Reference Customer Discovery
Build with a fixed number of real customers until they'll stake their reputation on your product
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 6
- Confidence
- 95%
Instead of running fake-door tests, opportunity solution trees or research interviews, you find people who actually have the problem, embed with them, and refuse to leave until you have solved it for them. The success criterion is not a survey score but a reference: a customer willing to put their reputation on the line and tell other people about your product. Christian Idiodi sets hard target counts — 6-8 references in B2B, 15-25 in B2C — and treats hitting that number as the definition of product-market fit.
Origin
Christian Idiodi's own practice, refined across a claimed ~205 products and taught through Silicon Valley Product Group (the firm Marty Cagan founded). The adoption-threshold logic behind the reference counts draws on Geoffrey Moore's technology adoption curve, which Idiodi names explicitly, and on early enterprise-buying research he attributes to IBM's supercomputer sales.
Core principles
- 01Discover WHO has the problem and discover the SOLUTION at the same time — the two are one activity, not sequential phases.
- 02A reference customer is the only honest signal of love: they risk their own reputation to recommend you.
- 03What people say is different from what they do; asking for a public five-star review surfaces the truth an interview hides.
- 04You cannot do this in a silo — you must be immersed in the environment of the problem.
- 05The recruiting step is itself the test: if you cannot find enough people with the problem, it isn't a problem worth solving.
- 06Build only what ALL the reference customers want — that is how you find the true minimum.
How to run it
- 1
Define the problem and who has it
State the problem crisply and name the single target market that has it. Do not generalise; you are looking for one type of person, not everyone.
Pro tip Idiodi's opening qualifier with a prospective customer is a money question — asking Starbucks whether they'd pay a million dollars to fix their hiring problem instantly sized the pain.
- 2
Validate the problem exists beyond one caller
Get out of the building and go talk to other people who might have the same problem, in person, before designing anything. Idiodi literally drove around with his designer and engineer talking to a McDonald's construction site and a Macy's manager.
Pro tip Take the designer and the engineer with you. Because they were present when the problem was defined, they invent the solution without any requirements document ever being written.
Watch out If you cannot find enough people with the problem, stop. That is your natural pivot-out, not a reason to widen the target.
- 3
Recruit the target number of candidate references
Recruit early adopters — technologists who love new things, and 'evangelists' who believe they'll look good using it. Aim for 6-8 in B2B, 15-25 in B2C. Recruit more than you need (30-50) because some will drop out.
Pro tip Ask up front: 'If I build the thing you absolutely love, are you willing to tell people about it — video testimonial, written review?' The willingness to commit is a qualifier.
Watch out One or two customers is not enough — they may be the only two people on earth with that weird problem.
- 4
Deliver and iterate with them, in their environment
Work alongside the customers, shipping and tweaking until the solution genuinely works for them. Manual, unscalable delivery is fine — Idiodi's team ran the whole first version on spreadsheets, phones and email with no software at all.
Pro tip Watch the failures closely: the McDonald's interview no-show rate (half don't show, one in five gets hired) became the core insight that sized the entire product.
- 5
Build only what everyone in the cohort wants
If one customer asks for a feature and the others don't, do not build it. Unanimity across the cohort defines the minimum product; anything less is customisation for one account.
Pro tip The unanimity rule doubles as a sales defence later: when a prospect asks for a missing feature, you can point to N happy references who don't need it.
Watch out Customising for a single loud customer is how you lose the leverage of the reference set.
- 6
Cash the references — and only then call it product-market fit
Go back and actually ask each person for the review, testimonial or reference call. When you have the full number (6-8 B2B / 15-25 B2C) genuinely delivered, you have product-market fit and a launch-day proof set.
Pro tip Idiodi never launches an app without 25 five-star reviews live on day one — the references ARE the launch.
Watch out Hesitation at this step is the most valuable feedback you will get. People say yes to be polite; only when you ask them to actually publish it does the real objection appear.
In the wild
The head of global staffing at Starbucks called Idiodi with a crisis: an acquired Bay Area bakery had ~800 employees who might be undocumented and would need replacing fast. Instead of writing requirements, Idiodi grabbed a designer and an engineer, drove around validating the problem with a McDonald's site manager (120 hires needed on opening day) and a Macy's manager (20,000 seasonal hires), then ran a live manual pilot for McDonald's — sending 40 candidates, of whom fewer than 20 showed up and only four or five were hired. That failure exposed the real mechanic: hourly-job candidates don't show up, so you must send ~10x. They doubled down on the channels that worked, sent 120-130 the next week, and the manager hired 45-50. Idiodi then told Starbucks he'd need to send 3,000 people for their 800 roles, staffed it manually with hotels and spreadsheets, and won the contract. Los Angeles International Airport was deliberately fired as a customer (demographic-matching rules, three months to staff) because it wasn't the target market.
→ After ~9 months of building software around the validated manual process, the product booked $32 million in sales in its first 90 days; the Starbucks contact copied then-CEO Howard Schultz on an email saying the team had saved them. The product went on to serve McDonald's new-store openings and high-volume event hiring.
Idiodi uses a steakhouse to explain the reference-count logic: one online review and you assume nothing; two and you assume it's the chef and their spouse. He asks how many positive reviews from people who look like you it takes before you'd try it. His answer for consumer products is 25 — so he works with 25 steak lovers, iterating the menu with them until each is willing to write the review unprompted.
→ The reference set both proves the product and pre-loads the word-of-mouth flywheel, making the eventual sales job trivial: 'let me show you 20 people that look like you that recommend this steak.'
Common mistakes
Treating a usability test as proof of value
Teams run a test with 300 users, get a 90% satisfaction score, and declare the idea validated. Being able to use something is not the same as buying it, choosing it, or actually using it. Value risk is the hardest and most overlooked of the four risks, and usability scores say nothing about it.
Accepting a polite yes
People will say almost anything to avoid hurting your feelings or to get you out of their face. A verbal 'sure, I'd give you five stars' means nothing. Only the moment you ask them to actually publish the review do you discover the hesitation — and that hesitation is where the real discovery happens.
Building for one loud customer
If a single member of the cohort demands a feature the others don't need, building it inflates scope and destroys the unanimity that makes the reference set persuasive.
Chasing a customer outside the target market
The LAX airport contract looked bigger than McDonald's but carried demographic-matching and security constraints that made it a different problem. Idiodi took three months to staff it and concluded 'that is not our customer, never again.' A big logo that doesn't share the problem corrupts the discovery.
Is it for you?
Best for
Founders and product managers building a genuinely new product or entering a new market, who can get physically close to a small number of customers and are willing to do unscalable manual delivery first.
Not ideal for
Optimising an existing product at scale, low-ACV self-serve products where you cannot get face time, or teams whose roadmap is already dictated to them (in which case there's no value question left to answer).
From the transcript
“the Holy Grail of product work is really a reference customer this is somebody that has used your solution or your product loves it enough…”
“if you really want to solve a problem you need to get out of your building go spend time with someone that has the problem…”
“for B2B I want six to eight references for b2c I want you know maybe 15 to 25 references as an indication that we've achieved…”
“all 25 have to want the same thing”
“because people will say or do anything just to avoid hting our feelings”
“if I can't find 25 people that love St why in the world am I building a Sak house”
“I was discovering who had the problem and developing the customers that have the problem at the same time I was discovering and delivering a…”
From the episode
The essence of product management
Christian Idiodi (SVPG)