Blog: Stop Hiring Agencies to Run Channels. Hire Them to Answer One Question.
Key Takeaways
- The job of a marketing agency has changed, and most evaluation processes haven't caught up.
Running paid search or paid social well used to be the thing worth paying a premium for. Google and Meta have automated a lot of that work, so tactical execution is now the floor rather than the differentiator. It still matters, but if your agency search is built entirely around channel questions like "who manages accounts best?", you're shopping for something you can largely take for granted and skipping the part that's actually hard. - Platform-reported metrics tell you which channel got credit for a sale, not whether the sale needed the ad.
Branded search looks like the best performer in almost every account, because anyone typing your company name into Google was usually already on their way to buying. The channel collects demand the rest of your marketing created and reports it as a win. Silverback cut 42% of one client's branded search budget and ran it as a controlled test in matched markets. Organic purchases in those markets went up 61%, which told us most of that spend had been buying customers who were already coming. - Brands should expect their marketing agency to run the channels and prove what they're worth.
Those aren't competing skills, and you shouldn't have to pick. Ask any firm you're considering how it reports marketing's impact on the business, what it looks at when deciding to move budget between platforms, and how it optimizes toward revenue instead of leads. You want an answer that covers both the account work and the evidence behind it: what a customer is worth, how they'd test whether a channel is earning its keep, and what they'd do if the test came back ugly.
Table of Contents
Most companies use a flawed set of decision criteria to select a marketing agency.
The search starts with a scope built around channels:
- Who can run paid search?
- Who can run paid social?
- Who has the cleanest account management process?
Finalists get judged on their tactical polish and platform-reported ROAS numbers in their case studies, and the team with the most confident answer to “how will you manage our accounts” usually wins. Every step of that selects for channel competence, which is the part of the job of a media buyers which platforms have spent 10+ years automating.
The thing actually worth paying for sits a level above an agency’s ability to pull channel levers, and it rarely comes up until the contract’s signed and a year has passed without the business moving.
What you actually get when you hire an agency to run a channel
Hire an agency to manage paid search and you get someone who manages paid search well. They’ll structure the account, write the headlines, tune the bids, and send a monthly report with strong ROAS on branded terms and a healthy platform-reported cost per lead. Every figure will look defensible. Almost none of it tells you whether the channel produced revenue you’d have missed without it.
Platform-reported metrics show you which channel got credit for a conversion. They say nothing about whether that conversion would have happened anyway. Branded search is the clearest case. It posts the highest return in nearly every account because every sale that touches a branded keyword gets counted as its win. But the person typing your company name into Google was usually already going to buy. The channel collects demand that other work created and reports it as its own.
An agency hired to run that channel will run it well. They’ll drive the cost per conversion down quarter after quarter and show you the trend. They’re doing the job you defined. The job you defined just doesn’t have much to do with whether the company is growing.
The one question to build your agency evaluation around: Did marketing drive net-new sales?
A marketing investment has to answer one question: did our marketing drive net-new sales we wouldn’t have gotten anyway? Impressions, clicks, leads, and platform-reported ROAS are proxies that may or may not point at that answer.
A partner who works this way starts with your economics. They ask how you define a profitable customer, what your close rates look like, and which conversions in your current reporting are real incremental wins versus demand you’d have captured for free. They treat every channel as a way to produce incremental revenue and stay loyal to none of them. When the data shows a channel taking credit it didn’t earn, they tell you, and they recommend cutting it, even when it’s the channel they were hired to run.
We saw this with a client running heavy paid brand search campaigns with a reported ROAS that made the spend look untouchable. We thought the budget was harvesting demand rather than creating it. We cut 42% of their brand search spend through a matched market geo-lift test. Organic purchases rose 61% in the test markets. An agency hired to manage paid search would never have proposed cutting brand search. As a partner hired to grow the business, we had no choice.
3 questions that separate measurement partners from channel operators
The questions you ask during evaluation decide what kind of firm you end up with. Most evaluations test for channel fluency. Instead, you should test for marketing measurement expertise. A few questions separate partners from operators quickly:
- How do you report on marketing’s impact on business performance?
- What KPIs do you use to make budget allocation decisions across platforms?
- What is your approach to optimizing campaign performance based on revenue?
Operators steer back to tactics like account structure, audience strategy, and creative cadence. That work matters, but the platforms increasingly handle it. The firms worth hiring talk about setting KPIs based on business unit economics, measurement design, incrementality testing, and feeding qualified lead and revenue data back into the platforms. They’ll tell you plainly that some of your current spend probably isn’t working, and they’ll explain how they’d prove it.
What an agency’s pitch tells you about how it will work
The opening of a pitch tells you what a firm thinks it’s selling. When the slide after the introduction is a wall of platform-reported ROAS and cost-per-lead trends, you’re being sold channel reporting. When the team opens by asking what a customer is worth to you and how you currently know whether marketing is working, you’re talking to a firm built around the right question.
How a marketing agency handles uncertainty is the other signal. Operators promise to improve the metrics they control. Real measurement partners will warn you that the first honest incrementality test usually delivers an uncomfortable result, and that finding those results before your CFO does is the whole point of the engagement. That candor is the value, and it’s worth more than a flawlessly tuned account aimed at the wrong outcome.
You’re buying an answer
The agencies worth their fee from here on are the ones who can stand in front of your finance team and show, with evidence, that the marketing produced sales the company wouldn’t have made otherwise.
That’s what your evaluation should be built to find. Judge every firm in the room against the one question that decides whether the budget is worth spending at all, and you’ll stop selecting people to run channels and start selecting people to grow the business.
FAQs
What is an incrementality test in marketing?
An incrementality test measures whether a marketing channel is creating sales that wouldn't have happened without it. The usual setup is a matched market test: you turn spend up or down in one group of similar cities or regions, hold it steady in a comparable group, and compare total business results across the two. Media spend is the only thing that changed, so the gap between the groups is what the channel actually contributed. It's a different question from the one attribution answers. Attribution tells you which ad got credit for a sale. Incrementality tells you whether the sale needed the ad at all.
How much conversion volume does a business need before an incrementality test is worth running?
Enough that a real effect will show up above ordinary week-to-week noise. A company doing a few hundred conversions a month, spread across enough geographies to build two comparable groups, can usually get a clean read on a major channel in four to six weeks. Below that, the test still works, it just needs more time, larger regions, or a bigger swing in spend so the effect is easier to see. Smaller advertisers often do better with simpler versions: hold out a handful of markets, pause one campaign type entirely, or read the before-and-after on a budget change you were already planning. What limits most companies is patience, not size.
Will competitors take the traffic if a company cuts its branded search spend?
Sometimes, which is why it's worth testing rather than arguing about. In some categories, competitors bid hard on your company name and pulling back genuinely costs you sales. In others, nobody is bidding at all and you're paying for a click that would have landed on your free organic listing a half inch below. The way to find out is to cut spend in a subset of markets, leave it running everywhere else, and watch total revenue instead of paid clicks. One Silverback client cut 42% of branded search spend in a matched market test and saw organic purchases rise 61% in the test markets, meaning most of that budget had been buying customers who were already on their way in. Another company's numbers could come back the other direction, and that's the whole reason to run it.
Are platform-reported metrics like ROAS and cost per lead still useful?
Yes, for the job they were built to do. Data from Google Ads, Meta, and similar platforms is the right tool for managing campaigns day to day, catching things that break, and comparing keywords, audiences, and creative inside a single channel. It stops being trustworthy the moment it's used to decide how much of the total budget each channel deserves, because every platform counts the conversions it touched and none of them can tell you what would have happened without the ad. A reasonable split: platform metrics for running the channels, incrementality tests and media mix modeling for deciding where the next dollar goes.
Should a company build marketing measurement in-house or hire an agency to do it?
Either can work, as long as the company owns the standard. What in-house teams usually lack isn't intelligence, it's repetition: a test design that holds up when the finance team pushes on it, enough experiments run to know what a normal result looks like, and enough distance from the media buying to say plainly when spend isn't working. A firm running tests across dozens of accounts builds that faster. The arrangement to avoid is the one where an agency buys the media and then grades its own performance using platform-reported numbers. Whoever ends up running the campaigns, the measurement approach and the reporting behind it should be something a CFO can inspect.
