Before you prototype
Pretotyping: Test If It’s the Right Idea Before You Build It Right
A polished prototype answers “can we build it?” Pretotyping answers a ruder question first: would people actually want this, use it as you expect, and keep using it? Alberto Savoia’s method — named on pretotyping.org — is a set of cheap, fast tests so you spend money on The Right It, not on building It right too early.
Last updated: October 13, 2026
Direct answer
What is pretotyping?
Pretotyping is a set of cheap, fast techniques to validate whether people would want and use a product before you invest in building a real prototype. Prototyping answers “can we build it / will it work?” Pretotyping answers “would people be interested, use it as expected, and keep using it?” The goal, in Savoia’s phrasing on pretotyping.org, is to make sure you are building The Right It before you build It right.
Key takeaways
What to remember
- Write the key assumption first: the one claim that, if false, means this is definitely not the right it. Then pick a pretotype that isolates that claim — not a mini-version of the whole product.
- Write a market-engagement hypothesis as X% of Y will do Z before anyone sees the artifact. Opinions collected in a slide deck are not a test.
- Pretotype vs prototype is a different question, not a cheaper prototype. Jeff Hawkins’s wood-and-paper Palm Pilot was a usage test, not an engineering build.
- ILI (initial interest to try) is not OLI (continued use). A waitlist can pass; repeat use can still fail. Do not invent a numeric bar from someone else’s workshop slide.
- A passed pretotype is evidence on one assumption in one setting. It is not product-market fit, and a Yibud score is not a pretotype. Neither predicts success.
Two different questions
Pretotype vs prototype
Pretotyping.org’s methodology page is blunt: the main objective of prototyping is buildability — can we build it, will it work, how cheaply, how fast. The main objective of pretotyping is appeal and usage — would people be interested, will they use it as expected, will they continue to use it. Mixing the two is how you spend a quarter proving you can ship something nobody reaches for.
| Artifact | Question it can answer | Palm example (pretotyping.org) |
|---|---|---|
| Pretotype | Would people be interested? Will they use it as expected? Will they keep using it? | Jeff Hawkins of Palm Computing mocked up a Palm Pilot with wood and paper and carried it for weeks, simulating the core experience with the smallest investment of time and money — to see whether he would actually carry and use such a device. |
| Prototype | Can we build it? Will it work as expected? How cheaply / how fast can we make it? | A working engineering prototype of the same device. That answers buildability. It does not tell you whether the form factor survives a real week in a pocket. |
Hawkins did not use the word pretotyping — pretotyping.org notes the term did not exist yet. The site still treats the wood-and-paper Palm as the method: simulate core experience before the expensive, time-consuming working prototype.
Why bother
The Law of Market Failure, and The Right It
Pretotyping.org’s one-sentence law: most new ideas fail in the market even if they are competently executed. That is the Law of Market Failure on that site — a qualitative claim, not a percentage this page will invent. Their prescribed fight is to test the market objectively, rigorously, and quickly before you invest to develop the idea. Pretotyping, they write, was originally developed at Google in 2010. The book framing is Alberto Savoia’s The Right It. At Stanford’s Entrepreneurial Thought Leaders seminar, Savoia listed seven strategies for fighting market failure. This page names them as the site names them. It does not add an eighth.
- 1Obey the Law of Market Failure.
- 2Make sure you are building The Right It.
- 3Don’t get lost in Thoughtland.
- 4Trust only in Your Own DAta (YODA).
- 5Pretotype It.
- 6Say it with numbers.
- 7Think global, test local.
Thoughtland is the slide-deck country where everyone nods. YODA is data you collected on this idea, this week, from people who could have said no. “Say it with numbers” is the market-engagement hypothesis, not a dashboard of hits. “Think global, test local” is the Provincial instinct: one place, one time, before world-wide.
The method
Five steps from pretotyping.org
These five steps are the methodology page, not a Yibud invention. Do them in order. Skipping the written hypothesis is how a fake door becomes a marketing site.
- 1
Isolate the key assumption
What is the one assumption about your idea that, if false, means it is definitely not the right it? If you have twelve, you do not have a pretotype plan. You have a backlog. Name one.
- 2
Choose a type of pretotype that isolates that assumption
The artifact should make that one claim cheap to observe. A Fake Door isolates initial interest. A Pinocchio isolates whether you would actually carry the form factor. A Mechanical Turk isolates whether people would use the experience if a human fakes the backend. Do not build a little of everything.
- 3
Make a market-engagement hypothesis
How many, and what kind of, people will do what with your pretotype? Pretotyping.org’s template is as simple as: X% of Y will do Z. A solid hypothesis takes guesswork and opinion out of the sitting. Write X, Y, and Z before the first visitor.
- 4
Test in the real world — start small
Put the pretotype where people already are. One place, one time. A parking-lot tent beats a global launch. You are collecting YODA, not a press cycle.
- 5
Learn, refine, hypozoom
Evaluate against the hypothesis you wrote. Refine the pretotype with the new data. If the hypothesis held, decide which other situations you should test to get a complete picture — pretotyping.org calls that hypozooming. One passing weekend in one Slack is not a market.
Technique map
A short map of pretotype types — not a new taxonomy
The names below are from Savoia’s published pretotyping quick-reference materials. This page does not invent types. Where Yibud already has a how-to, use that page. This table is only “which assumption does this isolate.”
| Type | What it isolates | Existing Yibud how-to |
|---|---|---|
| Fake Door | ILI — initial interest via an artifact that suggests the product already exists and is available. | Fake-door test guide (full procedure). |
| Pinocchio | A non-operational form factor. Hawkins’s wood Palm: would I actually carry this, and for what? | No dedicated Yibud how-to — use the method on this page. |
| Mechanical Turk | Humans simulate a complex backend. Savoia’s classic: IBM speech-to-text with a hidden typist, so you learn usage before the machinery. | Wizard of Oz MVP (curtain up; human in the back). |
| Provincial | A tiny geography or sample before a world-wide launch. Think global, test local. | No dedicated Yibud how-to — one place, one time is step 4. |
| One-Night Stand | A short-run live service before any long-term commitment — days, not a lease. | Overlaps concierge MVP when delivery is openly manual. |
| Impostor / Impersonator | Modify an existing close product so it can impersonate the new one, then collect behavior — not compliments. | No dedicated Yibud how-to — still write X% of Y will do Z. |
| MVP (as used in pretotyping literature) | A stripped first iteration: the minimum core that would be useful to early adopters, used here as a pretotype, not as “ship the company.” | Cheapest-validation-test compares MVP cost to other tests. |
A smoke-test landing page is often a Fake Door pretotype. A week of manual delivery is often a One-Night Stand or Mechanical Turk. The method is the hypothesis and the key assumption — not the nickname of the artifact.
Say it with numbers
X% of Y will do Z, ILI, and OLI
Pretotyping.org requires a market-engagement hypothesis. Savoia’s workshop materials name two interest metrics: ILI and OLI. This page uses those names as definitions, not as industry conversion rates. There is no published “good” ILI percent here. You write X before you look.
Market-engagement hypothesis (copy this)
X% of Y will do Z. Y = [qualified people you can actually reach]. Z = [observable action: join waitlist / carry the mock / complete one job / come back once]. Window = [one place, one time]. If we miss X, we [kill / rewrite the assumption / change Y].
X is your bar, written first. It is not a benchmark copied from a PDF. Y must be people who could have said no. Z must be something you can count without a compliment.
ILI — Initial Level of Interest
Interest to try: will they reach for an artifact that suggests the product exists? A Fake Door is built to read ILI. A click, an order attempt, a waitlist email. ILI is not continued use.
OLI — Ongoing Level of Interest
Continued use after the first try. Savoia’s materials treat OLI as a different measurement from ILI — people can want it at first and then stop. A pretotype that only captures ILI has not answered “will they keep using it?”
Celebrating ILI while you have no OLI plan is how a waitlist becomes a vanity number. Neither ILI nor OLI is a Yibud score. Neither predicts that the company will succeed.
Not the same page
Pretotyping vs fake door, Wizard of Oz, concierge, cheap tests, kill lines, and smoke tests
Several Yibud pages sit next to this one. Mixing them up turns one tactic into “we pretotyped” — or turns a method page into a duplicate how-to.
Fake door — one technique, not the method
That page is the operational playbook for one pretotype type: a CTA that suggests the product exists, so you can read initial interest. This page is the umbrella: key assumption, X% of Y will do Z, hypozoom, ILI versus OLI. Run their steps when Fake Door is the right type. Do not skip the hypothesis because the page already exists.
Fake-door test guide →Wizard of Oz / concierge — simulation and manual delivery
Wizard of Oz is the curtain: the customer thinks it is automated; a human does the work. Concierge is openly manual. Both can be pretotypes (Mechanical Turk / One-Night Stand). Those pages teach how to run the week. This page is when to pretotype at all, and how to write the number that week will honor.
Cheapest validation test — a cost ladder
That page ranks tests by cost and signal so you pick a first instrument. This page is Savoia’s pretotyping vocabulary and Right It framing. A cheap test with no key assumption and no X% of Y will do Z is still Thoughtland with a Stripe link.
Cheapest validation test →Critical assumption / kill criteria — what to kill, then how to test
Those pages pick the claim that would kill the idea if false, and write the Stop contract. Pretotyping is the cheap desirability/usage test once that claim is named. A kill line without a pretotype is a sentence. A pretotype without a killable assumption is theater.
Smoke test / landing page — demand experiments that often are pretotypes
A landing page with one CTA is frequently a Fake Door pretotype. Those pages teach the surface: copy, traffic, what a click means. This page names the method around them: isolate the assumption, write X% of Y will do Z, start small, hypozoom. A conversion rate with no hypothesis is just a percent.
Prototyping — buildability, not a Yibud route
There is no /prototyping page on purpose. Prototyping asks whether you can build it and whether it will work. If your expensive unknown is still “would anyone use this?”, a working prototype is the wrong instrument. Buildability tests come after The Right It is no longer the open question — or they run in parallel on a different assumption, with a different hypothesis.
Worked example
One fictional fake-door pretotype you can copy (illustrative)
The sitting below is labeled fiction. Names, product, and the 15% bar are invented so you can see the setup. They are not a study, not an industry waitlist rate, and not a Yibud report. Copy the method, not the numbers.
The key assumption, written before the page exists
Illustrative names: founder Priya Nair at Noonboard, a lunch-planning tool for office workers who skip lunch three or more days a week. The tempting next move is a polished calendar-sync prototype. The key assumption — the one that, if false, means this is not the right it — is desirability: qualified skippers will reach for a waitlist that promises a Monday plan from their calendar, before any software exists. That is an ILI question. It is not a buildability question.
The pretotype type, picked to isolate that assumption
Fake Door: one landing page, one promise, one “join the waitlist” CTA. No calendar sync. No app. The page says this is early access, not a shipping product. Traffic: one named Slack community of operations managers, one week, no ads blended in. Provincial on purpose — think global, test local.
The market-engagement hypothesis, written Monday
“15% of Y will do Z.” Y = operations managers in that Slack who skipped lunch at least three workdays last week (self-reported in a one-line qualifier). Z = leave a work email on the waitlist during the same week. Window = that Slack, Monday–Friday. Action if miss: kill the calendar-sync prototype; rewrite who skips lunch, or stop. 15% is Priya’s bar, written first. It is not a benchmark from pretotyping.org or from this site.
What a result can and cannot say
If the week misses 15%, the key assumption took a hit. Do not “need more traffic” from a second community without restarting the count. If it clears 15%, you have ILI in one Slack — not OLI, not product-market fit, not permission to build the sync. Hypozoom: a second place, or a One-Night Stand that actually delivers five manual Monday plans and asks whether they come back. Interview the people who signed. Compliments from friends do not count as Y.
The method is the dated hypothesis plus one pretotype that can kill the assumption. A pretty prototype that answers “can we sync a calendar?” still leaves the Right It untested.
Where Yibud fits
Name the weak assumption first — then pretotype it
Pretotyping needs a killable key assumption before you pick an artifact. Yibud is a free, no-signup startup idea validator. Scores come from a deterministic rule engine on a one-line idea plus five questions — not from a language model guessing success. Optional AI, when it is used, only polishes prose. It does not invent the number. A Startup MRI report can surface a weak dimension (demand, distribution, monetization) so the next pretotype has something specific to isolate. The score is a flashlight. It is not a pretotype, not ILI, not OLI, and not a forecast that you will succeed. You can run the five steps on this page without ever opening the analyzer.
Analyze my idea →Common mistakes
What usually wastes a pretotyping week
Building a polished prototype to answer a desirability question
If the open question is “would they want or use this?”, a working build is the expensive way to learn they will not. Pretotype appeal and usage first. Prototype when buildability is the assumption you can actually kill that week.
Collecting opinions in Thoughtland instead of YODA
Surveys that ask “would you use this?”, advisor nods, and Slack polls from friends are Thoughtland. YODA is behavior you watched: they clicked, they carried the mock, they came back. Trust that — not the room that wants you to feel encouraged.
Skipping the written hypothesis
Without X% of Y will do Z, every result is a story. Write X, Y, Z, the window, and the miss action before the first visitor. A kinder bar on Friday is a recap.
Celebrating ILI while ignoring OLI
A waitlist is interest to try. Continued use is a different measurement. If your key assumption is that they would keep using it, a Fake Door that only captures emails has not tested that assumption. Plan the OLI pretotype — or admit you only tested ILI.
Treating a passed pretotype as finished product-market fit
One passing hypothesis in one place is a hypozoom candidate, not PMF. It is not a Yibud GO. It does not mean the startup will succeed. Run the next cheapest test on the next assumption, or write a kill line that can read this number.
Sources
Where these ideas come from
- Pretotyping.org — home — Used for the definition (tools, techniques, and tactics to validate a new product idea quickly, objectively, accurately); the goal of building The Right It before you build It right; origin at Google in 2010; the Law of Market Failure (most new ideas fail in the market even if competently executed); the fight (test the market before investing to develop); The Right It as the book framing; and the seven strategies from Savoia’s Stanford ETL talk as listed on the site. This page does not invent a failure-rate percentage.
- Pretotyping.org — methodology — Used for the Palm Pilot wood-and-paper pretotype by Jeff Hawkins; pretotyping vs prototyping objectives (appeal/usage vs buildability); the five steps, including the market-engagement hypothesis “X% of Y will do Z” and hypozooming. Hawkins did not use the word pretotyping; the site still treats the sitting as the method.
- Alberto Savoia, summary of pretotyping techniques (PDF) — Used for the named types on this page (Fake Door, Pinocchio, Mechanical Turk, Provincial, One-Night Stand, Impostor, MVP as used in pretotyping literature) and for ILI / OLI as names for initial interest to try vs continued use. This page does not copy workshop examples as industry conversion rates and does not invent numeric ILI/OLI benchmarks.
- Alberto Savoia, pretotyping quick reference for Stanford MS&E 277 (PDF) — Companion quick-reference for the same technique names and metric vocabulary. Prefer pretotyping.org pages as the primary citations; these PDFs are listed because the technique names and ILI/OLI labels are not fully spelled out on the public methodology HTML.
In one paragraph
Summary you can quote
Pretotyping is a set of cheap, fast techniques to validate whether people would want and use a product before you invest in building a real prototype. Prototyping answers buildability; pretotyping answers interest, usage, and continued use. Pretotyping.org: originally developed at Google in 2010; most new ideas fail in the market even if competently executed (Law of Market Failure); fight that by testing the market before you invest to develop; goal = The Right It before you build It right. Methodology: isolate the key assumption; choose a pretotype that isolates it; write X% of Y will do Z; test in the real world, one place, one time; learn, refine, hypozoom. Jeff Hawkins’s wood-and-paper Palm Pilot is the usage-test example, not an engineering prototype. ILI is interest to try; OLI is continued use — named in Savoia’s materials, without numeric benchmarks on this page. A Yibud score can surface a weak assumption to pretotype next. It does not predict success.
FAQ
Questions founders actually ask
What is pretotyping?
Cheap, fast techniques to test whether people would want and use a product before you build a real prototype. Pretotyping.org’s goal line: make sure you are building The Right It before you build It right. It is not a smaller prototype. It is a different question.
How is pretotyping different from prototyping?
Prototyping asks whether you can build it and whether it will work. Pretotyping asks whether people would be interested, would use it as expected, and would keep using it. The Palm wood mock is pretotyping; a working device that proves engineering is prototyping. Same product idea, different unknown.
What is a market-engagement hypothesis?
Pretotyping.org’s template: X% of Y will do Z. Y is a kind of person you can actually reach. Z is an observable action. Write it before the test. Without it you are collecting opinions in Thoughtland. X is your bar, not a published industry rate.
What are ILI and OLI?
Named in Savoia’s pretotyping materials. ILI (Initial Level of Interest) is interest to try — reaching for an artifact that suggests the product exists. OLI (Ongoing Level of Interest) is continued use after the first try. This page does not publish a “good” percent for either. A waitlist can be ILI and still miss OLI.
Does Yibud run a pretotype? Does a score predict success?
No and no. Yibud scores come from a deterministic rule engine on your idea inputs. They can point at the weakest assumption to test next. They are not a pretotype, not ILI, not OLI, and not a forecast. You write X% of Y will do Z and run the sitting yourself. You do not need an account to use this page.
Isolate the assumption, then pretotype it
Yibud scores weak dimensions with a deterministic engine. You still write X% of Y will do Z and put a cheap artifact in the real world. The report does not replace the sitting.
Generate a Yibud report →