Reference calls: what to ask an agency's past clients

Every agency hands you the same thing when you ask for references: two or three clients who love them. Of course they do. Nobody referees you their disasters. This doesn't make reference calls pointless — it makes question design everything. A scripted call with a friendly referee produces the warmth you already had. A good call produces information you didn't.

Here are the twelve questions I've watched squeeze real signal out of even the most curated reference list, plus what the evasions sound like. (Yes, we hand out our own reference list. Yes, the people on it genuinely like us. Make them work for it anyway.)

Before the call: two setup moves

**Ask for one reference who ended the engagement.** Not because something went wrong — engagements legitimately end — but because a former client has nothing to protect. "Finished," "moved in-house," "budget cut": all fine. You want someone who can speak freely because the agency can no longer help or hurt them. A confident agency supplies this without blinking. We do.

Send the questions in advance to the agency, not the referee. Counter-intuitive, deliberate. It tells the agency you're thorough, discourages referee-coaching gossip, and — most usefully — their reaction is data. An agency that bristles at "what did the invoice look like in month three?" has just answered a question you hadn't asked yet.

The twelve questions

The opening (gets them talking like a person, not a testimonial)

1. "What had you tried before them, and why did it stop working?" Establishes the referee's reference points — the comparison class they're judging against. An agency praised by someone who's only ever worked with agencies is a different data point from praise by someone who's been burned twice.

2. "What did you hire them to do, and what did they actually end up doing?" The gap between the two is the agency's real shape. "We hired them for a website and they ended up restructuring our onboarding" tells you this team follows the problem. It also surfaces scope discipline — did the expansion happen through clean change control or through drift?

Money and time (where the truth lives)

3. "Tell me about the first date they missed." Not whether — the first. Everyone misses a date eventually; pretending otherwise is a flag in itself. You're listening for three things: how early the agency said it, what they offered (replan, added capacity, scope trade), and whether the referee called it a surprise. An agency that surfaces slippage in the weekly demo rather than in a Friday-evening email is structurally honest.

4. "What did the invoice look like in month three versus month one?" This is the single highest-yield question on the list, because invoices can't be spun. You're probing for surprise fees, scope-creep billing, timesheet opacity. Good answers sound boring: "Same shape, plus one change order we'd agreed." Hedged answers contain the word "eventually" — as in "the invoicing was eventually explained."

5. "How did the final cost compare to the proposal?" Listen for the percentage and the narrative. ±10% with clear change orders is a healthy project. ±40% with "well, our requirements evolved" means the agency quotes low and bills the Delta. We've written about reading SOWs before you sign — this question is how you audit one retroactively.

The people (staffing honesty)

6. "Did the people who pitched stay on the work? Roughly until when?" The bait-and-switch is the industry's oldest trick, and references are the only place it's visible. Quantify it: "the creative director dropped off after week three" is information; "there was good continuity… mostly" is a hedge with a heartbeat.

7. "Who on their team would you hire?" My favourite disguised question. If the referee names the junior designer and goes quiet on the leads, the senior layer was decoration. If they light up about a specific engineer's unglamorous diligence — that's who did your project.

8. "How did they take feedback — especially the first time you hated something?" Every engagement has a moment where the work gets rejected in a room. The answer tells you whether you get defensiveness, compliance ("OK, whatever you say"), or the third thing you actually want: pushback with reasoning. An agency that never disagrees with you is billing you to abandon their expertise.

The aftermath (what it's like when it's real)

9. "What broke after launch, and how was that handled?" Post-launch behaviour is invisible in pitches and painfully visible in references. Was there hypercare? Did fixes get priced as ransom, or absorbed as their own bugs? Did anyone from the agency check in at week six, or did the relationship end exactly at the final invoice?

10. "Six months later — what's decayed, and what's still earning?" This separates durable craft from launch-day glamour. Sites decay through handover gaps; growth programmes decay through measurement nobody maintained. What survived tells you what the agency builds for.

**11. "Would you hire them again — and would you hire them for something bigger?"** "Yes, again" is cheap. "Yes, for something bigger" costs the referee something and therefore means something. The hesitation here is almost always honest: "Again for web, yes. For the data platform… I'd want to think about it." Believe the hesitation. It knows things.

12. "What should we watch out for?" Asked plainly, near the end, when the referee likes you by now. Close the notebook, lean back. The answer is usually the most candid sentence of the call: "They go quiet when estimates are hard — chase them," or "Brilliant, but they'll outpace your content team; hire the writer first." That's the agency, accurately described, from the only chair that sees it whole.

Reading the hedges

A short field guide to evasion, because referees rarely lie — they soften:

• "Overall it was a positive experience" with no specifics = it was fine, which for the price you're discussing is not enough.
• Frequent past tense about enthusiasm ("we were really excited") = the excitement has a date of death.
• Praise of process, silence on outcomes = well-run work that didn't move the numbers. For growth engagements especially, this is disqualifying.
• "I'd have to check" on anything invoice-related = you don't have to check. That was the answer.

And the strongest negative signal of all: the referee who asks whether you've signed yet before answering question twelve.

One bonus move

Ask the agency to arrange a reference who matches your shape — same stage, similar in-house capability, a comparable problem. A glowing call with an enterprise client is nearly useless if you're a 20-person company about to be an agency's smallest account. Fit mismatches show up in month two as exactly the neglect you're hiring to avoid — something we flag in choosing an agency: signals over showreels for a reason.

If the agency can't produce a shape-matched reference, you have your answer about their client mix — and if a pilot is on the table, that's where pilot design earns its keep.

Key takeaways

• Reference lists are curated; questions are how you un-curate them.
• Ask for one reference whose engagement has ended — former clients hold the rawest data.
• Send questions to the agency, not the referee. Their reaction is a free data point.
• Money and time questions beat vibe questions: invoices, date-misses, and final-versus-proposal costs can't be spun.
• "Would you hire them for something bigger?" is the question that costs the referee something to answer.
• Learn the evasion grammar: unspecified positivity, dead past-tense enthusiasm, process-praise with outcome-silence.

FAQ

How many reference calls should we do? Two, done well, beats four done as ceremony. One current client, one former. Past that you hit diminishing returns and start projecting indecision — which good agencies also notice.

Isn't it rude to ask a referee about invoices? No — it tells them you're serious, and serious clients are the kind good referees genuinely want to help. Frame it as mechanics, not suspicion: "how did billing compare to the proposal?" reads as diligence because it is.

What if the agency only offers written testimonials? Politely treat that as a no. Testimonials are marketing copy by structure; a reference is a conversation by definition. An agency with happy clients and no callable ones is telling you something about how its relationships end.

Should we do the calls before or after the proposal stage? After proposals, before contracting — and before any budget conversation gets too invested. References exist to test a specific agency, not to generate a longlist. That's what the brief was for.

As an agency, doesn't publishing this hurt you? The opposite. We want clients who test us properly, because properly tested trust holds in month seven. If you want to test us live: the brief form is here.