Judge an agency's onboarding by how fast it learns

Most agency onboardings are the same. A kickoff call, a fortnight of access requests, then a deck with a 90-day plan on it: audit in month one, new campaigns in month two, optimisation in month three. It looks thorough, so everyone nods.

If you’re the client, that plan is a list of activities. Nothing in it says what the agency expects to learn, or when. The team's rate of learning is the better test: how quickly it works out the most important things about your business, starting from what you've already learned.

You write a script and pass it off as a playbook

When you run an agency, the easy thing is to plan around the team you've already got. You decide exactly how you'll use them and in what order. You write a script and pass it off as a playbook. Then you tell the client a playbook is what they need.

If the numbers haven't moved a month in, the answer's ready: "This was always a journey. Once we finish the playbook, that's when we'll have the learnings."

It's an easy way out. A fixed plan is easy to staff and easy to bill. It keeps costs down, and it spares the agency the communication overhead and the variance a new client brings. Most agencies aren't equipped to work any other way, and frankly, they don't want to. If you've been burned by an agency before, there's a good chance this is why.

It's very important to be thorough in onboarding: to map the universe of things you could do, and to give your own stakeholders a direction they can explain upwards. But planning in isolation from results is a silly idea. Once the plan exists, you can't put your head down and plough through it.

The KPI is the rate of important learnings

The right way to run an onboarding is to treat it like work from day one, and to KPI it on the rate of important learnings.

Anyone can launch a test quickly. Plenty of agencies don't, and that's a problem. But a quick test on something that doesn't matter only keeps the team busy. The skill is choosing what to learn first.

We score every hypothesis on its potential impact and on how hard it is to run. Then we balance the list. Some tests build on what already shows promise. Some go after the opportunities that what we know points to. Some go into white space nobody has tested yet, because it has high potential too. It's very easy to get lost on things that could have low impact.

One financial services app came to us wanting paying subscribers. The first thing we tested was their brand messaging. Within a week, we'd shown it didn't land with the people they were targeting. Within two weeks, we'd validated a new positioning that did, and that became the platform for hitting the subscriber goal by month three. The plan changed because the result did.

Look at the first month of any plan and ask which tests would change what you do next, whether they win or lose. Those should go first.

The first test doesn't have to be new

Your account already holds a lot of learning: campaigns that worked, landing pages that didn't, creative you've paid to test. Most agencies get access to all of it without understanding what's been learned, so they start the learning process all over again.

If you've switched agencies before, you can see it in the ad account. A set of campaigns gets paused and tagged "former agency", and a new set goes live beside it. You don't have to build things from scratch just to show you've taken over.

We worked with an e-commerce client that had only ever advertised in the UK. The plan we'd discussed was all UK: new segments, new creative, new funnels. Within the first week and a half, we tripled their revenue by running the same ads and the same funnels in the US. They'd assumed it was a different market with different audiences, and there's probably nuance to that. But it's the clearest kind of low-hanging fruit: something nobody had learned yet that took very little work to try.

Small wins stack up. Build tests on what's already running as well as the bigger swings, and hand over the tests that failed along with the logins.

Time to first test is an early signal

There's no excuse not to be faster. There's no reason your agency shouldn't know everything there is to know about your organisation within a couple of days.

In my experience there's a really high correlation between starting to move quickly and keeping a client. It depends on the client, and big companies with long approval processes move slower. But how a team runs its first month is usually how it runs the whole relationship.

Say a test takes two weeks to read (an illustrative figure). A team with its first test live in week two gets about six rounds of learning in its first 90 days; a team that starts in week eight gets two or three.

So time to first test is worth watching. It's an early signal: six rounds only count if they're spent on the things most worth learning.

Write down the three most important things you've learned about your own marketing in the last year, failures included. Then look at the plan you've been sent, and check whether it starts from them or from scratch.

‍