Questions to Ask Before Hiring a Digital Marketing Agency

Hiring a digital marketing agency is a decision about who will spend your money, touch your data and speak for your brand for months at a time. The questions to ask a digital marketing agency before you sign are the cheapest protection you have: an hour of structured interviewing costs far less than six months with the wrong partner, and it surfaces most of the problems that otherwise appear in the third monthly report.

This guide is for business owners, marketing leads and in-house teams comparing agencies for paid media, conversion rate optimization (CRO), or both. It is organized as an interview guide. Each question comes with why to ask it, what a good answer sounds like and what a bad one sounds like, grouped by theme: fit and process, craft and methods, team, measurement and reporting, contracts, rights and pricing, and aftercare. A scorecard near the end lets you compare suppliers side by side.

The aim is not to catch agencies out. Good agencies welcome hard questions because the answers show their work. The aim is to find out, before money moves, whether the agency's way of working matches your traffic, your data, your team and your budget.

Why the choice matters more than the pitch suggests

Agency pitches tend to look alike. Most show a case study, a process diagram and a team slide. What separates them is invisible in the deck: how they handle a test that cannot reach significance, whose numbers they report against, who actually does the work after the senior people leave the room, and what happens to your accounts and data if you part ways.

Three features of digital marketing make the choice unusually consequential. First, the constraints are hard. Traffic volume is the limit on how fast you can learn from testing, and no amount of budget changes that. An agency that promises weekly wins on a low-traffic site is either misunderstanding the math or planning to report noise. Second, measurement is genuinely difficult. Multiple platforms, offline conversions and consent requirements make attribution hard and expensive to get right, and every ad platform will happily claim credit for the same sale. Third, the work compounds. Research informs tests, tests inform the next round of research, and paid media learns from reconciled conversion data. Switching agencies midstream throws much of that away.

Before any interview, it helps to know what you want the agency to achieve. If your goals are still vague, our article on setting marketing goals is a useful first step, and if you have not yet settled whether the work belongs outside the business at all, read how to get agency or in-house right first.

The short list: the questions that matter most

If you only have time for a handful of questions, ask these. Each is covered in more depth in the themed sections below.

1. What would you need to know about our traffic before promising anything? It shows whether the agency understands that traffic is the hard constraint.
2. Whose numbers will you report against? The right answer is your own orders or leads, reconciled with platform data, not platform data alone.
3. Who will do the work each week, and how much of their time is ours? It separates the pitch team from the delivery team.
4. Who owns the ad accounts, analytics properties, test code and research? You should, from day one.
5. How do you get winning tests shipped? Without development capacity, a testing program produces documents, not results.
6. What does the first six months look like, week by week? A credible plan starts with a data audit, not with ads or tests.
7. How do we leave? Notice periods, handover and access should be settled before you start.

Fit and process: can this agency work the way your business needs?

Fit questions establish whether the agency's model matches your situation before you spend time on craft. An agency can be excellent and still wrong for you: too large, too small, built for high-traffic retail when you sell to a handful of enterprise buyers, or built for lead generation when you need e-commerce.

What would you need to know about our business before making any recommendations?

Why ask it: An agency that recommends tactics in the first meeting is selling a template. The answer shows how the agency diagnoses before it prescribes, and whether it asks about traffic, margins, funnels and markets.

A good answer sounds like: The agency asks for access to analytics and order data, asks about monthly traffic and conversions per funnel, wants to know how many markets and funnels you run, and asks what engineering time you have. It explains that each funnel and market is its own research, its own tests and its own sample size problem.

A bad answer sounds like: The agency already knows what you need: a new landing page, more budget on a particular platform, or a set of tests it has run for other clients. It does not ask about traffic at all, or treats it as a detail to sort out later.

What does the first six months look like?

Why ask it: The sequence of work tells you more than any promise. Rushing to launch tests or campaigns before the data is trustworthy wastes the first months.

A good answer sounds like: A plan that starts with a data audit of around 1–2 weeks, moves into research of around 3–5 weeks, gets a first test live somewhere around week 5–7, and treats each test as running 2–6 weeks, set by traffic and not by preference. The agency says meaningful program results take around 4–6 months and explains why.

A bad answer sounds like: Tests live in week one, a promise of results within a month, or a plan with no audit because the agency will "fix tracking as we go." Another bad sign is a plan identical to the one in its sales deck, with no reference to your data.

What kinds of clients do you do your best work for, and which do you turn away?

Why ask it: Every agency has a sweet spot. Asking which clients it turns away is the fastest way to find out whether you are inside it.

A good answer sounds like: A specific description: the traffic levels, categories, business models and budgets where the agency's methods work well, and a clear statement of where they do not, such as sites with too little traffic for testing, where it would recommend research and qualitative work instead.

A bad answer sounds like: "We work with everyone." Or a list of logos with no explanation of what the work involved or why it suited those businesses.

How will you work with our team day to day?

Why ask it: Many agency relationships run into trouble on process rather than skill. You need to know how requests, approvals and feedback move, and how much of your team's time the agency will need.

A good answer sounds like: Named contacts on both sides, a regular meeting rhythm, a shared place for requests and decisions, clear approval steps for creative and test launches, and an honest estimate of how much of your time the agency will need each week. The agency asks about your marketing calendar so launches do not collide with promotions.

A bad answer sounds like: "We'll handle everything, you won't need to do anything." No agency can run tests or campaigns without your approvals, your data and your product knowledge.

Craft and methods: how does the agency actually do the work?

Craft questions test whether the agency's methods are sound. You do not need to be a specialist to ask them; you need to listen for specifics, trade-offs and the willingness to say "it depends" and then explain what it depends on.

How do you decide what to test or change first?

Why ask it: Prioritization is where research turns into results. An agency that tests whatever a senior person believes will waste traffic on low-value ideas.

A good answer sounds like: The agency describes research that combines analytics, session review, expert review and user testing, and a way of ranking findings by expected impact, confidence and effort. It can show an anonymized example of a ranked findings list and explain why the top items were chosen.

A bad answer sounds like: "We start with best practices." Or a fixed list of test ideas that the agency runs for every client, such as button colors and headline swaps, regardless of what the research says.

What happens when our traffic is too low to test?

Why ask it: This question separates agencies that understand statistics from those that do not. Low traffic means tests take longer or cannot reach significance at all.

A good answer sounds like: The agency explains how it calculates the sample size and duration for a test before running it, and what it does when the numbers do not work: testing bigger changes, testing higher in the funnel, combining pages, or relying on research and qualitative evidence to make changes without a formal test. It says plainly that some sites should not run a testing program yet.

A bad answer sounds like: "We'll just run it for longer." Or calling a test after a few days because one variant is ahead. Peeking at results and stopping early is one of the most common ways programs report wins that are not real.

How do you manage paid media against conversion data?

Why ask it: Ad platforms optimize toward whatever conversion signal you give them. If that signal is wrong, the platform gets very efficient at the wrong thing.

A good answer sounds like: The agency manages search, social and shopping against reconciled conversion data, meaning platform-reported conversions are checked against your own orders or qualified leads. It explains how it handles offline conversions, how it separates brand from non-brand search, and how it tests creative.

A bad answer sounds like: Reporting platform-reported return on ad spend as the result, with no reconciliation against your own numbers. Or optimizing to clicks or cheap leads without asking which leads turn into customers.

How do you get tests built and winners shipped?

Why ask it: Tests have to be built and winners have to be shipped. Without engineering time you get a document. This is the most common reason CRO programs stall.

A good answer sounds like: The agency explains who builds tests (its own developers, yours, or a mix), what testing tool it uses and how it handles quality assurance across devices and browsers. It asks about your release process and agrees how winning variants move from the testing tool into your production code, rather than being left running in the tool indefinitely.

A bad answer sounds like: Vague answers about "handing recommendations to your developers," or a program where winners stay inside the testing tool for months because nobody owns shipping them.

If CRO is the bigger part of your brief, our companion guide on questions to ask before hiring a CRO agency goes deeper on research and experimentation methods.

Team: who will actually be doing the work?

Team questions close the gap between the people who pitch and the people who deliver. The pitch is often led by senior staff whose time on your account after signing is limited. That is normal, but you should know it in advance.

Who will work on our account each week, and what share of their time is ours?

Why ask it: An agency can only give you the attention it has scheduled. Knowing names and allocation lets you judge whether the team is big enough for the scope.

A good answer sounds like: Named people with roles, an honest description of how many other accounts each person handles, and which specialists (analysts, developers, designers, researchers) are shared across clients and how their time is booked. The agency offers to introduce the day-to-day team before you sign.

A bad answer sounds like: "A dedicated team" with no names, or an organization chart full of titles that turns out to be one account manager passing work to freelancers you will never meet.

What happens if a key person leaves?

Why ask it: Staff turnover at agencies is a fact of life. The question is whether knowledge about your account lives in people's heads or in documents.

A good answer sounds like: Written documentation of your accounts, tests, decisions and learnings, a named backup for each key role, and a defined handover process with notice to you when someone changes.

A bad answer sounds like: "That won't happen." Or a pause while a new person relearns your account from scratch.

Do you subcontract any of the work?

Why ask it: Subcontracting is common and not a problem in itself, but it affects quality control, confidentiality and who has access to your data.

A good answer sounds like: A clear list of what, if anything, is subcontracted, to whom, under what confidentiality terms, and how the agency reviews that work before it reaches you.

A bad answer sounds like: Discovering after signing that your paid media, development or research is handled by a third party you did not know about.

Measurement and reporting: whose numbers count?

Measurement questions matter most in paid media and CRO, because the work is judged on numbers, and there are many ways to make numbers look good. The core question is simple: will the agency report against your reality or against the platforms' view of it?

What will you check in our data before you start?

Why ask it: If tracking is broken, every decision built on it is suspect. A data audit is the foundation, and it is typically where the first 1–2 weeks go.

A good answer sounds like: The agency lists what it will check: whether conversion events fire correctly and only once, whether analytics numbers match your order or CRM records within a reasonable tolerance, how consent affects what is recorded, whether offline conversions are imported, and whether ad platforms receive the right conversion signals. It commits to sharing the findings before building on them. Our guide to auditing the marketing stack covers the same ground from your side.

A bad answer sounds like: "Your tracking looks fine" without having looked. Or a plan to install a new tag manager or tool as the first step without understanding the one you already have.

Which numbers will you report, and how will they reconcile with our own?

Why ask it: Each ad platform tends to count conversions in its own favor. Added together, platform-reported conversions often exceed the orders you actually received.

A good answer sounds like: Reporting against your own orders or qualified leads, with platform figures shown alongside and the gap explained. The agency agrees the definitions in writing: what counts as a conversion, which attribution model is used, and which date range and time zone the numbers follow.

A bad answer sounds like: Screenshots of platform dashboards presented as results. Or a reporting deck where every channel claims credit and nobody explains why the total is larger than your revenue.

What will a monthly report look like, and who is it for?

Why ask it: Reports are for decisions. A report that nobody can act on is a cost, not a deliverable.

A good answer sounds like: A short report that states what was done, what was learned, what changed in the numbers that matter, and what happens next, with a plain-language summary for leadership. The agency shows an anonymized example and asks who reads your reports. If you report upward, it can shape the summary to match; see our guide to reporting marketing to boards.

A bad answer sounds like: A long deck of charts with no conclusions, or a report that changes metrics from month to month depending on which ones look best.

How do you report tests that lose or come out flat?

Why ask it: Many tests in a healthy program will not produce a clear winner. How the agency reports them reveals whether it is learning or selling.

A good answer sounds like: Losing and inconclusive tests are reported as clearly as winners, with what was learned and how it changes the next round of research. The agency describes its rules for calling a test before it starts, including sample size and duration.

A bad answer sounds like: Only winners appear in reports. Or the agency reports projected annual revenue from each winning test without saying how it was calculated or whether the result held up once shipped.

If you want a shared view of the numbers rather than monthly decks, agree up front on a dashboard both sides use. Our article on marketing dashboards explains what to include and what to leave out.

Contracts, rights and pricing: what are you actually buying?

Commercial questions are where most avoidable disputes begin. Ask them early, in writing, and compare the answers with the draft contract rather than the proposal.

Who owns the ad accounts, analytics, test code, creative and research?

Why ask it: If the agency owns your ad accounts or analytics properties, leaving means losing years of history and learning. Ownership should never be ambiguous.

A good answer sounds like: You own every account and property, created under your business and with the agency given access as a user. Creative, test code, research findings and reports produced for you are yours to keep and reuse. The agency may keep its own general methods and tools, and says so plainly.

A bad answer sounds like: Accounts created under the agency's own business manager or login, or contract terms that give the agency ownership of creative or research until all invoices are paid in full or indefinitely.

How is your pricing structured, and what does it include?

Why ask it: Pricing structures change incentives. A percentage of spend rewards higher spend; a flat retainer rewards efficiency but can drift out of line with scope. You need to know what the fee covers and what costs extra.

A good answer sounds like: A clear breakdown of fees by service, with what is included and excluded, and how the fee changes if scope or spend changes. The agency explains how its rates compare with typical market ranges and what drives cost: traffic volume, number of funnels and markets, research depth, development capacity, media spend under management and measurement complexity.

A bad answer sounds like: A single monthly number with no breakdown, fees that rise with spend without any explanation of extra work, or add-on charges for reporting, creative or tooling that appear after signing.

What are the notice period, minimum term and exit terms?

Why ask it: Programs take time to produce results, so some commitment is reasonable. But you should know exactly how to leave.

A good answer sounds like: A minimum term that reflects how long the work takes to show results, a reasonable notice period, and a written handover that includes account access, documentation, test history and open work.

A bad answer sounds like: Long lock-ins with automatic renewals, exit fees, or no mention of handover at all.

What typical pricing looks like

Prices vary with scope, but it helps to know typical ranges before you compare proposals. The table below sets our published starting rates next to reviewed typical US market ranges. Our rates are starting prices, not guaranteed totals; the final figure depends on the cost drivers above.

A worked example: the first six months at starting rates

This is an illustrative scope, not a quote. Suppose a business buys conversion research, then a testing program, and runs paid media management alongside, all at our published starting rates.

• Conversion research, one engagement: $6,500.
• Testing program from month 2 through month 6, five months at $5,500 per month: 5 x $5,500 = $27,500.
• Paid media management for all six months at $2,400 per month: 6 x $2,400 = $14,400.
• Total at starting rates: $6,500 + $27,500 + $14,400 = $48,400.

That figure is a floor, not a forecast. It excludes media spend itself, and a business with more funnels, more markets, moderated user testing with recruitment and participant costs, or complex measurement would pay more. It also shows why the timeline matters: with research taking around 3–5 weeks and meaningful program results taking around 4–6 months, the first six months are an investment in learning as much as in results. When you compare proposals, lay each one out the same way so you are comparing like with like. Our guide to allocating marketing budget can help you decide how much of the total should go to fees versus spend.

Aftercare: what happens after launch, and after you part ways?

Aftercare questions cover the long middle of the relationship and its end. They are easy to forget during selection and expensive to discover later.

How do you keep improving once the early wins are in?

Why ask it: Many programs slow down after the first few months once the obvious issues are fixed. You need to know how the agency keeps learning.

A good answer sounds like: Research that continues alongside testing, regular reviews of what the program has learned, and a refreshed prioritized list at a set interval. The agency describes how it avoids testing the same ideas repeatedly and how it spots when a channel or test area has run out of headroom.

A bad answer sounds like: The same monthly activity forever, with no sign of new research, or a sudden push for more budget without new evidence.

How do you handle changes to our site, products or compliance rules?

Why ask it: Your business will change during the engagement. Site releases can break tracking and tests; new products, markets or rules can change what the agency is allowed to say.

A good answer sounds like: A process for checking tracking and running tests after each site release, a way for your team to flag changes early, and familiarity with review steps for regulated claims. The agency asks what compliance training your team has and how approvals work.

A bad answer sounds like: Finding out weeks later that a release broke conversion tracking, or ads going live with claims your legal team never approved.

What does a handover look like if we part ways?

Why ask it: Every relationship ends eventually. A clean handover protects the value you paid for.

A good answer sounds like: A written handover within the notice period: confirmed account access, a log of every test and its result, open work, documentation of tracking and conversion setup, creative files and a call with whoever takes over. The agency treats this as normal, not as a concession.

A bad answer sounds like: Reluctance to discuss it, or a handover limited to "we'll remove our access." Your own data governance policy should cover this too.

Scoring agencies side by side

Interviews blur together after the third one. A scorecard forces you to write down what you heard while it is fresh and to compare agencies on the same criteria. Score each criterion from 1 to 5 immediately after each interview, and have at least two people score independently before comparing.

Weight the criteria to suit your situation. A business with low traffic should weight testing discipline and research heavily, because the answer to the low-traffic question decides whether a testing program is worth buying at all. A business spending heavily on paid media should weight paid media method and measurement. Whatever the weighting, a score of 1 or 2 on ownership and exit should rule an agency out regardless of its other scores.

All the questions on one page

Print this list and take it into each interview. Tick each question off as it is answered, and note anything the agency deflected.

Preparing your side of the interview

An interview is only as good as the information you bring to it. Agencies can give specific answers only if you give them specifics. Before the first meeting, prepare a short brief covering your goals, your monthly traffic and conversions by funnel, the markets you serve, your current media spend, the tools in your stack, the engineering time you can offer and who will approve work on your side. Our guide to briefing an agency walks through what to include.

Share the same brief with every agency so their answers are comparable. Offer read-only access to analytics under a confidentiality agreement if you want the agency to give an informed view before signing. Some agencies will ask for a paid audit or discovery phase before they commit to a full proposal; that is often a sign of seriousness rather than a sales tactic, provided the scope and fee are clear.

Finally, agree internally what a good outcome looks like before you meet anyone. If your leadership expects results within a month and the credible agencies all say meaningful program results take around 4–6 months, that is a conversation to have inside your business, not a reason to pick the agency with the most optimistic answer. You can browse more of our Digital Marketing & CRO articles for background on each area, or see how our own Digital Marketing & CRO work is structured if you want to compare it against this list.

How to make the final call

Once the scorecards are in, the decision is usually clearer than it felt during the interviews. Discard any agency that will not give you ownership of your accounts and data. Among the rest, prefer the one that asked the best questions about your business, gave the most specific answers about method, and was most honest about what it could not promise. Check references by asking former clients how the agency handled a test that failed or a month that went badly, not just whether they were happy.

Price matters, but compare proposals on the same scope and the same assumptions. A lower monthly fee with no research, no development capacity and platform-only reporting is not cheaper if it produces nothing you can use.