How to verify a PR agency through references

Use references to verify specific claims about how the proposed team performed under a comparable scope, especially during slow periods, changes, and disagreements. Combine agency-supplied references with independently sourced feedback when practical. Record dates, team members, scope, and evidence so a favorable relationship is not mistaken for proof that the agency fits this assignment.

Reference relevance matters more than brand recognition

Ask for a reference that matches the objective, company stage, market, service mix, operating model, budget shape, or risk level. No reference will match every factor, so name the two or three that matter most. Confirm that the people proposed for your account actually worked on the cited engagement and whether their roles have since changed.

Ask the agency to obtain permission before sharing contact details or nonpublic engagement information. Confirm the reference's identity, company role, dates, and relationship to the provider. Disclose and account for any investor, portfolio, personal, financial, or referral relationship that could affect the feedback.

An agency may be unable to name a client or share results because of confidentiality. Request substitute evidence:

an anonymized sample or report;
a comparable nonconfidential reference;
confirmation of the proposed team member's role;
public work with the provider's contribution explained;
an independently sourced reference with permission;
a small paid project with defined acceptance criteria.

An explained confidentiality limit and useful substitute evidence is different from refusing all verification. A newer agency or solo operator may reasonably have fewer references, which shifts more weight to work samples, relevant experience, operating controls, and a bounded initial scope.

Comparable questions reveal how the team operated

Use the same core questions across candidates:

Starting point. What objective, baseline, scope, timeline, and internal constraints shaped the work?
Delivery team. Who performed strategy and daily work? Did staffing, capacity, or senior oversight change?
Working rhythm. How did the team handle approvals, urgent requests, disagreements, and missed client dependencies?
Work quality. Were the audience rationale, claims, content, opportunities, and advice relevant and accurate?
Observed learning. What changed after an approach produced weak evidence or conditions shifted?
Reported results. Which activities, outputs, audience responses, outcomes, or business effects were observed? What else may have contributed?
Commercial control. Did scope, expenses, invoices, or out-of-scope work create surprises?
Difficult period. What happened during a quiet month, missed result, staff change, correction, or hard conversation?
Exit or renewal. Why did the relationship continue or end, and what accounts, files, rights, and status transferred?
Best fit. For which assignment would the reference hire this team again, and where would they choose another model?

Ask for concrete examples and dates. A general rating carries less evidence than an account of how the team corrected an inaccurate draft, staffed a launch delay, or reported weak results.

Differences should be reconciled before selection

A proposal, case study, and reference may describe different periods, teams, definitions, or scopes. Build a short claim log with the agency statement, reference evidence, unresolved difference, and owner. Return factual differences to the agency and record its explanation.

The startup's selection owner controls reference outreach, comparable questions, notes, and the final assessment. Legal, people, or security owners set any limits on contact data, sensitive allegations, access, and retention. Keep notes within the selection team and use them only for the stated purpose.

One negative reference may reflect poor fit, a difficult former relationship, or a material delivery problem. Several references can repeat the same unsupported story. Weigh specificity, recency, direct observation, incentives, and consistency rather than averaging sentiment.

Compare the final evidence with the case-study review, agency vetting scorecard, and proposed scope of work. Record which claims became contractual commitments and which remain selection assumptions.