Resources
How to Choose a Marketing Agency
The questions to ask in the first call, the answers that should end it immediately, and how to run a paid pilot properly — with success criteria agreed in writing before anybody starts.
What should I ask in the first call?
Five questions, asked in the first conversation, tell you more than any case study on a website. The good answers are specific to your business; the bad ones are reusable.
- "What is my conversion rate, and how do you know?" A good answer names a figure and the method. A vague answer means they have not looked at your business.
- "What would you remove from my current plan?" Good consultants cut. A plan that only adds is a plan that has not been prioritized.
- "How would we know within 90 days whether this was working?" The answer should name specific metrics and a threshold. "We will monitor progress" is not an answer.
- "Who exactly will do the work?" If the answer is a team or an account manager rather than a person, ask who, and see whether they join the call.
- "Show me a report you gave a client that went badly." If every engagement is a success story, the reporting is curated. Being able to show failure is the strongest single trust signal available.
What are the warning signs?
| Warning sign | What it usually means | What to do |
|---|---|---|
| Guaranteed rankings or traffic figures | No legitimate practitioner can promise this | Leave. It implies short-term tactics with manual-action risk |
| "Secret techniques" | If it were defensible, it would be describable | Leave |
| Percentage of media spend as the fee | Their revenue grows when you spend more, regardless of results | Decline. Fixed fee for defined scope instead |
| Reporting via screenshots | They cannot see the data they would act on, so they are not acting on it | Ask for account access as a condition of engagement |
| Case studies with no named clients | Client permission issues are normal; total absence suggests invention | Ask for specifics. Anonymised detail is fine |
| Link volume as the headline metric | Volume is not quality, and is frequently how networks are detected | Ask what they will do about off-site, not how many links |
| Everything is included for one price | Scope varies enormously; a single price means something is being excluded | Ask what is explicitly out of scope |
| No mention of measurement or attribution | They are not measuring outcomes | Treat as a serious problem, not a detail |
| Pressure to sign quickly | Standard terms are being overridden | Insist on your own review period, however long it takes |
What do good answers sound like?
| Positive signal | Why it matters |
|---|---|
| They ask about your margin, not just your traffic | They understand the business, not just the channel |
| They want access to analytics, Search Console and server logs | They intend to work from evidence |
| They tell you what they would not do | Clear professional boundaries |
| They name the specific person doing the work | Accountability, and you can check their expertise |
| They explain their reasoning, not just the recommendation | The reasoning is what survives when circumstances change |
| They say the problem needs a conversation first | They are not selling a predetermined package |
| They push back when the goal is unrealistic | Willingness to disagree is a professional signal |
| They have a written process with named phases | Repeatable, rather than improvised per client |
| They discuss what happens when it does not work | Measurement culture rather than reassurance culture |
How do I run a paid pilot properly?
A pilot is the only reliable way to evaluate an agency, and the majority are run badly — usually with no written success criteria, which lets both sides interpret success differently afterwards.
The single rule: **agree success criteria in writing before anyone starts.** Not after, not in the final debrief. If the criteria are not written down before the work, the pilot proves nothing.
Agree success criteria in writing
Specific metrics, the baseline they are measured against, and the threshold that counts as success. Agree before the pilot starts.Agree how it will be measured
Including who owns the data and who can access accounts during the pilot.Agree what happens on failure
What you learn either way, whether you continue, and whether the fee is credited against a longer engagement.Review weekly during, not after
A pilot that ends with a single presentation has had no chance to be corrected.What should the contract say?
Most agency contracts are longer than they need to be and say less than they should. Six things must be explicit.
- Deliverables, named specifically. "Ongoing SEO" is not a deliverable. "Twelve optimized product pages per month, each including keyword research, brief, copy and internal linking" is.
- Scope boundaries, stated as exclusions. What is explicitly not included. Ambiguity here is where invoices grow.
- Fees, and what they exclude. Media spend, third-party tools, and travel should all be named as separate.
- Term and notice period. Prefer 30 days' notice from either side, and be suspicious of twelve-month lock-ins for a new relationship.
- Account and data ownership. Analytics, Search Console, ad accounts and content should be yours, with access granted to the agency rather than the reverse.
- How success will be reported. Metrics, frequency, format — so reporting quality is agreed before it is delivered.
One clause worth insisting on: if you leave, you keep everything — your analytics, your content, your ad accounts, your documentation. Agencies that resist this are building a dependency rather than a service.
Should I hire an agency or do it in-house?
| Situation | Better option | Why |
|---|---|---|
| No one internally owns marketing | Agency or fractional leader | You cannot buy tools and hope; someone must decide |
| Someone internal owns it well | In-house, with selective specialist support | Capability exists; do not outsource the decisions |
| One channel, well understood | In-house | Enough volume to be worth automating, small enough to manage |
| Four or more channels | Agency plus in-house owner | Specialisms multiply; nobody internally knows all of them |
| Content production is the bottleneck | Hybrid | Strategy in-house, production outsourced |
| Very small budget | In-house plus free resources | Agency fees become a disproportionate share |
| Urgent turnaround needed | Agency | Time cannot be hired for a deadline |
The failure mode on both sides is the same: neither the agency nor the internal team owns the outcome, so nobody does. Whatever you choose, name one person who is accountable for results — and if that is an agency, hold them to it.
Frequently asked questions
How do I know if a marketing agency is any good?
Ask five questions in the first call: your conversion rate and how they know it, what they would remove from your plan, how you'd know within 90 days whether it worked, who exactly will do the work, and to see a client report that went badly. Good answers are specific to your business. Bad answers are reusable, which is the tell.
Should I sign a 12-month contract with an agency?
Be cautious. A twelve-month commitment before either side has worked together once puts all the risk on you. Prefer an initial short term with 30 days' notice from either side. If you want longer commitment, make it conditional on performance criteria agreed in writing at the start rather than unconditional.
Is it bad if an agency wants a percentage of my ad spend?
Yes, and you should decline. Their revenue then grows when you spend more, regardless of whether it worked. That is a misaligned incentive and it is the opposite of what you want from someone advising on budget. Fixed fee for defined scope, with media spend paid directly to the platform, keeps incentives aligned.
How do I evaluate a pitch when I don't know enough to judge the tactics?
You do not need to know the tactics. Evaluate the reasoning: does the pitch demonstrate an understanding of your economics, does it say what it would remove, does it specify deliverables, and does it come with a process rather than a list of activities? A strong proposal from someone who explains their reasoning is safer than a weak one from someone who promises more.
Should I run a pilot before committing?
Yes, and if you do, agree success criteria in writing before the work starts. If the criteria are not written down first, the pilot proves nothing and both sides will interpret the result differently afterwards. Review weekly during the pilot rather than in a single debrief at the end.
What if an agency won't show me a client account?
Client confidentiality is legitimate and they may decline. But they should be able to describe a specific engagement without naming the client, show the shape of their reporting, and discuss a case that did not go well. A blanket refusal to give any detail at all is different from protecting a name, and is worth asking about.
Can I work with an agency that also works with my competitors?
Often yes, and it is normal for specialists. What matters is whether they share your data, your strategy or your learning between accounts. Ask directly. A specialist working across several clients in the same market often produces better work, because they see patterns you cannot — but the boundary on confidential information should be explicit rather than assumed.
What if the agency is doing work I don't understand?
That is a problem worth fixing rather than accepting. You should be able to understand the reasoning behind each recommendation — not the implementation detail, but the logic. If you cannot follow why something is being done, you cannot evaluate whether it is working, and you cannot make the decisions when circumstances change. Ask them to explain it; if they cannot, that is a serious concern.
Keep reading
Want this applied to your business?
Stop guessing what to prioritize. Get a straight answer on where your marketing is leaking money, and what to fix first.
Book a free 30-minute audit