Operations
Part of Digital marketing agencies: choosing and governing
How to choose digital marketing agencies without guessing on fit
How to choose digital marketing agencies in 2027 by scoring records, not pitches, with weighted criteria and a decision table for cross-channel work.
What to take away
- Weight criteria before final proposals arrive, then score against those weights. Scoring after you read the pitch is how fit gets decided by presentation.
- Ask for records, not claimsbaseline, period, metric definition, and who did the work. A number without those four is a story.
- Normalize cost across fees, media, production, technology, data, and expenses before comparing totals.
- Keep a written record of scores, dissent, and approval. A defensible choice is one you can explain a year later.
- Pay for a bounded diagnostic when uncertainty is material. It is cheaper than a wrong retainer.
Choosing among digital marketing agencies is a procurement decision, not a taste decision. The work below borrows its structure from public-sector source selection, where the rule is simple: decide what matters, tell every candidate, and judge only on that.
Weight the criteria before you read anything
Write your categories and their weights first. A workable set for cross-channel work:
Weight the criteria
- problem understanding
- channel specialty
- assigned team
- evidence quality
- measurement approach
- security and data handling
- commercial value
- collaboration
- exit readiness
Assign each a weight that sums to 100. Then define, per category, what a strong record looks like. "Evidence quality" might mean a baseline measured before the work started, a stated period, and a named metric definition.
A sample weighting:
FAR 15.303 on source-selection responsibilities governs federal acquisition, not a commercial agency hire. Its logic reduces to one question: did you disclose the evaluation factors before proposals arrived?
The decision table
| Stage | Required record | Decision test |
|---|---|---|
| Weight | Criteria, weights, and scoring definitions | Could two reviewers score alike? |
| Brief | Same facts, questions, and factors to all | Did every candidate see them? |
| Verify | People, work, references, and access | Is evidence current and comparable? |
| Compare | Total cost, controls, and exceptions | Are proposals normalized? |
| Decide | Scores, narrative, dissent, and approval | Can the choice be defended? |
Worked example. Agency A scores 71 against the weights and leads on evidence quality. Agency B scores 76 and leads on security and team continuity. Both price the same scope. The panel picks B and records a dissent note on A's stronger evidence. That file is what you defend a year later.
Check the evidence behind each claimed result
Treat provider case studies as self-reported until checked. For each claimed result, ask for the baseline and the period. Ask for the metric definition and what the client supplied. Also ask who performed the work.
No public directory ranks agencies on your weights. The record does the ranking.
Then contact permitted references about problems and corrective action, not just satisfaction. Repeated patterns across reviews matter more than any single review, positive or negative. A weak result explained clearly is better evidence than a strong result explained vaguely.
Ask which tools the team will use for reporting, such as Google Analytics 4, Looker Studio, or BigQuery. The brand matters less than who owns the account and maintains the data.
Compare total cost on the same scope
Fees are the visible part. Totals diverge in several areas. Here is where:
Compare total cost
- Media
- Production
- Technology
- Data
- Expenses
- Taxes
- Markups Ask each candidate to price the same scope so the comparison holds.
A hypothetical $100,000 first year shows the spread. The figures are an example only:
Compare total cost
| Line | Hypothetical amount | What moves it |
|---|---|---|
| Media | 50,000 | Paid mix, reach, and frequency targets |
| Agency fees | 30,000 | Retainer hours, seniority, and fee model |
| Production | 10,000 | Asset volume and revision rounds |
| Technology and data | 6,000 | Licenses, tracking, and warehouse work |
| Expenses and taxes | 4,000 | Tools, travel, and jurisdiction |
Media is the largest swing. A low-media performance retainer inverts the split. Ask for the same line items in the same units.
CISA's SMB vendor risk template resource gives a worksheet for supplier review. Use it to answer one question: what access will this agency hold? Read-only dashboard access and server-side tagging access need different controls.
Run a controlled handoff
The GOV.UK technology selection guidance answers one question: who controls the data and the accounts? Its ownership-cost tests map onto digital marketing agencies pricing, where scope and fee model decide the real total.
The CISA software acquisition fact sheet covers development practice, supply-chain exposure, deployment, and vulnerability management. One question for the agency: what will it build, host, or maintain for you? Ask it without treating the fact sheet as approval.
Common questions
How many agencies should we shortlist?
Use the smallest number that preserves real choice and permits careful scoring. Complexity, risk, market depth, and your own procurement capacity matter more than a fixed count.
Should every agency get the same brief?
Yes. Identical facts, questions, and evaluation factors reduce presentation bias. They make assumptions, exclusions, teams, pricing, and risk comparable.
Is an unpaid pitch a fair test?
A short clarification call is reasonable. Substantial speculative work should be paid, bounded, and designed to test how the team behaves without demanding free strategy.
