A paid-acquisition relationship usually goes wrong before the first campaign launches. The failure is rarely “the agency could not buy ads.” It is more often a mismatch that nobody forced into the open: the client thinks it is buying profitable growth, the vendor thinks it is being paid to hit a platform KPI, account access sits in the wrong hands, conversion data is weak, creative responsibility is vague, and the exit terms are discovered only after performance slips.

That risk matters in a mature market. IAB/PwC reported U.S. internet advertising revenue of $294.6 billion for 2025, up 13.9% year over year. In a market that large, competent media buying is table stakes. The more useful due-diligence question is whether a partner can operate inside your economics, measurement limits, data rules, and decision cadence.

Before comparing retainers or promised ROAS, force the vendor to answer the questions below in writing.

Start with ownership, not tactics

The first six questions determine whether you are buying a service or accidentally renting access to your own growth engine.

Question A useful answer should make clear Warning sign
1. Who owns each ad account? Your company owns the primary accounts; the vendor receives appropriate access. Vendor insists campaigns must live only in its account.
2. Who owns pixels, tags, audiences, feeds and historical data? Ownership and export rights are explicit. “We will sort that out later.”
3. What access level do you need? Least privilege that still permits the agreed work. Blanket admin access with no operational reason.
4. Who can add or remove users? Named client-side administrators remain in control. Only one vendor-side administrator exists.
5. What happens on termination? Access, exports, creative files and documentation have a dated handoff process. Handoff is undefined or billable at an unknown rate.
6. Which assets are portable? Campaign history, audiences where platform rules allow, naming maps, creative source files and reports are listed. “Our methodology is proprietary” is used to block ordinary account portability.

Google Ads lets account owners invite users, edit access levels and remove access later. Google also warns that relying on only one administrator can create continuity risk. That is a practical governance point, not a theoretical one: keep at least one qualified client-side admin and document who owns every critical account.

The rule is simple: a vendor may operate the system, but the advertiser should not lose control of the system merely because the contract ends.

Ask what the vendor is actually optimizing

“ROAS” is not an objective until both sides define the revenue, attribution window, margin treatment and customer type behind it.

Ask questions 7 through 10:

7. Which business metric is the primary optimization target?
Possible answers include first-order contribution, qualified lead value, new-customer revenue, profit-adjusted conversion value, booked revenue, or a proxy because better data is not yet available. The vendor should be able to explain why the chosen metric is close enough to the business outcome.

8. How will brand and non-brand demand be separated?
A campaign that captures people already searching for your brand can look excellent without creating much incremental demand. The partner should show how branded demand, retargeting, existing customers and prospecting are reported separately where the channel allows it.

9. How are returns, cancellations, offline sales and lead quality fed back?
If the platform sees only the initial order or form fill, it may optimize toward cheap but low-quality outcomes. Ask which downstream events can be imported and how often.

10. What measurement changes are expected in 2026?
This question tests whether the vendor maintains the plumbing rather than just the campaigns. Google Ads, for example, consolidated enhanced conversions into an account-level setting in April 2026 and accepts user-provided data through website tags, Data Manager and API connections. A partner does not need to use every feature, but it should know which implementation is in place, who maintains it, and how consent and privacy requirements are handled.

A good answer contains a map: business outcome → conversion event → data source → platform signal → reporting view → decision rule.

Do not let “measurement” become permission creep

A technically possible integration is not automatically an appropriate one. Ask the vendor to list every data source it wants, the fields it needs, the system that receives them, the retention or deletion path, and who can change that configuration. Customer email, phone, CRM status, order value and offline conversion data can improve measurement in some setups, but they also increase the consequences of weak access control. The vendor should be able to explain the minimum data necessary for the agreed objective and what still works if a source is withheld. “Give us admin access to everything” is not a measurement plan.

This is also where legal and consent requirements become market-specific. The correct setup can differ by jurisdiction, product category, customer relationship and platform. A media partner can help operate the tooling, but it should not pretend that a platform feature by itself answers the advertiser’s privacy or compliance obligations.

Force the fee model onto one page

The media fee is rarely the whole acquisition cost. Question 11 should be: What exactly will appear on an invoice, and what necessary work sits outside it?

Request a one-page cost map that separates:

  • media spend;
  • percentage-of-spend fees;
  • fixed retainers;
  • minimum monthly fees;
  • creative production;
  • creator or affiliate payments;
  • landing-page work;
  • feed management;
  • tracking or analytics implementation;
  • third-party software;
  • call tracking;
  • marketplace fees;
  • travel or production expenses;
  • rush work;
  • onboarding and offboarding charges.

Then ask 12. What happens to fees when spend falls, rises sharply, or pauses?

Percentage-of-spend pricing can create a structural conflict if the vendor earns more whenever the client spends more. A fixed retainer can create a different conflict if workload rises but compensation does not. Neither model is automatically bad. The question is whether incentives and scope are visible enough that both sides know when the commercial model stops fitting.

Do not compare proposals on the headline retainer alone. Compare fully loaded acquisition operating cost and the amount of internal labor still required.

Make creative responsibility painfully specific

For paid social and creator-led acquisition, creative operations can be the bottleneck. IAB’s January 2026 creator-economy measurement work describes fragmented metrics and proxy-based ROI as structural problems at scale. Similar ambiguity appears in everyday vendor relationships when nobody owns the production loop.

Questions 13 and 14:

13. Who is responsible for generating, editing, approving and refreshing creative?
Ask for expected monthly volume by format, not “ongoing creative support.”

14. How does creative learning get converted into the next brief?
A useful process connects ad comments, search queries, sales objections, landing-page behavior and performance by concept. If the vendor reports only “Ad 7 won,” the learning is too shallow to compound.

For products that require demonstrations, trust or explanation, this may matter more than small differences in bidding technique.

Ask how testing decisions are made before seeing the first test

Question 15: What makes you stop, scale or extend an experiment?

The partner should specify:

  • the decision metric;
  • minimum observation window or sample logic;
  • what counts as a material implementation change;
  • how seasonality or promotion periods are handled;
  • whether tests are channel-native, holdout-based, geo-based or simply directional;
  • which conclusions will not be drawn from weak data.

There is no universal sample size for every acquisition test. Traffic volume, conversion rate, variance, sales cycle and business risk all change the answer. A serious vendor should say “we cannot know yet” when that is the correct answer.

Question 16: How do you test incrementality rather than only attribution?

Do not demand perfect causal measurement from a small budget. Do demand intellectual honesty about the difference. Ask how the team distinguishes platform-reported conversions from conversions that would likely have happened anyway. The right method may be a formal experiment, a geo test, an audience holdout, a brand/non-brand split, a matched-period analysis, or simply a lower-confidence directional read when the data is thin.

Reporting should make a decision easier, not create a meeting

Question 17: What will we receive every week and every month?

A useful weekly view is short:

  • spend versus plan;
  • the business KPI and its trend;
  • new versus returning customer mix where relevant;
  • major creative/query/product concentration;
  • material tracking problems;
  • tests opened, closed or extended;
  • one sentence on what the team will do next.

A monthly review can go deeper into cohort quality, margin, attribution differences, channel saturation and budget reallocation.

If every report has 40 charts but no explicit decision, the reporting system is decorative.

The last question is the one most buyers forget

18. What would make you tell us to spend less?

This is a strong vendor filter because it tests independence. Useful answers might include:

  • marginal acquisition cost is rising faster than contribution;
  • payback breaches the agreed cash threshold;
  • creative volume cannot support additional spend;
  • inventory or fulfillment cannot absorb demand;
  • brand search is masking weak prospecting;
  • tracking quality falls below the threshold required for automated bidding;
  • the next dollar has a better expected return in conversion work, retention or another channel.

A partner who cannot imagine recommending lower spend may not be acting as a buyer’s agent.

One more practical check belongs in the selection meeting: ask who will attend the first 30 days of work. The people selling the relationship are not always the people operating it. Get the names or roles of the media buyer, analyst, creative lead and account owner, and ask which of them can make budget, tracking and escalation decisions without waiting for another team.

A quick scorecard for the selection meeting

Score each candidate 0, 1 or 2 on five dimensions:

  1. Control: account and data ownership remain with the advertiser.
  2. Measurement: the partner can connect platform events to business outcomes and describe uncertainty.
  3. Economics: all fees and internal workload are visible.
  4. Operating system: creative, testing, reporting and escalation have named owners and cadence.
  5. Exit quality: the relationship can end without destroying access, history or documentation.

A 10/10 score does not guarantee performance. It only means the operating conditions are unusually clear. A lower score may still be acceptable when the partner has a narrow specialty, but the trade-off should be deliberate.

What changes the answer

This checklist is stricter for a brand that spends across several channels, uses customer-level data, runs regulated offers or depends on complex offline revenue. A small local advertiser may need far less process. A marketplace seller may also accept less data ownership because the marketplace itself controls the customer environment.

Channel mix changes the questions too. Paid search needs query and brand/non-brand discipline. Paid social needs heavier creative operations. Retail media introduces retailer economics and data limits. Creator programs create additional contracting and measurement questions.

The mistake is not choosing a small vendor, a large agency, a freelancer or an in-house operator. The mistake is signing before you know who owns the system, what the system is optimizing, how much it truly costs, how learning compounds, and how you leave.

Sources

Related Reading