Concierge and wizard of oz MVP: learn before you automate
A concierge MVP and a wizard of oz MVP both test whether the outcome you promise is worth paying for before you automate it: you deliver the result manually, or behind a simple front that looks like software, while you watch what people actually do.
What is a concierge and wizard of oz MVP?
A concierge MVP means you do the work yourself (or with a small team) for a short list of customers: spreadsheets, email, exports, phone calls, whatever matches the job. The customer knows they are in a pilot or a service, and you learn whether the finished result matters enough to repeat.
A wizard of oz MVP uses the same manual engine but often hides it behind a product-shaped surface: a form, a chat box, a bare dashboard. The visitor believes they are using an early product; you fulfil the request by hand and study friction, language, and drop-off. The label comes from the story where the wizard is a person behind a curtain. In practice the two terms overlap; teams argue about naming more than they change the work.
This is for founders with an idea or pre-seed traction who do not yet have a technical co-founder and need evidence beyond friendly encouragement. For landing pages, interviews, and the full menu, see how to validate a startup idea. Here the focus stays on manual delivery, honest limits, and when to stop pretending the spreadsheet is the product.
How to run a concierge or wizard of oz MVP
Keep the cohort small, the promise narrow, and the log honest. You are buying learning, not revenue at scale.
Write the outcome in one sentence
Who receives what, by when, and what would count as failure. If you cannot describe the deliverable without naming your future tech stack, simplify.
Choose five to fifteen people, not the internet
Pick a segment you can reach without pretending you already have a platform. Mix strangers with warm intros only if you label which is which in your notes.
Decide how visible the manual work is
Concierge pilots often say “we do this for you this month”. Wizard-of-oz tests keep the curtain unless trust requires transparency. Deception without a follow-up plan burns the next launch.
Charge or commit something real
Free favours teach politeness. A modest fee, a signed pilot letter, or a clear time commitment from the buyer separates curiosity from need.
Log every step you repeat twice
Those repetitions become your automation backlog. Steps you hate are often the ones customers barely notice.
Set an end date and a stop rule
Example shape: stop manual delivery if fewer than three of ten pilots renew or expand, or if fulfilment takes more than twice the time you assumed.
What a manual MVP proves, and what it does not
- ProvesOutcome value
Someone received the result and came back, paid, or referred a peer under fair terms.
- ProvesWorkflow detail
You see exceptions, data formats, and approval steps surveys never surface.
- Does not proveScale
Your hands do not predict cloud cost, support load, or onboarding once many users arrive.
- Does not proveSelf-serve product
People who like a white-glove pilot may refuse a self-service app with the same logic.
A strong concierge run is weaker than repeat usage in deployed software and stronger than a fake door test that only records a click. It sits well after a smoke test landing page when strangers already accept the promise and your open question is whether the delivered outcome changes behaviour.
A made-up example: weekly supplier summary for a small manufacturer
A made-up example: a founder in Aargau targets operations leads in firms with 40–90 staff. The hypothesis is that buyers lose half a day each week reconciling supplier delivery notes with invoices. They offer a four-week pilot: every Monday morning the lead receives a one-page PDF listing mismatches from the previous week. There is no app. The founder collects exports by secure upload and builds the PDF in a spreadsheet template. Five firms agree; three pay CHF 200 for the month, two join free through a trade contact.
Week two shows that one buyer only cares about three suppliers, not the full list. Week three kills a fancy chart nobody opens. Week four, two paid pilots ask to continue; one free pilot drops out when the internal buyer changes. The founder does not claim software margins. They claim that a narrow reconciliation report is worth renewing for two of five, and that automation should start at ingestion rules, not at a mobile dashboard. The next step is either a thin wizard-of-oz upload page with the same human backend, or a scoped build conversation, not a pitch deck about AI.
Who should try manual delivery, and who should skip
Who this is for
- Founders whose hypothesis is about the value of the outcome, not only the headline on a page
- B2B ideas where trust and exceptions dominate, and a polite survey would lie
- Teams that finished a cheap demand test and need to see whether fulfilment breaks the promise
Who should skip or shorten it
- Ideas where the risk is “will anyone click?”; start with a measured page instead
- Products that only work at scale (marketplaces, real-time matching, network effects)
- Founders who cannot sustain manual work for the agreed window; burnout invalidates the data
Remember that ideas as such are not protectable in Switzerland; execution and brand are (IGE on ideas). Manual delivery does not hide your concept. It buys time to learn before you fund code.
Is wizard of oz the same as concierge?
Often in conversation, yes. Concierge usually stresses visible service; wizard of oz stresses a product-like front with manual fulfilment. Pick the label that keeps you honest with customers, then run the same disciplined pilot.
When do you stop manual delivery?
When you have a written rule hit (renewals, referrals, or time spent per customer), or when the same manual steps repeat cleanly enough that automation is the next experiment, not a default twelve-month roadmap.
When pilots show the outcome is real and the open question is behaviour inside deployed software with measurement, see the product validation package.
Written by
Aurum Avis Labs
Builds and ships at Aurum Avis Labs. Writes here about what we learn working with founders and SMEs in the DACH region.
Related Articles
You might also be interested in these articles
Fake door test: measure demand before you build
A fake door test shows whether visitors click as if they would buy or sign up, before the product exists. Steps, limits, and when it is enough.
Smoke test landing page: test demand before you build
A smoke test landing page checks whether strangers take one clear action on your offer. Steps, limits, and what email signups do not prove.