You can check a lead generation agency's results in three steps before you sign anything. First, read each case study for three things: a hard number, the time it took, and the business outcome it led to. If one of them is missing, you are reading a testimonial.
Second, find out which number the headline is: calls booked, calls that actually took place, or contracts signed. Third, get someone outside the agency to confirm it: the client on a short phone call, a verified review on a third-party platform, a recommendation on LinkedIn from a person you can look up.
An agency with real results gets through all three easily, because the proof already exists. One that stalls on any step has answered your question for you.
Smirnov Consulting Group is a Prague-based B2B outbound lead generation agency that runs cold email and LinkedIn campaigns for founder-led B2B companies and books qualified sales calls. It is my company, so this is a test I expect to pass myself. Below is how to run it properly, including six questions for the reference call that an agency cannot rehearse for.
What a real case study has in it
When I wrote down the questions I would ask before hiring an agency, the case-study red flag took one line: "No case study with a real number and a timeframe in it. Not a testimonial. A number, a period of time, and a named outcome."
Each of the three parts catches a different trick.
- The number catches adjectives. "Significant pipeline growth" cannot be checked by anyone. 215 can.
- The timeframe catches slow wins dressed up as fast ones. 24 contracts in 3 months and 24 contracts in 3 years describe two very different agencies.
- The named outcome catches vanity metrics. The outcome should be something your business can bank: calls that took place, signed contracts, revenue. Leads, opens and accepted connection requests are activity.
Here is the test applied to three results from our own campaigns, each next to a vague write-up of the same result, the kind that fills plenty of agency sites.
| Vague version | The version that passes the test |
|---|---|
| "Helped a Swiss SEO agency scale its pipeline with LinkedIn" | 215 sales calls and 24 contracts in 3 months, LinkedIn outbound from 2 accounts |
| "Transformed lead flow for a financial services firm" | First 7 days: 26 booked calls, 4 contracts. By day 30: 132 booked calls. By day 60: 344 booked calls |
| "A great partner who filled our calendar" | 30 days on LinkedIn and email: 2,187 people contacted, 14 calls booked, 11 held, 2 contracts signed |
The left column sounds pleasant and proves nothing. The right column could be wrong, which is exactly why it is worth something: somebody was willing to be held to it.
Two of those rows already have a post of their own. I took the financial services campaign apart step by step in a teardown of the 344-call campaign.
The 30-day row comes from my post on the audit buyers run before they reply, where the same cold email got 2% replies in one business and 7.5% in the next. That post also describes a first call with a UK design agency whose own case studies ran two sentences each with no results in them. From a buyer's seat, nothing set it apart from anyone else.
Ask what the typical client got
Published case studies are the best work an agency has done. That is fine. Nobody puts their average month on the home page.
But a serious agency can tell you, without squirming, what a typical engagement looked like and what a weak one looked like. If every answer is another highlight, you are looking at a shop window and still know nothing about the shop.
Booked, held or signed: which number is the headline?
The most useful question in this whole process is also the shortest. When an agency quotes a number of calls or meetings, ask: booked, held, or signed?
- Booked means a calendar invite exists.
- Held means the call took place with the person who was invited.
- Signed means a contract came out of it.
Each number is smaller than the one before it, and the gaps are normal. The third row of the table above shows all three at once.
As I wrote in my list of hiring questions, a gap between "the booked figure and the showed-up figure" is "a coaching conversation about reminders and timing, not something to hide in a monthly summary."
Part of that gap usually sits with the client. A call nobody can take for weeks tends to slip until it quietly dies, which is why calendar room for those calls belongs on the list of things to settle before you hire.
What is not normal is an agency that can only give you one of the three.
The Swiss SEO result above shows why the last number matters most. 215 calls and 24 contracts means roughly one call in nine turned into a signed client. That ratio says far more about the quality of the calls than 215 does on its own.
Use this as a quick filter for any figure you see in a case study:
| Figure in the case study | What it proves | What it leaves open |
|---|---|---|
| Emails sent, connection requests, opens | The agency sent things | Whether anyone wanted them |
| Replies, "leads", interested prospects | The message reached someone who answered | Whether they were the right people |
| Calls booked | A prospect agreed to talk | Whether the call happened |
| Calls held | Real conversations took place | Whether your offer converts |
| Contracts and revenue | The whole chain worked for that client | Whether it repeats for you |
How to confirm a number from outside the agency
Everything so far you can do by reading. This part means contacting people, which is why buyers skip it, and it is the step that is hardest to fake.
Ask for the client behind the case study
Ask for a reference from the case study closest to your own business. Then ask for a second one from a client who is no longer with them.
The first confirms the headline. The second shows you how the agency behaves when an engagement ends, which is the part of the relationship no deck will show you.
A reference call is only useful if you ask things the agency could not have rehearsed. These are the six questions I would use:
- "What number did they report to you at the time?" Then compare it with the case study on the website. If the number has grown since, ask the agency why.
- "Was that booked, held or signed?" Ask the client, not the agency. The client knows which calls actually happened.
- "What was already working before they started?" Referrals, a known brand, a founder with a big network, an existing list. A strong result on a strong base still counts, but you need to know how much of it the agency built.
- "What did you have to provide?" Case studies, a rebuilt LinkedIn profile, salespeople free to take calls within days. If the reference had a sales team and you do not, the outcome will not transfer as it stands.
- "How long until the first call you would call a good one?" This gives you the real ramp, separate from the headline period.
- "Are you still working with them, and would you sign again on the same terms?" A pause before the answer is information. A clean yes from someone who left on good terms is worth more than a warm speech from a current client.
Fifteen minutes is enough. If the reference cannot answer the first two questions, treat the headline number as unconfirmed.
Read third-party reviews with two filters
Testimonials on the agency's own website are chosen by the agency. Review platforms such as Clutch, G2 or Google Business Profile are harder to curate, so they carry more weight. Read them with two filters.
- Verification status. Clutch says that when it cannot verify a reviewer's identity or the project details, it may still publish the review but labels it "Not Verified" (Clutch Help Center, checked September 2026). Weigh those less.
- Consistency across places and dates. A wall of glowing quotes on the agency's own site and a thin or silent profile everywhere else is a mismatch worth one direct question. So is a profile where most of the five-star reviews arrived within a few weeks of each other. Ask what happened that month.
For US companies there is also a legal floor under all of this. In August 2024 the Federal Trade Commission finalized a rule that bans reviews and testimonials claiming to come from people who do not exist, AI-generated ones included, or from people with no actual experience of the business (FTC, checked September 2026).
The rule is written around consumer reviews, so how it applies to a particular B2B case study is a question for a lawyer. The direction is clear enough. This is not legal advice.
Look at the people as well as the company
Open the founder's LinkedIn profile, and the profile of whoever would actually run your account. Recommendations are a small check that gets ignored. They are written by named people who put their own profile behind the words, and anyone can read them.
My rule on this is short: zero recommendations is a red flag, and recommendations are proof.
If a case study does name the client, check the easy things. The company exists. The person quoted works there, or did at the time. The size of the company fits the size of the claim.
One strong case study is not a track record
A single great result can come from a hot market, a founder with a large network, or an offer that was already selling itself. It proves the agency can do it once. What you want to see is the same shape repeating.
A practical way to weigh what you find:
- One case study shows capability. Get the reference and treat it as a single data point.
- Three or more, from different clients, start to show a method. Look for the same logic in each: who was targeted, which channels, how the numbers moved from contacts to calls to contracts.
- At least one close to you matters more than any number of distant ones. Same kind of buyer, similar deal size, similar sales cycle. A result selling small packages to five-person firms tells you little about selling six-figure contracts to enterprise IT.
- Recency. Ask when each campaign ran. Deliverability rules and buyer habits shift, so a best case from years ago says less about what the agency can do this quarter.
One consistency check on top. When we start with a client, case studies with real numbers are on the short list of things that must exist before the first message goes out. An agency that asks you for that kind of proof should have its own ready.
Red flags that end the check early
Some answers make the rest of the checking unnecessary. The first two are quoted from my own list of hiring red flags.
- Any guarantee on lead or meeting counts. "Whoever promises it does not control most of what decides it: whether you show up to the calls, whether your team closes, whether your market wants the thing."
- Paying per lead when nobody has defined a lead. "You will receive exactly what you paid for, which is rows in a spreadsheet."
- Case studies that are two sentences of praise. Kind words, no number, no period of time.
- Only one number, ever. The headline says "meetings", and nobody can tell you how many were held or signed.
- No reference, or only current clients who were clearly briefed.
- Testimonials that live only on their own website and nowhere you can verify.
Weeks of free work is a signal too
The usual advice is to push for a pilot before a long contract. A short, paid pilot with agreed definitions of what counts is sensible. Weeks of unpaid work is a different offer, and it deserves a question of its own.
I wrote about this from the seller's side in five days of free work, and the buyer's half is one paragraph: "If a vendor offers you weeks of unpaid work to win your business, do not congratulate yourself on the negotiation. Ask why their pipeline allows it. People with proof and demand do not audition, and the ones worth hiring will qualify you back."
An agency whose results are real has already built the proof once and shows it to everyone. As I put it in that post: "Case studies with a real number, a real timeframe and a real outcome. Two sentences of praise from a happy client proves nothing."
An agency with time for weeks of free work has either little else going on or a plan to make the hours back later. Either way, ask them which.
Questions buyers ask about agency case studies
Is it normal for case studies not to name the client? Yes. Plenty of B2B clients prefer that competitors do not know who fills their calendar. An anonymous case study is fine if it keeps the numbers, the period and the channel (ideally the industry too), and if the agency can put you on a call with the client once you are close to signing. None of the three examples above names the client.
Can I trust reviews on Clutch or G2? More than the agency's own testimonials, because the agency does not write them. Prefer reviews marked as verified, read the critical ones to see how the agency responded, and check that the reviewers look like the clients in the case studies.
What if the agency will not give me a reference? Ask why. Some clients do refuse. An agency with several happy clients can usually find one willing to spend fifteen minutes on the phone. If it cannot, weigh everything else it shows you lower.
If you only have ten minutes
Before your next call with an agency, open its case studies and pick the one closest to your business. Write three questions under it: booked, held or signed; what was already working before they started; and can I speak to this client.
Most agencies sort themselves out on those three before the call is over.