How to Choose a Digital Marketing Agency
The short answer
Ask three things before you sign anything: who will actually do the work each month, what your business owns at the end of it, and which two or three numbers both sides agree will decide whether it worked. Nearly every agency relationship that goes wrong went wrong on one of those three, left vague at the start. A pitch deck answers none of them.
Most businesses choose an agency from a pitch. Which is awkward, because the pitch is the one piece of work every agency is genuinely excellent at. It has been made more times than anything else they produce, sharpened over years, and it is usually built by their strongest people — none of which tells you who will be writing your captions in March.
So the decision gets made on a deck and reviewed on the third month's report, by which point the money is gone and the relationship is already awkward to leave.
What follows is not a list of things to look for. It is a list of things to ask, which is a different exercise, because the answers are much harder to prepare in advance.
What should you ask a digital marketing agency before hiring them?
Three things, in this order: who is actually doing the work, what your business owns at the end of it, and how both sides will know whether it worked. Almost everything that sours an agency relationship later was one of those three, left vague at the start because asking felt impolite.
The nine questions below sit under those three headings. None of them is clever, and that is the point. The useful questions in this process are boring ones that are difficult to answer smoothly.
Who does the work
- Who, by name, is on this account every month, and what else are they carrying?
- Which parts are done in-house and which are passed to somebody else?
- When something is wrong, do we speak to the person doing the work or to a manager between us?
What you own
- Whose ad account does the spend run through?
- Who holds the domain, the hosting, the analytics and the source files — during, and after?
- Do we receive editable files or only exports?
How you will know
- What exactly arrives each month, and what counts as a revision rather than a new request?
- Which numbers are you willing to be judged on?
- What does ending this look like, and what do we receive on the last day?
Who is actually going to do the work?
Ask for names, and ask early. The people who present are frequently not the people who deliver, and there is nothing improper about that so long as you learn it before signing rather than in week two.
Then follow it with a question that sounds rude and is not: how many other accounts is that person carrying? You will not get a precise number and you do not need one. Watch how the question lands instead. An agency that has thought about capacity has an answer ready; one that has not goes vague, and vagueness about capacity is the most reliable early predictor of slipping deadlines there is.
Ask what gets subcontracted too. Video editing and photography commonly are, sometimes performance media. Subcontracting is ordinary and often the right call. Undisclosed subcontracting is the problem, because it quietly inserts somebody you have never met into your approval cycle and adds a day or two to every round of changes.
Whose accounts is this running on?
The question people skip, and the most expensive one to skip.
Ad spend should run through an ad account owned by your business, with the agency given access to it — not through the agency's account with you named on an invoice. Several things become true at once when it is the wrong way round, and none of them get mentioned at the start. The performance history belongs to them. The audiences and the pixel data belong to them. If you leave, you restart from nothing on a platform that specifically rewards accumulated history. Set up properly, changing agencies costs you a month of onboarding. Set up badly, it costs you everything the account has learnt since day one.
The same logic runs through everything else. The domain should be registered to the business, not to whoever happened to buy it. Analytics should live in a property you own. Design files should come back editable rather than flattened, and that is far easier to agree in a proposal than to negotiate during a handover that is already tense.
There is a clean test for all of it. Ask what happens to each item on the day the contract ends. Any answer containing the phrase we would have to look into that is worth looking into now.
If the agency pays the platform bills on their own card and invoices you afterwards, that is workable. Plenty of agencies do it, and it saves everybody a failed payment at two in the morning. Just ask to see the platform's own invoice alongside theirs. An agency doing it honestly will not mind at all.
What does "we will handle your social media" actually mean?
Nothing at all, until somebody itemises it.
Scope disagreements end more agency relationships than bad creative does, and they nearly always start the same way: both sides used one phrase to describe two different quantities of labour. Twelve posts a month sounds specific. It is not. Does it include the copy, the design, the shoot, the scheduling, replying to comments? Are stories counted in the twelve? Is a carousel of eight slides one post or eight?
Settle it in the proposal. Settle what a revision is while you are there, because the distance between make the logo bigger and actually, could we try a completely different concept is several hours of somebody's day, and the argument about which one just happened is a miserable one to have in month four.
Volume as the headline of a proposal is itself a mild warning. Thirty posts a month is a manufacturing figure, not a marketing plan. It is easy to price, easy to compare across three quotes, and close to unrelated to whether anything gets sold. A proposal built around what the business actually needs looks less tidy in a comparison table and is usually worth considerably more.
Which numbers should an agency report?
The ones connected to money, plus the ones that explain them. Reach, impressions and follower growth are not worthless. They are simply very easy to move for reasons that have nothing to do with your business, which is exactly why they fill the report during the months when the other numbers are not going well.
Before the first invoice, agree the two or three that genuinely decide this. Enquiries, qualified leads, bookings, orders, the cost of getting one. Then agree how each gets counted, because a lead that is a phone call, a lead that is a filled form and a lead that is somebody replying to a story with a fire emoji are three different things, and only one of them pays for anything.
Then ask the question that actually separates agencies: which of these are you willing to be judged on? Watch what gets negotiated away, and how quickly. That conversation predicts the next year better than any case study will.
Agency, freelancer or in-house, compared
| Freelancer | Agency | In-house hire | |
|---|---|---|---|
| Suits | One discipline, clearly specified | Several disciplines that must agree with each other | Work that is continuous and central to the business |
| Cost shape | Lowest per piece, hardest to forecast | Fixed monthly, predictable | Highest fixed cost, cheapest at volume |
| Range | One skill, occasionally two | Several under a single brief | Whatever you hired for, and no more |
| Continuity risk | High. One person, one illness, one better offer | Lower, though the individual can still change | Lowest, right up until they resign |
| A small urgent change | Fastest | Slower, there is a process | Fast |
| Outside perspective | Some | Most | Least, and it fades |
| Where it usually fails | Anything beyond that one skill | Sliding into order-taking | Nobody left to say the idea is bad |
Most businesses end up running a combination, which is sensible rather than indecisive. The recurring mistake is a mismatch: paying agency rates for a single well-defined deliverable a freelancer would have done better, or handing a freelancer something that quietly needed four people to stay in step.
What are the warning signs?
Six, each common enough to be worth naming plainly.
A guarantee. Nobody controls Google's rankings, so nobody can promise a position in them. A guaranteed follower count is worse, because that one is achievable, and the only way to achieve it is to buy accounts that will never purchase anything from you.
No questions about the business. An agency that can produce a quote before finding out what you sell and who buys it is quoting from a rate card. The proposal will be competent and generic in precisely the proportion that the conversation was.
A portfolio with no roles attached. Handsome work, presented with no indication of which parts the agency did.
Everything on WhatsApp and nothing in writing. WhatsApp is excellent for speed and a poor place to store an agreement about scope. Both are true at once.
Any hesitation about access. If owning your own ad account is treated as an unusual request, you have just learned what happens to clients who leave.
Chasing. If it takes four follow-ups to get a proposal out of them while they are still trying to win you, picture month seven.
How do you tell whether a portfolio is real?
Ask which parts they did. That is the whole method, and it works because the honest answer is easy to give and the dishonest one is not.
Who shot the photography. Was the strategy yours or the client's. What did this look like before you started. Did you build the website or design it for somebody else to build. None of that is hostile — it is the same question a good agency puts to its own prospective hires — and it is remarkable how quickly a padded portfolio thins out under it.
Then ask for one project that did not go well, and what changed afterwards as a result. That answer is worth more than five successes. An agency that has never had a difficult project has either not been working long enough or is not going to be straight with you, and you would rather find out which one now.
What to settle before you sign
Six items, in writing, inside the proposal itself rather than in a follow-up email:
- Names. Who is on the account, and who covers when that person is away.
- Itemised scope, with a revision defined in a sentence both sides would read the same way.
- Ownership of accounts, domain, analytics and files, stated rather than assumed.
- The two or three numbers that define success, and how each one is counted.
- What you owe them. Approval turnaround, access, information, somebody empowered to say yes. Campaigns miss dates because clients miss approvals far more often than anyone in this industry admits publicly.
- Notice period and handover. Specifically, what arrives on the last day.
Not one of those requires trust, which is the point of them. They are what makes trust unnecessary at the stage where you do not have any yet.
Applying this to any agency, including this one
Two things published on this site are answers to questions on that list, given before anybody asks. The Sequence is the studio's process written out as numbered stages, so the shape of the work is visible before a contract exists rather than after it. The facts page states plainly who the team is, by name and role, alongside what the studio does and does not do, which covers the portfolio question and the who-is-doing-the-work question in advance.
Run the rest of the list against this studio too. Ask about ownership. Ask which parts of the published work were done here. Ask what the reporting will contain. If any of it fails to hold up, that is genuinely useful information, and far better arrived at now than in month four.
If you are earlier in the process than this and still working out what to commission, brand identity vs logo design covers the most commonly misread line item on a proposal, and why a business needs a website covers the next one. You can see what the studio actually offers in The Craft, or start a conversation and put the nine questions to us directly.