Auditing an inherited ad account: a two-hour script

Inheriting an account means inheriting a working system nobody documented. There are campaigns running, decisions someone made for some reason, and no record of what the reason was.

This is a roughly two-hour script that separates what is working from what is merely running — and that gives you the baseline without which you will not be able to prove anything you do afterwards.

First of all: do not switch anything off

The temptation to start from scratch is strong and almost always wrong.

A running campaign carries learning. Switching it off and recreating it with the same settings costs days of unstable delivery and money spent relearning.

Switching off transfers responsibility. If results drop in week 1, the cause is you. If they drop with everything as it was, the cause is what you inherited — and you have the baseline to show it.

Two exceptions: spend happening with zero conversions for days, and clearly broken measurement. Act on both on day one.

Hour 1 — what the account is

Measurement (20 min)

Start here, always. Half of inherited accounts have a problem here, and no number is worth anything until it is checked.

• Have conversion events fired in the last 7 days?
• Is any event duplicated — same event arriving via pixel and server with no deduplication?
• Is what counts as a conversion what the business calls a conversion?
• Is a conversion being counted in more than one campaign at once?

Submit a test lead. It is the fastest way to know whether the whole chain works.

Structure (20 min)

Draw what exists on paper:

The most revealing signal is the date of the newest creative. An account still running an ad from eight months ago says a lot about the care it was getting.

Money (20 min)

Last 90 days, by campaign:

• spend
• result (conversions, and sales if there is a CRM)
• cost per result
• trend: improving, flat, worsening

This is where campaigns that spend without producing show up — the most common finding and the easiest to justify acting on.

Hour 2 — what is hidden

The questions nobody answers on their own

What was tested and failed? Without this you will repeat tests already run. Ask the previous agency if the relationship allows, or look through paused campaigns.

Why is this audience excluded? An unexplained exclusion usually has a good reason nobody wrote down.

What is the seasonality? Twelve months of spend and results show whether the current month is a good or bad one for the sector.

Who else edits the account? Three people editing with no coordination explains more inconsistency than any analysis.

The five most frequent findings

1. A campaign spending with no conversions for weeks, unnoticed because the overall dashboard looked reasonable.
2. Overlapping audiences competing against each other in the same auction.
3. A single creative taking nearly all the spend, with the others getting no delivery.
4. The wrong event — optimizing for signups when the business needs sales.
5. Unreadable naming, which blocks any by-source reporting.

The audit report

One page, three sections:

STATE
· 6 active campaigns, 11 ad sets, 9 ads
· Newest creative: 14 Aug (2 months old)
· Measurement: lead event OK; purchase event stopped since 2 Sep
· Naming has no standard: source cannot be separated in the CRM

COST OF THAT STATE
· Campaign "Brand-2" spent $8,400 over 60 days, 3 conversions
· 2 ad sets with overlapping audiences competing in the same auction
· Purchase event stopped = blind optimization since September

PLAN
Week 1  fix measurement · pause Brand-2 · observe
Week 2  new naming on active objects · 2 new creatives
Week 3  separate overlapping audiences
Week 4  first comparable reading

The middle section is the one that matters politically: it translates the state into money, which is the language of whoever approves your contract.

How to talk about the previous agency

Describe the state, not the intent. "Spent $8,400 over 60 days with 3 conversions" is verifiable. "They abandoned the account" is opinion, invites a defense, and makes you look like the professional who wins ground by criticizing.

Admit what you do not know. "The purchase event has been stopped since September; I cannot tell whether that was a site change or a configuration change" is stronger than an accusation.

Give credit for what is good. If the search structure is well built, say so. Credibility for criticism comes from the praise preceding it.

The first week

Observe, fix only what is urgent, and write down the baseline: spend, results and cost per result for the week you changed nothing.

That number is what will justify everything you do the following month. Without it, any improvement turns into an argument about whether it was already like that before.

For the routine that starts after the audit, see a weekly optimization routine. For fixing naming without breaking history, see campaign naming conventions.