How to tell whether your agency is doing a good job

An agency's work does not show in the account ROAS

Whether your agency does good work shows in the non-brand part of the account and in the comparison with your shop, not in the overall ROAS in the monthly report. The overall ROAS mixes searches for your shop's name with searches for products, and the brand searches lift the average whatever the agency does. You need no specialist knowledge for this, just two figures from the shop back end and two from the account, plus about an hour for spot checks.

I sell Google Ads management as a freelancer myself. The checks below are chosen so that you can carry them out without me and without a new agency.

Two ratios you can work out yourself

Ratio 1: reported revenue against shop revenue. Take the conversion value Google Ads shows for the last three months and divide it by your shop's online revenue for the same period. If the result is close to 100 percent or above, the account is claiming revenue that would have come through other routes anyway: newsletter, repeat customers, organic search, direct visits. I would then first ask which conversion actions feed into the value and whether anything is counted twice. How to do that is covered under reconciling conversion data with the shop system.

Ratio 2: brand share of reported revenue. Have the conversion value of the brand campaigns shown separately and divide it by the total conversion value. The higher this share, the less the overall ROAS says about the agency's work. In a large field experiment at eBay, ads on brand terms had no measurable short-term benefit: buyers arrived via the organic result anyway. That is a retailer with a very strong brand, and for smaller shops with competition on their own name the picture can differ. As a check, it is still sufficient: brand revenue is mostly not the agency's achievement.

The worked example shows why the two ratios together say more than any report. The figures are invented and rounded.

The report shows ROAS 6 and 90 percent of shop revenue, which is already suspiciously high alongside newsletter and repeat customers. Without brand, ROAS 2.6 remains. Whether that is good depends on your margin: if you earn 35 percent of revenue before ad spend, break-even is at ROAS 1 / 0.35 = 2.9, so the non-brand part loses money. At a 45 percent margin it is at 2.2 and the part pays for itself. More in brand revenue is not automatically Google Ads performance.

What the change history reveals about the work

The change history shows whether anyone is working on the account and who that is. According to Google, it lists all changes made to your account, campaigns and ad groups during the past two years. For changes made through the interface, the User column shows the email address of the person. Automated changes appear as Google Ads system, changes made via the API with the name of the interface or the tool.

From this you can read three things without understanding every change:

• Who makes changes: the person you talk to, a colleague you have never met, or almost only the system.
• When changes are made: spread across the month or bunched in the days before the report is due.
• What is changed: negative keywords and ad copy as well, or only budgets and bids.

My assessment: lots of changes are no mark of quality. A well-built account with an automated bid strategy does not need daily intervention, and hectic readjusting can even disrupt the strategy. The warning sign is rather the opposite case: months in which no human changed anything while the report talks about optimisation.

Ten minutes of search terms

The search terms report shows which searches triggered your ads. Sort it by cost over the last 30 days and read the top 50 rows. You know your range better than any agency. If you find terms with costs for which you sell nothing, job searches, how-to queries or competitors' products you do not stock, ask why they have not been excluded yet.

One limitation: terms with too few searches are left out of the report for privacy reasons, so you never see everything. An agency that explains this works properly. One that considers the report unimportant probably does not. More under analysing the search terms report.

Which warning signs are clear-cut

Most findings are a reason for a conversation. Some are not, because Google itself regulates them in its transparency requirements for third parties or because they concern your property.

Google has the access levels Email-only, Billing, Read-only, Standard and Admin. Read-only is enough for the check. To be sure you can take the account with you when switching, you need admin access on an address belonging to your company.

Questions you can ask as a non-specialist

You recognise good answers by the fact that they give figures, admit limits and fit your business. Evasive answers point to Google, the algorithm or the season.

• How much of the reported revenue comes from brand searches, and what is the ROAS excluding brand?
• Which margin do you work with, and where is the break-even ROAS for my shop?
• Which conversion actions count towards the conversion value, and how does it differ from shop revenue?
• What did you change last quarter that I can find in the change history, and what did it achieve?
• What would you cut with 20 percent less budget?

I consider the last question the most revealing. Someone who knows the account can say straight away which part contributes least. Someone who only administers it needs a week for that. A report that answers these questions unprompted looks like the one described in a monthly report for management.

When this check is misleading

The comparison works over several months, not for individual weeks. Google Ads attributes conversions to the date of the click, not to the purchase date, while your shop records by order date. At the end of a month, the account is therefore missing purchases that are added later.

The brand ratio says little if competitors deliberately bid on your shop's name. In that case the brand campaign can protect revenue, and a high share is no reproach to the agency. Only a test brings clarity then, not a ratio.

And none of these checks judges the strategy: whether the agency pushes the right products, wins new customers or simply pays again for repeat customers. The eBay experiment found that for non-brand terms it was mainly new and infrequent buyers who responded to ads, while frequent buyers accounted for most of the spend. Whether your agency knows this pattern and counteracts it is a good question for the next meeting. If the answers fall short, an audit with read-only access by someone independent is the next step, before you give notice.