoring
Rate each row missing (absent or broken), partial (present but incomplete, inconsistent or out of date) or solid (matches the "what good looks like" column). The scale is your judgement against that column, not a number. Rows marked gate decide the band on their own when they score missing.
Ownership and access
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Business Center ownership (gate) | Business Center, the ad account, the pixel and the brand's TikTok account sit under a business login with a second admin from the business | Missing: assets in a freelancer's, agency's or one person's login / Partial: owned, but one admin or a personal billing card / Solid: owned, a second business admin, recorded |
| Partner and staff access | Agencies and staff work through roles the business grants, scoped to their job, with leavers removed and no shared passwords | Missing: shared logins / Partial: roles granted but never reviewed / Solid: narrow roles, reviewed |
The Business Center tutorial sets out what owned looks like, including linking the brand's account so its own posts can run as Spark Ads without codes.
Measurement
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Pixel events and parameters (gate) | Standard events fire on the right actions with values and content identifiers from the data layer, proved with test events | Missing: no pixel, or purchases fire on page views / Partial: events fire but values or identifiers missing / Solid: tested and correct |
| Events API and deduplication | Key events also arrive from the server with the same event identifier as the browser, inside the consent pattern, so each conversion counts once | Missing: browser only, or both routes counted twice / Partial: server events without matching identifiers / Solid: both routes, deduplicated, consent respected |
| Attribution setting record | The attribution setting in use is recorded with its date and held constant across the periods being compared | Missing: nobody knows the setting / Partial: known but changed mid-period / Solid: recorded and stable |
Score the deduplication row against what the pixel and Events API tutorial describes from TikTok's documentation on the pixel and the Events API, not against third-party claims. The TikTok measurement hub explains why a changed attribution setting changes the numbers, which is why the record matters.
Structure and bidding
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Objective and optimisation event | Each objective follows the business outcome and each ad group optimises for an event that happens often enough to learn from | Missing: objective chosen by habit, or an event that rarely fires / Partial: right objective, event starved / Solid: deliberate on both |
| Ad group design | One ad group per hypothesis, broad with sensible exclusions unless the hypothesis is specific, and no ad groups bidding against each other | Missing: many narrow, overlapping ad groups / Partial: some redundant splits / Solid: each ad group has one reason to exist |
| Budget level and learning | Budgets sit at the level that fits the job, are large enough for ad groups to leave learning, and change in modest steps | Missing: thin budgets, frequent large edits / Partial: some ad groups stuck learning / Solid: stable and sufficient |
| Naming and hygiene | Names carry objective, market, audience type, optimisation event and version, so reports can be filtered without opening the account | Missing: default names / Partial: inconsistent / Solid: one convention throughout |
The campaign structure guide explains the Learning PhaseThe period after launch or a significant edit when an ad set's delivery is still being optimised. and why significant edits restart it; TikTok's help centre gives the current volumes, which this scorecard does not repeat.
Audiences
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Purchaser and lead exclusions | Prospecting excludes existing customers using an uploaded customer file as well as pixel events | Missing: no exclusions / Partial: pixel-based only / Solid: customer file and events, kept current |
| Retargeting sequence | Retargeting rungs exclude one another and are judged on what they add, not on their own reported cost per result | Missing: overlapping rungs / Partial: exclusions but judged on platform cost / Solid: separated and tested |
| Lookalike seeds and consent | Lookalikes are seeded from buyers or qualified leads, and customer files are uploaded only after a consent check | Missing: seeded from all visitors, files uploaded unchecked / Partial: buyer seeds, consent unclear / Solid: buyer seeds, consent checked |
The custom and lookalike audiences guide covers each Custom AudienceAn ad audience built from the advertiser's own sources: customer lists, site visitors or engagers. source, the exclusions that stop paying for existing customers and the seed a Lookalike AudienceAudience similar to a seed list. copies.
Creative and Spark Ads
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Native openings | Ads earn the first seconds with a visual, spoken or text hook made for the vertical feed, with native text, sound and captions | Missing: repurposed landscape adverts / Partial: vertical but slow to start / Solid: made for the feed |
| Creative testing method | Rotation chooses and split tests prove; concepts are varied before hooks, and ad names carry concept codes | Missing: no method / Partial: budget favourites treated as winners / Solid: method and concept codes in use |
| Spark Ads authorisation and rights | Brand posts run from a linked account; creator codes, expiry dates, sound clearance and usage terms are tracked | Missing: expiries unknown / Partial: tracked for some creators / Solid: a record for every code |
| Fatigue watch | Frequency, click-through and cost per result are read by ad, and tired ads are replaced in the live ad group before results slide | Missing: averages only / Partial: replaced after results fall / Solid: read weekly by ad, planned rotation |
Openings are judged against the content formats and hooks guide, testing against the creative testing guide and authorisation against the Spark Ads guide. When cost per result is already climbing, the rising cost per result diagnosis separates Creative FatigueThe decline in an ad's response as the same people see it again and again. from an expired Spark authorisation or a restarted learning phase.
Reporting and evidence
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Reconciliation with the business | TikTok's conversions are compared each month with the business's own orders or leads, with view-through credit on its own line | Missing: platform figures only / Partial: occasional checks, view-through folded in / Solid: monthly, view-through separate |
| Incrementality evidence | Spend decisions rest on a holdout, a lift study or outside signals chosen before the results, not on TikTok's column or last click alone | Missing: no evidence beyond attribution / Partial: outside signals only / Solid: a planned test behind the main spend |
The monthly TikTok report tutorial shows the join that removes double counting, and the guide to what TikTok adds beyond last click sets out the routes to evidence of IncrementalityConversions caused by marketing..
How to score and prioritise
Score the two gates first. If Business Center ownership is missing, every later row describes an account the business could lose, so ownership is fixed before anything else. If pixel events are missing, delivery has been learning from the wrong signal and the structure, audience and budget rows are provisional: what looks like a structural fault may be the system compensating for bad events. Mark them so.
Then rank the remaining weaknesses in this order:
- Whatever stops delivery learning: the wrong optimisation event, budgets too thin to leave learning, and ad groups that compete with each other.
- Spend on people who would buy anyway: missing purchaser exclusions, and retargeting judged on its own reported cost.
- Creative depth, fatigue and testing, because they decide results once the foundations hold. Production takes time, so schedule it early even if it lands later.
- Naming and report polish last, unless they hide a problem the other rows need to see.
Within each group, rank by the spend the weakness touches. A fix on the ad group that carries most of the budget outranks the same fix on a small test.
Reading the result: what to do with each band
Red means a gate scored missing. Hold budget increases and new launches, move the assets under the business's ownership or repair and deduplicate events, and verify with test events before trusting any result. Rescore after a full month on the repaired set-up.
Amber means the gates hold at least partially but several other rows are partial or missing. This is the usual outcome. Work a short ordered list, one area at a time, so the effect of each change can be read.
Green means the gates are solid and most rows are too. Move effort into a steady creative testing rhythm and into the landing pages behind the ads, and rescore each quarter to catch drift.
From findings to a prioritised fix list
Write each partial or missing row as a finding in plain words: what you saw, the evidence, and why it matters. Then turn the findings into one fix list with a line per fix: the finding, the fix, the owner, the guide that explains it and the rescore date. Order the list by the priorities above, not by how easy each fix is.
Group the lines by who does the work, because ownership fixes sit with the business owner and admins, event fixes with whoever runs the tag manager and the site, structure and audiences with the media buyer, and creative with the people who make it. Present the band, the gate findings and the first few fixes to the budget holder, and keep the full scorecard as the working copy for the rescore. The academy exercise on auditing creator work and an inherited TikTok account for an oral care brand practises scoring an account taken back from an agency that created the pixel in its own login, beside a creator programme audit, with rescore dates on the same evidence windows.
Illustrative example
A lip balm brand inherits a TikTok account
Kivdorra Balms, an invented lip balm brand spending about £6,000 a month on TikTok, takes the account back from a departing freelancer and runs the audit over the last eight weeks. Business Center ownership scores missing: the pixel was created in the freelancer's own Ads Manager. Pixel events score partial, and Events API and deduplication score missing, because browser and server purchases carry different event identifiers; TikTok reports 410 purchases for the period against 260 orders in the shop. Purchaser exclusions score missing, so prospecting is paying to reach existing customers. Creative depth is fair but fatigue watch is partial: the two ads carrying most of the spend have run unchanged for six weeks. The band is red. The fix list puts a new business-owned pixel and matching event identifiers first, a customer-file exclusion second and three new concepts third, and holds a planned budget increase. After a full month, reported purchases sit within a few orders of the shop's records, the gates score solid and the band moves to amber.
Common mistakes
- Scoring results before ownership and events. An account that reports well but sits in someone else's login, or counts each purchase twice, is not a good account. Score the gates first.
- Treating the budget favourite as a proven winner. Rotation decides where spend goes; only a split test proves a difference. Score the testing row on method, not on which ad got the most money.
- Marking broad targeting as a failure. The criterion is whether each ad group has one reason to exist and sensible exclusions, not whether interests are stacked.
- Judging retargeting on its own reported cost. It reaches people likely to buy anyway. Score it on what it adds.
- Comparing periods on different attribution settings. The numbers change with the setting. Check the record before reading any trend.
- Fixing everything in one week. Large simultaneous edits restart learning and blur cause and effect. Work the list in order and rescore on the same evidence window.
