An agency case study can show useful work and still fail as selection evidence. It may omit the starting point, blend several channels, report a result without a denominator, or use a client logo without permission to share the underlying proof. Diagnose the claim before treating the story as a reason to shortlist a provider.
Define the selection decision
Write what the case study is supposed to prove: capability, process, industry familiarity, measurement discipline, speed, creative quality, or commercial outcome. One story rarely proves all of these. Record the role of the evidence in the shortlist and the alternative proof you would accept.
Build the Agency Case Study Verification Map
| Claim layer | Verification question | Safe conclusion | | — | — | — | | client and scope | who was served, when, and for what work? | identity and scope may be confirmed | | starting point | what was the baseline and comparison period? | magnitude needs context | | method | what changed, by whom, and under what constraint? | process may be inspectable | | measurement | which source, denominator, filter, and lag? | result may be bounded | | contribution | what else changed during the period? | causality may be uncertain | | permission | can the evidence be shared and checked? | public reuse has a rights boundary | | transfer | what conditions match the buyer’s situation? | relevance is not a guarantee |
Mark each field verified, supported, asserted, missing, or not transferable. Do not let an attractive result compensate for an absent baseline.
Test identity, scope, and ownership
Ask whether the agency names the client, period, service, team, and role. A logo can establish recognition but not the agency’s contribution. If several suppliers worked together, record which work the case attributes to the agency and which remains unknown.
Check whether the named client has approved the story, metrics, screenshots, and quotation. The FTC advertising FAQ is a useful general boundary for substantiation and responsibility for advertising claims. It is not legal advice; escalate material claims and permissions appropriately.
Test the baseline and metric chain
For every result, ask: compared with what, over which period, using which source, for which segment, and with what denominator? “More visibility” is not the same as qualified demand. “Revenue influenced” is not necessarily incremental revenue.
Separate impressions, clicks, sessions, enquiries, accepted leads, opportunities, payments, and retention. The Salesforce lead implementation guide illustrates why capture and qualification stages need explicit definitions. Use the client’s own stages when evaluating relevance.
Test method and alternative explanations
Ask what changed, when, and what else changed. A case may coincide with seasonality, pricing, product availability, a platform update, a sales-team change, or a broader market shift. Do not demand a perfect experiment for every engagement, but require the agency to state what it knows and what it cannot attribute.
A credible case can be valuable even when it does not prove causality. It may show disciplined research, a clear handoff, a useful test design, or an honest limit. Grade the evidence for the capability you are actually buying.
Test evidence custody and repeatability
Ask where the underlying export, dashboard view, brief, or approval record lives; who can provide it; and whether the evidence is preserved in a form that another reviewer can inspect. A screenshot may show a number but not its filter, date, property, or denominator. A testimonial may show satisfaction but not a measurable outcome.
Check whether the method could be repeated in a smaller buyer-controlled pilot. The agency should be able to state the first diagnostic step, required access, expected lag, and stop condition. If verification depends entirely on a provider’s private dashboard or a customer who cannot be contacted, lower the evidence status and request another proof route.
Look for consistency across the case-study page, proposal, sales call, and contract. Different scopes, dates, metrics, or team descriptions are not automatically deception, but they are a reason to ask for the versioned evidence chain. Keep the question and answer with the shortlist record.
Ask for the smallest artefact that would let the buyer test the method: a redacted brief, a measurement map, a sample report, a change log, or a pilot acceptance record. The purpose is not to obtain confidential client data. It is to see whether the agency can explain inputs, decisions, evidence, and limits without relying on a polished narrative.
Test transfer without copying the outcome
Compare audience, offer, market, sales cycle, channel access, budget, team, measurement, and timing. A case that matches the buyer’s mechanism may inform a pilot; a case that only shares an industry label may not. Google’s people-first content guidance is a useful quality boundary for content examples: reader purpose, usefulness, originality, and evidence still matter. It does not guarantee ranking or commercial transfer.
Decide how the case can be used
| Evidence state | Meaning | Shortlist action | | — | — | — | | verified and relevant | identity, method, evidence, and conditions are inspectable | use as one input | | useful but bounded | process is clear, outcome or attribution is limited | request a capability sample | | asserted | story is plausible but material proof is missing | do not use as performance proof | | not transferable | conditions materially differ | ask for a closer example or pilot | | unsafe | permissions or claims cannot be established | exclude until resolved |
The diagnosis is complete when the buyer can state exactly what the case proves, what it does not prove, and which next check would reduce uncertainty. A shortlist should reward verifiable capability and fit, not the confidence of the storytelling.
How did this article land?
Choose one reaction. You can change it anytime.
