Field Notes

How Do You Choose a Performance Marketing Agency?

How to choose a performance marketing agency: test their measurement, shopping and catalogue ads and creative testing, spot the red flags, and use the checklist.

Published Read time 11 min

Short answer: Choose a performance marketing agency by testing what generic agency checklists miss: whether they can explain your tracking and judge results on blended revenue rather than platform ROAS, whether they understand product feeds and catalogue campaigns, and who makes and tests the creative. Then rule out anyone who guarantees a ROAS before seeing your data.

Key takeaways

  • A performance agency is paid to move revenue from paid search, paid social and shopping ads, so judge it on measurement, feeds and creative, not on its brand story.
  • Ask them to explain how they would check that your conversion tracking is correct. Vague answers about "the pixel" are a warning sign.
  • Good agencies report blended figures like MER and new customer CAC next to platform ROAS, and they can tell you why the two disagree.
  • For ecommerce, the product feed is the targeting. Ask feed-level questions and listen for custom labels, product sets and margin.
  • Ask who makes the ads, how many new concepts ship each month, and what spend threshold they use before judging one.
  • Walk away from guaranteed ROAS, case studies that only quote platform ROAS, or an agency with no opinion on your feed or creative.

Most advice on choosing an agency is about chemistry, culture and case studies. That matters, and we cover it in our guide on how to choose the right marketing agency. A performance marketing agency is a narrower hire: it runs paid search, paid social and shopping ads, and you hand over your ad budget expecting measurable revenue back. The questions that separate good from mediocre are technical, and most can be tested in one call.

How do you test whether they actually understand measurement?

Ask them to walk you through how they would check your current tracking, and how they would decide whether a campaign is working. A competent agency answers with specific steps and specific metrics. A weak one talks about dashboards.

Questions that reveal tracking competence

You do not need to know the answers yourself. You only need to notice whether they give concrete ones.

  • "How would you check that our purchase or lead events fire once, with the right value?"
  • "Do you set up server-side tracking, the Meta Conversions API and Google enhanced conversions? How do you handle deduplication between browser and server events?"
  • "What does our consent setup do to the data, and how do you account for it?"
  • "If GA4, Google Ads and Meta all report different revenue for the same week, which number do you trust and why?"
✓ A good answer sounds like

"First we'd place a few test orders and compare them line by line against what Meta, Google Ads and GA4 record. We'd check the Conversions API is sending the same event ID as the browser pixel so purchases aren't counted twice, and look at your event match quality. Then we'd reconcile a month of platform-reported revenue against your actual Shopify revenue. The platforms will always over-claim a bit, so we report them next to a blended figure."

✗ A weak answer sounds like

"We install the pixel and connect GA4, then you get access to our live dashboard. Our clients average a 6x ROAS."

Why platform ROAS is not enough

Every ad platform measures its own success, and each one claims credit for sales that other channels, or no channel at all, would have produced. When Google, Meta and an email tool all count the same order, the sum of platform-reported revenue can be far larger than your actual revenue.

That is why a good performance agency reports blended metrics alongside platform numbers:

  • MER (marketing efficiency ratio): total revenue divided by total ad spend, across all channels. It cannot be gamed by attribution settings.
  • New customer CAC: ad spend divided by first-time customers. It shows whether you are buying growth or just re-buying people who were already coming back.
  • Contribution after ad spend: revenue minus product costs, shipping, fees and ad spend. This is the number that pays salaries.
Worked example

What the platforms say

In one month you spend €10,000 on Meta and €10,000 on Google. Meta reports €50,000 in purchases (5x ROAS). Google reports €60,000 (6x ROAS). Together that is €110,000 of claimed revenue.

What the business saw

Your shop recorded €80,000 in total revenue that month, including organic, email and direct. Blended MER is €80,000 divided by €20,000, so 4.0. The platforms are claiming at least €30,000 more than exists.

What a good agency does with it

It tracks MER and new customer CAC month by month, and asks what happens to total revenue when spend changes. If spend rises 25 percent and total revenue barely moves, the extra spend is mostly buying credit, not customers.

The best agencies also talk about incrementality: geo holdouts or platform lift tests that show what would have sold without the ads. You do not need this on day one, but an agency that has never heard of it will optimise towards whatever the platform tells it.

The common mistake here is accepting a beautiful dashboard as proof of competence. A dashboard only displays numbers. Ask what is underneath it. If measurement is the gap you are worried about, it is worth reading what proper marketing analytics services cover before you talk to agencies, so you know what to listen for.

How do you check they understand shopping and catalogue ads?

Ask feed-level questions. For an ecommerce brand, the product feed is effectively the targeting for Google Shopping, Performance Max and Meta catalogue ads, so an agency that only talks about bids and budgets is missing the main lever.

The simplest test: give them view access to your feed and ask what they would change.

Feed-level questions to ask

  • "How would you structure our Performance Max asset groups and listing groups, and why?"
  • "What would you put in custom labels?" (Good answers mention margin bands, bestsellers, price tiers, seasonality or stock level.)
  • "How would you build product sets in our Meta catalogue for Advantage+ campaigns?"
  • "How do you handle disapproved products, missing GTINs and out-of-stock items?"
  • "Who edits product titles for search, and does that happen in Shopify, in a feed tool, or in Merchant Center?"
✓ A good answer sounds like

"Your feed has 900 products but most of the revenue comes from about 60 of them. We'd add a custom label for margin and one for sales velocity, split Performance Max so high-margin bestsellers get their own budget, and stop low-margin items from eating spend. Your titles also lead with the brand name instead of what people search for, so we'd rewrite them to put product type, material and size first."

✗ A weak answer sounds like

"We'll launch one Performance Max campaign with all products and let Google's AI find the best ones. Feed management is handled by your Shopify app."

Why the structure matters

Automated campaigns chase the cheapest conversions, not the most profitable ones, and often pour budget into low-priced items and branded searches that would have converted anyway.

Worked example

One campaign, all products

Product A sells for €30 with a €6 margin after costs. Product B sells for €120 with a €50 margin. Both convert at the same ROAS of 4, so the campaign treats them as equal. A €7.50 ad cost per sale of A loses money. A €30 ad cost per sale of B makes €20.

Split by margin

With a margin label in the feed, the agency gives Product B its own campaign or asset group with a lower ROAS target and more budget, and caps spend on Product A. Same total spend, more contribution.

The common mistake is treating feed work as a one-off setup job. Prices, stock and bestsellers change every week, and the feed has to keep up. Ask who monitors feed diagnostics in Merchant Center, and how often. For deeper context on how we approach this on Google specifically, see our Google Ads management services.

A performance marketing agency that starts with your tracking and your feed

We audit measurement and product data before we touch budgets and report blended revenue next to platform numbers. Not the right fit for every brand, and we will say so if that is the case.

See our performance marketing agency

Who makes the creative, and how do they test it?

Ask exactly who produces new ads, how many they produce each month, and how they decide what to keep. On paid social, creative now does most of the targeting work, so a slow or unclear creative process will cap your results however good the media buying is.

Three models you will hear

  • The agency produces it in house. Fastest loop. Check how many new assets per month the fee includes.
  • Your team produces it, the agency briefs and tests. Cheaper, but only works if you deliver to their briefs on time.
  • Creators produce it, the agency manages them. Good for volume on Meta and TikTok. Ask who finds, briefs and pays the creators.

None is wrong. The problem is not knowing which one you are buying, and both sides assuming the other is making the ads.

What a real testing process looks like

Area
✗ Weak process
✓ Strong process
Volume
A new batch "when performance drops"
A set number of new concepts each month, agreed in advance
What is tested
Small colour or button changes
Different angles, hooks and formats that could genuinely change results
Decision rule
Whatever looks best after two days
A minimum spend or conversion count per ad before judging, on cost per purchase or new customer
Learnings
Lost when the account manager changes
Logged in a shared document you can keep

A simple rule of thumb for the decision rule: if your target cost per purchase is €40, an ad that has spent €40 with no sale has told you almost nothing. A common rule is to wait until an ad has spent two to three times the target cost per purchase before calling it. Ask what their threshold is. If they do not have one, results are being judged by gut feel.

The common mistake is judging creative on click-through rate alone. A clickbait hook can double CTR and halve conversion rate. For more on how we run creative and buying on Meta, see our Meta ads services.

What red flags should end the conversation, and what goes on your checklist?

End the conversation if the agency cannot explain its measurement, guarantees a ROAS before seeing your data, or has no opinion about your feed and creative. Those three cover most bad outcomes we see when brands come to us after a poor agency experience.

Red flags specific to performance agencies

  • Guaranteed ROAS or revenue numbers in the pitch. Nobody can promise a return before auditing your tracking, margins and feed.
  • Case studies that only quote platform ROAS. Ask what happened to the client's total revenue and new customers.
  • Reporting that excludes brand search and retargeting share. If a large part of "results" comes from people already searching your name, you need to see that.
  • No opinion about your feed or your creative. If they have looked at your account and only talk about bids, they are not looking at the levers that matter now.

For the general questions that apply to any agency (team, communication, references, onboarding), use our list of questions to ask a digital marketing agency. The checklist below is the performance-specific layer to add on top.

Performance marketing agency checklist

Try this now

Run a 20-minute test on any agency you are considering

  1. Give them view access to one ad account and your product feed before the call.
  2. Ask: "What are the first three things you would fix, and how would we know it worked?"
  3. Ask: "If Meta, Google and our shop disagree on revenue, which number do you report to us?"

Specific, account-based answers are a good sign. Generic answers that could apply to any client mean they have not looked, or do not know what to look for.

Frequently asked questions

Two or three is usually enough, as long as you run the same test with each: the same account access, the same feed and the same questions. Talking to more rarely adds information and makes the answers harder to compare side by side.

Industry experience helps, but the business model matters more. An agency used to ecommerce feeds and catalogue ads thinks differently from one built around lead generation, so check they have run accounts like yours, then judge them on the measurement, feed and creative answers.

Often yes, because one team can judge both channels against the same blended numbers and stop them from claiming the same sales. Separate specialists can work well too, but then someone on your side has to own the overall budget split and the single source of truth for revenue.

Yes, it is one of the most useful tests. An audit of your real account and feed shows how an agency thinks before you commit to a retainer. Compare how specific each audit is: findings tied to your campaigns and products are a good sign, generic best practice is not.

Look for one that proposes to start there. Fixing conversion tracking, deduplication and consent handling is usually the first job, because better signals improve how automated campaigns bid. Be wary of any agency that wants to scale spend before the data is trustworthy.

What to do next

Before you speak to agencies, gather the basics so you can test them properly: your monthly ad spend, your true revenue for the same period, your product margins, and view access to your ad accounts and feed.

  1. Calculate last quarter's blended MER and new customer CAC, so you have a baseline that no attribution setting can change.
  2. Set up view access to one ad account and your product feed that you can share with each agency before the call.
  3. Shortlist two or three agencies and run the 20-minute test above with each of them.

If you want a second opinion first, we offer a free written audit within 48 hours covering tracking, account structure and feed. You can use it to judge us, or to brief whichever agency you choose.

✓ Link copied