How to tell if your ads agency is doing a good job: five checks you can run yourself

You can tell whether your ads agency is doing a good job in about an hour, using five checks that need no marketing training, only your own records and the willingness to ask for two reports. In the Google Ads accounts I took over during 2026, the expensive problems were visible in standard reports the whole time, and nobody had looked.

One thing before the checks. Plenty of agency people do honest work while stretched across too many accounts, and these checks are just as good at confirming a solid agency as they are at catching a poor one. Run them as due diligence, and hold your conclusions until you have the evidence, because accusations travel faster than corrections.

Which searches is my money actually buying?

Ask for the search terms report for the last 30 days, exported as a spreadsheet. It lists the real phrases people typed before clicking your ad, which is a different thing from the keywords the agency chose to bid on, because Google matches loosely. Read the top 50 rows by cost and mark anything you would never want to pay for: another company's brand name, job seekers, students doing homework, products you do not sell. Every account collects some junk. A managed account should show that junk being added as negative keywords (exclusions, phrases the account refuses to pay for again) month after month. If the agency cannot produce the report within a day, or the top rows are full of rubbish that has plainly been running for months, you have your first answer.

Do the reported conversions match my own records?

Take last month's report, find the conversions number, and compare it to your own count of enquiries: the forms in your inbox, the calls your receptionist logged, the WhatsApp messages. Expect some drift between the two numbers, since attribution is imperfect, but they should land in the same neighbourhood. When an agency reports 60 conversions and you can find nine enquiries, one of those numbers is fiction, and it is rarely yours.

What does each conversion actually fire on?

Ask the agency one precise question: "For each conversion action in the account, what does a person have to do for it to count?" The only good answers describe things a real prospect does, such as submitting the enquiry form or tapping the phone number. In one account I took over in 2026, 75% of all clicks were being counted as conversions because the tag fired the moment the page loaded. Every visitor was a "lead", the reporting looked spectacular, and the business was getting almost nothing.

You can check part of this yourself without asking anyone. Right-click your own website, choose View Page Source, and search the text for gtag, GTM- and AW-. Those strings show which Google measurement is installed at all. Finding them proves the plumbing exists; the question above establishes whether it measures anything real.

Did my budget run on Search or on Display?

Search ads appear when someone actively looks for what you sell. Display ads are banners scattered across apps and websites, and for most service businesses they produce cheap clicks that go nowhere. Ask for spend split by network over the last 90 days. In takeovers I did in 2026, I found between 43% and 59% of total spend leaking to the Display network as junk traffic the client never asked for, usually because a default setting was left switched on. If most of your money sits on Search, that is the healthy pattern. If half of it sits on Display and nobody ever discussed that choice with you, you have found a leak you can raise today.

What did the agency test last month?

Ask "What did you test last month, and what was the result?" A managed account should always have at least two ads running against each other, and someone should be able to tell you which one won and what replaced the loser. A vague answer about ongoing optimisation usually means the account has been coasting. This check has a second edge: a good agency will answer it instantly and will enjoy being asked.

What does good agency work look like?

Fairness matters here, because tracking is genuinely fiddly and suspicion is cheap. In one verification sweep I ran in 2026, 30 of 70 suspected tracking problems turned out to be false alarms once I checked inside the Google Tag Manager containers, the public files that hold a site's measurement setup. So treat any single red flag as a question to raise rather than a verdict to deliver.

Here is what a monthly fee should visibly buy: the search terms report gets reviewed and the junk gets negated, ads run against each other in live tests, the landing page loads fast and asks plainly for the enquiry, the conversion count roughly matches your own records, and the monthly report reads in five minutes without a glossary. On fees, percentage-of-spend arrangements commonly land somewhere between 10 and 20 percent of media budget. My standing rule is that below roughly R7,500 per month in media spend (about $400 to $500) you should not be paying a management fee at all, because the fee eats the budget that was supposed to buy customers.

If checks two and three worry you, I wrote a longer piece on the three pieces of evidence every advertiser should hold their reporting against, at /writing/the-three-receipts. I run paid media and rescue broken conversion tracking for a living, mostly in accounts exactly like the ones described above, so if a check turns up something ugly, you now know precisely what to ask whoever fixes it.