How to Spot Fake Agency Case Studies

An agency sends you three case studies. A clinic that “grew organic traffic by 312 per cent”. An e-commerce brand that “hit page one in six weeks”. A B2B firm that “generated RM1.2 million in pipeline”. Every number is probably accurate. None of them is checkable.

That gap is the actual problem with fake agency case studies, and it explains why the usual advice does not help much. You are told to watch for exaggerated claims and stock photos. Meanwhile the case study that costs you money is the one with a modest, entirely true number attached to a client you cannot name, over a period nobody wrote down.

The useful question is therefore not “is this true?” but “who could prove it is false?” If the answer is nobody, the document is marketing, whatever the figures say.

Malaysian buyers face this more often than they used to, because more of the market is now online to be marketed to. The Department of Statistics Malaysia reports that 74.4 per cent of establishments had a web presence in 2024, up from 72.7 per cent a year earlier. Every one of those owners is a prospect for somebody’s case study deck.

So this piece does the thing the decks avoid. It sorts the evidence into three piles: what you can check yourself, what an honest agency can hand over at no cost, and what nobody can settle either way. That is the sorting we run at IZI Digital Marketing before recommending anyone. First, why so few case studies can be checked at all.

PART 1 · DIAGNOSE

Fake, or Simply Impossible to Check?

IN BRIEFOutright invention is rarer than buyers assume. The common failure is a true number stripped of the details that would let anyone test it. Treat missing detail as the finding, not as an oversight, and apply the same rule you would when shortlisting a digital marketing agency.

Agencies remove detail for ordinary reasons. Clients ask not to be named. Nobody kept the baseline screenshot. The junior who ran the account has left. None of that is dishonest, and all of it produces the same result for you: a claim with no handle on it.

Three removals do nearly all the damage:

• The client disappears. “A leading clinic in Klang Valley” cannot be contacted, so the strongest possible check — asking the client — is closed off before you start.
• The dates disappear. Without a start and end date, you cannot tell whether the gain took four months or four years, or whether it sat on top of a seasonal peak.
• The baseline disappears. A 312 per cent rise from 40 visits a month is 125 extra visits. The percentage is true and the business impact is nearly nil.

Once you see the pattern, the reading gets faster. You are not judging plausibility any more. You are counting how many of the three are missing.

BENCHMARK BRIEFING 1 OF 4

What Must a Case Study Contain to Be Checkable?

IN BRIEFSeven elements carry the weight, and they sit at four different levels of proof. Two you can settle alone, three the agency can settle in minutes, one needs the client, and one cannot be settled by anyone. Sorting them this way turns vetting into a measurement problem rather than a character judgment.

Read the table below before you read the deck. Anything in tier one that is missing is a question you can answer without asking permission.

PART 2 · DIAGNOSE

What Can You Check Yourself in Twenty Minutes?

IN BRIEFMore than most buyers expect. The company register, the client’s live site and an archived copy of it settle tier one without anyone’s cooperation. Run these before the call, the same way you would normalise numbers when comparing SEO proposals side by side.

How to verify an agency case study before the first call

Five checks, in order. Stop early if the first one fails, because everything after it rests on the client being real.

1. Confirm the client exists. Search the named business on SSM e-Info, the Companies Commission of Malaysia’s official corporate information portal. A trading name with no registered entity behind it is worth asking about.
2. Open the client’s live website. Look for the pages, content or booking flow the case study says were built. Work that shipped leaves visible traces.
3. Pull an archived copy. Use the Wayback Machine to see the site as it stood at the claimed start date. Note its own limits: it does not archive pages behind logins or forms, so absence of a capture proves nothing on its own.
4. Reconcile the numbers against the site’s size. A twenty-page brochure site that supposedly generates six thousand organic sessions a month deserves a specific explanation.
5. Write down three questions. One about the baseline, one about the period, one about what else was running at the time. Send them before the meeting, in writing.

None of this is adversarial. An agency that has done the work will answer all three questions in a paragraph.

BENCHMARK BRIEFING 2 OF 4

Which Kinds of Proof Actually Hold Up?

IN BRIEFProof is strong when a third party controls it. Google controls a shared Search Console view; the client controls a reference call; the agency controls a screenshot. Rank the proof you are offered by who holds the pen, and the ordering stops being a matter of taste.

The model below scores common proof types on how hard each would be to fabricate or reframe. Higher means somebody other than the agency stands behind it.

PART 3 · DESIGN

Screenshot, Shared Report or Reference Call?

IN BRIEFAsk for one thing, not three. The shared report is the cheapest strong proof and the fastest to produce, so it is the right opening ask for most shortlists — including the ones drawn from a review of KL agencies.

Sequencing matters as much as the ask itself. Requesting reference calls from every bidder is how you get three polite delays and no evidence before the decision deadline.

BENCHMARK BRIEFING 3 OF 4

Where Does Framing Inflate a True Number?

IN BRIEFFive framing moves account for most of the distance between an impressive headline and a modest reality. Each is technically accurate, so calling them lies gets you nowhere. One precise question deflates each of them.

Keep this table open during the pitch and ask the question in the last column the moment the matching claim appears.

PART 4 · DEPLOY

Ask for the Share Link, Not the Screenshot

IN BRIEFGoogle already ships the mechanism. A shared Performance report view exposes one report and nothing else, so client confidentiality survives the request. Treat a refusal the way you would treat vague pricing in a Google Ads fee discussion — as information.

The usual objection to evidence requests is confidentiality. It is largely solved, and has been for years. Search Console’s share function grants access only to the current view of the report, not to the property, the account or any other page — Google says so in its own Performance report documentation.

Three requests cover almost every claim you will meet:

• A shared Performance report view covering the exact months in the case study, filtered to organic search.
• A dated export of the same period. Search Console exports every tab and the filters applied, which is why an export is harder to reshape than an image.
• The AI visibility report where the case study claims AI-era wins. Google introduced generative AI performance reporting in Search Console in June 2026, so those claims are now checkable against a real report rather than a narrative.

BENCHMARK BRIEFING 4 OF 4

How Long Does a Verification Run Take?

IN BRIEFTwenty minutes for the checks you control, then a few days of waiting for the ones you do not. The waiting is unavoidable, which is the argument for starting before the shortlist meeting rather than after it.

The model below assumes one case study and one person doing the checking.

PART 5 · DRIVE

What If Nothing Can Be Verified?

IN BRIEFDo not walk away automatically. Newer agencies genuinely have thin evidence, and some good ones sit under strict client agreements. Replace past proof with future proof: a short paid pilot with measurement you own from day one.

Unverifiable history is a risk to price, not always a reason to refuse. What you must not do is accept the risk silently and hope month six settles it. Three moves convert the gap into something manageable:

• Buy a small, separable first phase. An audit or a single-lane sprint with a defined deliverable tests the agency on your business rather than on someone else’s numbers.
• Own the accounts from day one. Search Console, analytics and ad accounts registered to you, with the agency added as a user. Then their next case study is one you can verify yourself.
• Write the baseline down before work starts. Export the current figures and file them. Most disputes at renewal are really disagreements about where things started.

Google’s own guidance on hiring an SEO makes a related point worth holding on to: nobody can guarantee a top ranking, and any pitch claiming a special relationship with Google is misrepresenting how search works. Realistic estimates beat impressive certainties.

THE VERDICT

Rank the Evidence, Not the Story

Fake agency case studies are less common than the internet suggests. Uncheckable ones are everywhere, and they cost you the same money, because a claim nobody can test does no work in your decision even when it happens to be true.

So change what you are looking for. Sort each claim by who could contradict it: you, the agency, the client, or nobody. Spend twenty minutes settling the first group, ask for a shared report to settle the second, and save the reference call for the finalist. Then price whatever is left as risk instead of arguing about it.

The habit outlasts the decision, too. Once you own the accounts and record the baseline before anyone starts, you stop needing other people’s case studies. You have your own, and yours is the only one you can actually check.

FAQ

Frequently Asked Questions

1. How do you know if an agency case study is real?

Check whether anyone outside the agency could contradict it. It depends on three details: a named client, a start and end date, and a starting figure. With all three you can verify most of the claim in twenty minutes; with none of them, treat the case study as a story rather than as evidence.

2. What should a real case study include?

A named client, a dated period, the baseline it started from, and a business outcome rather than only traffic. It depends on the engagement, since a paid campaign should also disclose the spend behind the result. Anything you cannot trace back to Google’s own data belongs in the weakest evidence tier.

3. Can agencies fake dashboard screenshots?

Yes, and that is why screenshots sit near the bottom of the proof ladder. It depends less on dishonesty than on control: an image is produced entirely by the agency, while a shared Search Console view is produced by Google. Ask for the shared view and the question mostly disappears.

4. Should I ask to speak to an agency’s clients?

Yes, but only for the one or two agencies you are seriously considering. It depends on how far along you are, because reference calls spend a client’s goodwill and take a week or two to arrange. Use cheaper checks to narrow the field first, then use the call to confirm the business outcome.

5. What if the agency says client work is confidential?

Ask which part is confidential, because the answer is rarely all of it. It depends on the agreement, though a shared Search Console report view exposes only one report and no account access. An agency unwilling to seek that permission is telling you something about the relationship, not about the contract.