How to tell if your marketing agency is actually working

An agency report can be accurate line by line and still not tell you whether your marketing is working. Its figures come from Google Ads, Meta or GA4, and each measures what that platform counted as a conversion. Your business runs on sales that cleared, net of refunds, and nobody has to lie for a report to look great in the space between the two.

So the way to tell if your marketing agency is working is to check its numbers from the side that pays the invoice. A first pass takes about 30 minutes, and I'd start with five checks:

1. Your company owns the ad accounts.
2. You know what the account counts as a conversion.
3. Total revenue against total ad spend holds up over months, not days.
4. Searches for your own name are reported apart from customers the ads had to win.
5. What the agency counted roughly matches your real orders.

Why can a good agency report hide flat sales?

Because each platform keeps its own score. When a customer clicks a Google ad on Monday, taps through from an Instagram ad on Wednesday and buys from an email on Friday, Google, Meta and your email tool can all claim the order, while your store recorded one sale.

Timing opens a second gap. Google Ads' standard columns date a conversion to the day of the click (opens in a new tab), not the day of the sale, so the agency's month and your bank's month can differ and both be right.

Refunds and chargebacks add a third. They land weeks or months after the sale, so the newest campaigns look best until the reversals catch up.

Do you own your Google Ads, Meta, GA4 and Tag Manager accounts?

Your company should own all of them, with the agency given access. If the agency owns them, leaving can cost you years of conversion data.

Google Ads

If a manager creates a new account, it "will automatically become the owner of that account," (opens in a new tab) Google says, and "a client account can only have one owner." So if your agency created the account inside its own manager account, the agency owns it, even though your card pays for it.

Linking an account you opened yourself doesn't transfer ownership by default, so open it in your company's name and invite the agency in. If the agency already owns it, ask them to transfer it to a manager account you control.

Meta, GA4 and Tag Manager

On Meta, your own business portfolio (formerly Business Manager) should hold the ad account and the pixel, with the agency added as a partner. In GA4 you want the Administrator role, and in Tag Manager at least two administrators, one of them you, since anyone who can publish there can change the tags that count your conversions.

Put it in the contract too: you own the accounts, and the agency hands over anything it still controls within a set number of days of parting ways.

What counts as a conversion in your ad account?

Whatever the account has been told to count, which often isn't a sale. When we audit an account, the conversion list is the first thing I open, because most of the tracking we inherit is broken somewhere.

In Google Ads, primary conversion actions fill the "Conversions" column and steer bidding. The "All conversions" column is broader (opens in a new tab): Google says it adds the actions you've kept out of "Conversions," plus store visits, certain phone calls "and more," and it's the only one of the two that includes view-through conversions.

So check which column the report's headline uses, and ask the agency what event sits behind each primary action, since page views and add-to-carts often get counted too.

One real outcome, such as a purchase or a funded deal, should drive the bids. My note on landing pages for paid clicks covers how to count it once in Google Ads, GA4 and Meta.

ROAS vs MER: which number should you trust?

Use MER to judge the business and ROAS to steer campaigns. ROAS (return on ad spend) is the revenue a platform credits to its own ads, divided by what you spent there. MER (marketing efficiency ratio) is your total revenue divided by your total ad spend, from your own books.

A made-up month shows the gap. Google reports $30,000 in conversion value on $8,000 of spend, a ROAS of 3.75, and Meta reports $22,000 on $6,000, about 3.7. Your store took in $40,000 in total, organic sales included, so the platforms claim $52,000 of a $40,000 month. Your MER is $40,000 ÷ $14,000, about 2.9.

Whether 2.9 is good depends on your margin. At a 40% gross margin, each dollar of revenue leaves 40 cents to pay for ads, so break-even is 1 ÷ 0.40 = 2.5, before rent or salaries.

Now count the agency's fee as spend. A $3,000 retainer takes MER from 2.9 to 2.35, under break-even. An agency that doesn't know your margin can't tell you which side of that line you're on.

MER includes sales you'd have made anyway and swings with the seasons, so read it over six to twelve months. If spend climbs and total revenue stays flat, the extra money is buying very little. The ad ROI calculator shows what a dollar of ads returns in gross profit once repeat orders count.

Are you paying for customers who were already looking for you?

An ad at the top of a search for your company's name takes the click and the credit, though many of those people would have clicked your organic listing just below it.

How many is where the research disagrees. In field experiments at eBay, economists found that brand-keyword ads had "no measurable short-term benefits" (opens in a new tab). Google's own research across 390 pause studies (opens in a new tab) is kinder: when the advertiser's organic result already ranked first, as yours usually does for your own name, about half of the ad clicks wouldn't have happened without the ad.

That study counted clicks, not sales, and eBay is a household name. If competitors bid on your name, your answer may differ, and only a test on your own account will tell you.

Meanwhile, ask for brand and non-brand search reported separately, and whether Performance Max excludes your brand name.

How much credit goes to people who never clicked?

Often more than the report makes obvious. A view-through conversion is a sale from someone who saw an ad, didn't click, and bought later. Some were nudged by the ad and plenty would have bought anyway, which is why retargeting people halfway through your checkout can look brilliant.

Meta stretched the word "click" too. Until this year it counted a like, a share or a save as a click for attribution, and in March it announced it would limit click-through attribution to link clicks (opens in a new tab) for website and in-store conversions, starting later that month. If your Meta results fell from late March with no campaign changes, check that before blaming the agency.

Would those sales have happened anyway?

Some of them would have. Attribution tells you which ad touched a sale; incrementality tells you whether the ad caused it, and an attribution report on its own can't show you that.

So look for an agency that's willing to test it. On most budgets that means a planned holdout: pause one channel, or brand search, in some regions and not others for a few weeks, then compare total sales.

Agree on how you'll read the result before it starts. On a small budget the answer will be rough, which still beats a guess, and at larger budgets platform lift studies and marketing mix modeling start to earn their cost.

Do the agency's conversions match your real revenue?

Rarely to the order, and that's fine if you know the gap. In Google Ads, add the "Conversions (by conv. time)" column, which dates each sale to the day it happened, and set it by week beside your Shopify or Stripe orders, net of refunds.

Do the same with Meta's purchases. A steady gap you can explain is normal; a growing one needs an answer.

If you sell through leads, check last month's counted leads in your CRM: how many became customers, and for how much? A campaign that fills a form with people who never answer the phone can look excellent on cost per lead.

The lasting fix is to send real outcomes back to the ad account, so bidding learns from closed sales instead of form fills. Google's classic route, offline conversion import, matches each sale to its ad click through the GCLID, so that ID has to survive from the landing page to your form. For a new setup, Google now recommends enhanced conversions for leads (opens in a new tab) instead.

Does your agency's fee reward spend or sales?

Look at how the agency gets paid, then expect more of whatever that is. Incentives work on everyone, me included.

A fee that's a percentage of ad spend grows with your budget, whether sales follow or not. Plenty of honest agencies charge this way, but when results are flat, the easy recommendation is a bigger budget, which also raises the fee.

The badges lean the same way. To hold a Google Partner badge, an agency's manager account needs $10,000 of ad spend over 90 days (opens in a new tab) across all its clients, an optimization score of at least 70%, and at least half its account strategists certified. Premier Partner goes to the top 3% of participating companies in a country, ranked partly on year-over-year ad spend growth among existing clients.

None of that makes a Partner agency bad. The badge measures spend, setup and training, not your sales.

Performance fees usually pay on what the platforms count, which puts you back in the same gap. One tied to your own books is a better deal. Failing that, I'd look for a flat fee or a fixed scope, though I'm biased: that's how we price our own work.

Short answers

How long should I give an agency before judging results?

Judge the setup within the first month: by then you should own the accounts, and tracking should count real sales. Judge results over a quarter or more, on the MER trend and at least one test.

What is a good MER?

There's no universal number, because it depends on your gross margin. Break-even MER is 1 divided by that margin, so 2.5 at 40% and 4 at 25%, and you want to clear it with room for your other costs.

What should a monthly agency report include?

Monthly, in dollars, bad months included: spend by channel with brand search split out, conversions by conversion time beside your real orders, the MER trend with the fee counted, and what was tested or changed.

What are the biggest marketing agency red flags?

The agency owns your accounts, won't say what counts as a conversion, reports view-through as sales, never separates brand search, or doesn't know your margin. A promised ROAS before anyone has seen your data is another.

The 30-minute check, step by step

Have monthly revenue and ad spend to hand, plus your own login to each account. If you can't get a login, that's your first finding.

1. Google Ads access. Under Admin, then Access and security: are you an Admin, and which manager accounts are linked? Whose name is on the payments profile in Billing?
2. The other accounts. In your Meta business portfolio, who owns the ad account and the pixel? Are you an Administrator in GA4? How many administrators does Tag Manager have?
3. Conversions. Under Goals, then Conversions, then Summary: which actions are primary, and what fires each one? Is view-through inside the headline number, and which attribution setting do the Meta figures use?
4. MER. For each of the last 6 to 12 months, divide total revenue by total ad spend plus the agency's fee, with platform-reported revenue beside it.
5. Brand vs non-brand. In the search terms report for the last 90 days, add up spend and conversions on searches containing your name.
6. The spot-check. Compare last month's "Conversions (by conv. time)" with real orders or closed deals. Then skim Change history for who changed what. If nobody has touched the account in six weeks, ask what the fee is paying for.

What to do with what you find

Send it to the agency as questions, not accusations. "Which conversion actions are primary, and what fires each one?" gets further than "your numbers are wrong."

A good agency answers quickly, shows where its numbers and yours part ways, and suggests a test. If the accounts turn out not to be in your name, sort that out before you argue about any numbers.

The Attribution Honesty Checklist goes further, with checks that need a developer, like whether click IDs survive to checkout.

Our free Growth Engine Audit is the long version of this check. We start with read-only access, trace your last 90 days of spend against the orders you actually took, and send you what we find. Your current agency can stay while we look.

We hold our own figures to the same test, and how we count results sets out the rules.