Prototyping Tactics: Fidelity & Pretotyping
Innovation & New Business Proposal - TUHH Institute of Entrepreneurship & Institute of Innovation Marketing, Hamburg · part of my Technology Management MBA · study notes for revision.
The strategy chapter dealt with direction and attitude: why an uncertain project has to be run in a way that allows failures to happen, and how those failures are then interpreted. This chapter drops one level down. Where the prototyping strategy provides the overall direction for the learning process, the prototyping tactics organise the actual activities - the concrete testing work that fills the weeks between one review and the next.
The course notes describe a tactic as a series of choices about different prototyping approaches. That is worth taking literally, because it means there is no single correct prototype. There is a set of decisions: which tool or process to use, which level of fidelity is appropriate, and how the results will be captured and interpreted. Get those choices right and a week of cheap work removes a real uncertainty. Get them wrong and a month of beautiful building removes nothing at all.
Underneath everything sits the reason prototyping exists in the first place. Entrepreneurial ventures and corporate innovation projects run at a high level of uncertainty, fed from outside by technology shifts, market dynamics and trade, regulation and the occasional black swan event, and from inside by uncertainty about the team’s own capabilities, its finances and its organisation. Reducing those uncertainties in a targeted way - learning what was previously unknown - is the heart of prototyping. A prototype that does not reduce an uncertainty is decoration.
1 · Strategy sets direction, tactics organise the work
Section titled “1 · Strategy sets direction, tactics organise the work”The two levels are easy to blur, so it helps to hold them apart deliberately.
- Sets the overall direction of the learning process for the venture or the innovation project
- Accepts that learning under high uncertainty always produces non-successful outcomes - ideas, projects and whole ventures fail
- To maximise learning, it must deliberately allow for the possibility of failure
- Deals with the emotional, cognitive and conceptual side of failure
- Organises what is actually built, shown and measured this week
- A series of choices: which tools and processes, with which pros and cons
- Which fidelity level is appropriate for the question at hand
- How the results will be captured and interpreted
2 · The double movement: problem space and solution space
Section titled “2 · The double movement: problem space and solution space”The framework the session builds everything on is a single timeline with two spaces on it. In design, ideation and prototyping form a constant cycle between divergent and convergent activities, and the cycle happens twice, once for the problem and once for the solution.
The second thing the diagram insists on, marked at every single stage, is that getting feedback should occur along the entire process. Feedback is not a gate at the end of development. It is attached to problem articulation, to problem selection, to concept generation, to concept selection and to the run-up to market introduction alike.
3 · Why teams rush into the solution space, and why the problem space pays
Section titled “3 · Why teams rush into the solution space, and why the problem space pays”Building feels like progress. Questioning does not. A team with a working material sample in its hands wants to make something out of it today, and the solution space is where making things happens - which is exactly why teams arrive there far too early, carrying a problem they never really pinned down.
The session teaches this the hard way, by opening the discussion of a design video with a question about the problem, not the product: how would you describe the problem that the team is trying to solve? And the case used in the same session is a story of a team that only escaped a weak problem because one member went and lived inside it for a week. Time in the problem space is not lost time. It is the cheapest place in the whole process to change your mind.
Three practical consequences I take from this:
- A wide problem set comes first. If only one problem statement was ever on the table, no selection actually happened.
- Test the problem, not only the product. Early prototypes can be aimed at whether the pain exists at all, long before any of them are aimed at whether your device removes it.
- The cost of being wrong rises with every step to the right. Being wrong during problem articulation costs a conversation. Being wrong after commercialisation costs the venture.
4 · The four choices that make up a prototyping tactic
Section titled “4 · The four choices that make up a prototyping tactic”Every prototype worth building is the answer to four questions asked in this order. The order matters, because teams that start from the fourth question end up measuring whatever their artefact happens to produce.
- Articulate the single biggest key uncertainty on the concept, in one sentence
- One prototype, one question - a prototype asked to settle four things settles none
- Is the question about desirability, about feasibility, or about both?
- Real users in the real situation, not classmates and not friends
- Both sides of a service count - the customer experience and the employee experience
- Enough of them that a result is a pattern rather than an anecdote
- Which of the techniques fits, and at which fidelity level
- Build the cheapest and fastest thing that can still answer the question
- Products, interactions, processes, environments, services and businesses each have their own natural forms
- Decide in advance what would count as evidence, and set the target
- Record the result - notes, data, photos, sketches, video
- Then interpret it as a team and decide what happens next
The course states the same thing as a loop it calls the general feedback gathering process: select the most interesting concept, articulate the biggest key uncertainty on it, design an experiment that tests the aspect relevant to that uncertainty, build the cheapest and fastest pretotype that can test it, execute the experiment and record the results - then repeat.
5 · Fidelity, the central dial
Section titled “5 · Fidelity, the central dial”Fidelity is how finished the artefact is. It is the one setting that most changes both the cost of a test and the honesty of the answer, and the guiding rule from the session is blunt: model to the lowest appropriate resolution.
- Fast, cheap and genuinely disposable - you can throw away five in a day
- Right for early, broad questions: is this the right problem, does the direction appeal at all
- Because it looks unfinished, people feel free to criticise it and to redesign it with you
- Nobody on the team is emotionally invested in a piece of cardboard
- Slower and more expensive, and every iteration costs real time
- Necessary for questions about detail, performance, cost, durability and real use over time
- Danger one: a polished artefact collects feedback on the polish - colours, wording, finish - instead of on the idea
- Danger two: the team becomes attached to what it has built and starts defending it instead of testing it
So the rule is: fidelity should match the question and the stage, never the team’s pride. If the question is whether a portable desk should be this big or that big, cardboard answers it perfectly, and a machined aluminium version answers it no better for a hundred times the price. The deeper framing from the session is that early on a prototype is a tool for learning, not always a tool for validation - and that only works if it is cheap enough to be wrong.
The three principles that go with the dial:
Right - focused on the issueRapid - explore quickly and simply, then iterateRough - lowest appropriate resolution, cheap
6 · The seven early prototyping techniques
Section titled “6 · The seven early prototyping techniques”| # | Technique | What it is | When to use it |
|---|---|---|---|
| 1 | Looks-like (the Pinocchio) | A non-functional version that mimics some or all elements of the concept - the classic being an early non-working block standing in for a handheld organiser, or a cardboard mock-up asking how big a portable desk should be | When you want reactions to how the product would look and work if the design were more complete; good for direction and desirability |
| 2 | Works-like (functional) | A model that lets you test one or more functions, usually with very little visual development - for example a motorised clothes steamer rig built only to test the guide assembly | When you want reactions to a function you are considering, or need to resolve feasibility questions; covers desirability and feasibility together |
| 3 | Mechanical Turk | Any prototype that replaces machinery with human beings to simulate a function, named after the eighteenth-century chess-playing cabinet with a person hidden inside; the class example is a ramen vending machine worked by a human behind the panel | When you want customer reactions to a function that would be expensive to produce mechanically. This is the same idea as a Wizard-of-Oz or concierge test: the service is real, the automation is not |
| 4 | Scale model | A build at a smaller scale, usually looks-like but sometimes functional - a scale vending machine, or a facility layout assembled in LEGO | When a full-size prototype is too expensive or too cumbersome at this stage, or when a small model is simply easier to carry to the customer |
| 5 | Role-play (behavioural or service prototype) | A prototype of an experience, in which people act out roles with each other - a mock-up operating room used to rehearse new surgical processes, or students acting out a new cultural enrichment programme for children | When you need to understand how people respond to new roles and behaviours, which is most of service design. Bodystorming is the same idea taken further: brainstorming services and processes with your bodies, as with rough seating layouts for transport and aircraft interiors |
| 6 | Storyboarding (2D prototyping) | A set of drawings that lay out a process flow step by step | Best for services and for concepts that are hard to describe in words; it lets a user react to individual steps and to the assumptions hidden inside them |
| 7 | Fake door | A looks-like artefact that appears more finished than it is, used to measure the response before anything exists behind it | Mainly for web and app products, to gauge real interest. Also called a smoke screen or a landing page test |
Digital wireframes are storyboarding. The session treats simple wireframe tools as a form of 2D prototyping and names the usual suspects for building low-fidelity wireframes, alongside the generative AI tools now used for the same job and for visualising ideas quickly. The example given is deliberately unglamorous: a wireframe for a group dog-running app, drawn at exactly enough fidelity to collect useful feedback and no more.
The fake door in full. Two real illustrations are given: a car-buying site whose founder tested demand through a fake entry before the business existed, and a student team that handed cards to ride-hailing drivers pointing them to a website purely to gauge interest. The session then works through what an online grocery venture could have done instead of building warehouses first - build a high-quality website with a polished front end and no back end, advertise locally in one major city and one suburb, and when orders actually arrive, buy the food at existing stores, rent delivery trucks and hire temporary personnel to deliver it, running the whole experiment for four weeks. Almost the entire cost of the real business is avoided while the only question that matters still gets answered.
Matching the technique to what you are prototyping, as the session tabulates it: products call for form models, breadboards and mock-ups; interactions for role-playing, scenarios and computer simulations; processes for bodystorming, storyboards and scale models; environments for bodystorming and scale models; services for storyboards and role-playing; and whole businesses for business-level walk-throughs and role-playing.
My own shorthand for the first two techniques: a looks-like prototype is a horizontal slice - it covers the breadth of the concept with no depth behind any of it, which is why it suits questions about desirability and direction. A works-like prototype is a vertical slice - one function taken all the way down to something that really runs, with nothing around it, which is why it suits questions about feasibility. Filming a storyboard as a short demo film is a natural extension of technique 6 for the times when the audience cannot be in the room.
7 · Pretotyping: the right it before it right
Section titled “7 · Pretotyping: the right it before it right”Pretotyping is not a cheaper prototype. It is a different question.
- Can we build it at all, and will it work?
- Will it work as intended?
- How small or how big can we make it?
- What would it cost to produce, and how long would the batteries last?
- How will people use it, and what will they use it for?
- Is this the right thing to build in the first place?
- Everything in the left-hand column is wasted effort if the answer here is no
The formal definition used in the session: testing the initial appeal and the actual usage of a potential new product by simulating its core experience with the smallest possible investment of time and money. In other words, testing an idea quickly and inexpensively by creating an extremely simplified, mocked or virtual version of it, in order to validate the premise that if we build it, they will use it. The informal version is easier to remember: pretend and test it before you make it, and make sure, as quickly and as cheaply as you can, that you are building the right it before you build it right.
Behaviour beats opinion. The experimentation framework the session cites forces an explicit choice between a say test and a do test. A say test collects statements: would you use this, would you pay for this. A do test collects actions, and actions are exactly what a fake door, a pre-order button or a sign-up form are built to capture - clicks, sign-ups, pre-orders, payments, repeat use. Enthusiasm is free; a click costs a second and a payment costs money, and that price is what makes the signal worth something.
8 · Define the evidence before you run the test
Section titled “8 · Define the evidence before you run the test”The steps of effective experimentation, as the session sets them out, come in five groups. Notice that two whole groups happen before anything is built.
The assignment version of this is a test card with four fields, and it is the cleanest checklist I have for any experiment:
- State the hypothesis qualitatively - what you believe, in a sentence.
- Describe the prototyping or pretotyping technique you will use.
- Describe the quantified measure you plan to take.
- Define the threshold that would give you the confidence to confirm the hypothesis.
Field 4 is the one people skip, and it is the one that makes the test real.
9 · Capturing, interpreting and acting on the result
Section titled “9 · Capturing, interpreting and acting on the result”The team process the session prescribes for watching a prototype being tested is three short instructions, and they are harder than they look:
The first one is the real discipline: the moment worth catching is usually the one nobody planned to measure - the user who holds the object the wrong way round, or who solves the problem with something else entirely. The second exists because memory rewrites results in your favour within a day, so documentation is expected in whatever form fits: notes, data, photos, sketches, video. The third is a team conversation rather than a private conclusion, and it deliberately separates what we saw from what we think it means. The session even asks the class to critique three things separately after watching a design team at work: the prototyping process, the prototypes themselves, and the data collection process - a reminder that a badly run test produces confident nonsense.
Handling a negative signal honestly. The threshold was set in advance precisely so that a miss counts. If the number comes in under the line, the options are to kill the concept, to change one variable and re-run, or to go back into the problem space - but not to quietly relabel the threshold. The strategy layer already made room for this: learning under high uncertainty always includes non-successful outcomes, and a prototype used as a tool for learning has done its job when it tells you no cheaply. The expensive failure is the one discovered after commercialisation. Then the loop closes with the shortest instruction in the whole deck: repeat.
Case corner
Section titled “Case corner”The situation. It is technically spring in Boston, but the temperature is below freezing and the wind makes it feel worse. Chloris Yang hurries across a car park to meet her team, wishing she had a hot coffee in her hand. She is a student on a master’s programme in entrepreneurial leadership, she grew up around her family’s food production business, she has already run and lost one start-up in China, and through a family contact she obtained samples of phase-change material - a substance that absorbs or releases a large amount of energy as it changes state, so it can pull a hot liquid down to a pre-specified temperature quickly and then hold it there for a while.
Her team of four, drawn from India, Russia and China, first aimed the material at baby milk formula: in all of their home cultures formula is mixed with hot water for safety, and parents then cool it down through an anxious manual cycle of feeling and tasting. They did the work properly - interviews with new mothers, nurses and nutrition specialists, plus a study of the competitive landscape of baby food containers - and then Chloris spent a week of spring break living with a family with a newborn. Observing the actual context and timing of feeds, she concluded that the problem mattered less than the team had assumed. They dropped it. The new target is a travel mug, aimed at a country of long car, train and bus commutes and heavy coffee and tea drinking. Many questions need answering. Which ones first, and how?
How I would prototype it early on.
- Which question first. Not can we build it - they already hold a material that does the physics. The open question is the one that killed the previous idea: does the temperature of a commuter’s coffee actually bother them enough to change what they buy? That question lives in the problem space, and the baby-milk detour is the team’s own evidence that skipping it is expensive.
- The riskiest assumption, stated so that it can fail: commuters experience the wait for coffee to cool, and its going cold later, as a repeated and real annoyance - real enough that they would pay a clear premium for a mug that removes it.
- The cheapest artefact that could test it. Two cheap things rather than one expensive one. First, a Mechanical Turk test with no product at all: hand commuters a cup that a person has already brought to the ideal drinking temperature, and observe whether anyone notices or values it. Second, a fake door: a polished single page for a named travel mug at a stated price with a pre-order or notify-me button, promoted to commuters near stations and car parks, with a looks-like Pinocchio mug - a shaped, non-working body - carried along to the in-person conversations so that people react to a real object rather than a description.
- Who to test it with. The people in the described situation: car, train and bus commuters who buy hot drinks on the way to work, plus the cafes that hand the cups over. Explicitly not classmates, who will be polite about it.
- What would kill or confirm it, agreed before running. Confirm if a clear majority of the intercepted commuters describe the temperature as a genuine repeated irritation and the fake door converts at the pre-agreed rate at full price, say two percent of visitors clicking through to a pre-order. Kill it if people shrug at the perfectly tempered cup and the page converts near zero - that is the baby-milk finding all over again, and finding it in a fortnight for the cost of a landing page is the best possible outcome.
- The honest exit. Any pre-order taken is a waiting list, described as such, and refunded immediately if the venture does not proceed.
Worked example
Section titled “Worked example”A small prototyping plan for a plausible venture: a subscription that collects broken small electronics from office buildings, repairs them and returns them within 48 hours. Three assumptions, ranked by risk.
| Rank | Assumption | Test | Fidelity chosen, and why | Signal measured | Threshold set in advance |
|---|---|---|---|---|---|
| 1 · highest risk | Office workers have enough broken devices lying unused that they would hand them to a stranger’s service | Fake door: a one-page site plus a poster in three office lobbies, offering a named price and a booking button | Low - the whole question is demand, and no repair capability is needed to answer it. A polished operations back end would answer nothing extra and would cost weeks | Bookings started per 100 lobby visitors over two weeks | At least 8 bookings started per 100 visitors, and at least 15 carried through to address entry |
| 2 | The 48-hour promise is what makes it worth paying for, rather than the price | Mechanical Turk: run the first 20 jobs entirely by hand with a courier and a bench technician, offering two advertised turnarounds in two buildings | Medium - the service has to be genuinely real to the customer, but nothing behind it needs to be systematised or automated yet | Split of take-up between the 48-hour and the 7-day option at the same price | The 48-hour option must take at least 60 percent of jobs, otherwise speed is not the differentiator |
| 3 | The hand-over and return steps feel safe enough that people part with a device holding their data | Role-play prototype of the full hand-over, plus a storyboard of the six steps shown to ten employees and two facilities managers | Low - drawings and acting cost an afternoon and expose the anxious moments better than a working app would | Number of participants who stop or hesitate at each storyboard step, and the reasons they give | Fewer than 3 of 10 hesitating at any single step; any step that stops 5 or more is redesigned before launch |
Three points about this table. The order is by risk, not by how interesting the test is - assumption 1 makes the other two irrelevant if it fails. The fidelity column always carries a reason. And every threshold is a number that was written down before the test ran, so a disappointing week cannot be talked upwards afterwards.
Apply it to your project
Section titled “Apply it to your project”-
Pick the concept, then write the uncertainty. Choose the most interesting concept you have and write the single biggest key uncertainty about it in one sentence. If you cannot get it into one sentence, you have more than one uncertainty and must pick.
-
Turn it into a falsifiable assumption. Rewrite the uncertainty as a statement you believe, phrased so that a specific result would prove you wrong. A vague belief cannot fail, and something that cannot fail cannot teach.
-
Rank your assumptions and take only the top one this week. List the critical assumptions, sort them by how badly the venture is damaged if each turns out to be false, and take the highest. Testing the comfortable ones first is procrastination with a spreadsheet.
-
Decide say or do. Ask what people would have to do to prove you right. If the honest answer is an action - a click, a sign-up, a pre-order, a payment, a second visit - design a do test and stop planning interviews.
-
Choose the technique and the lowest fidelity that still works. Pick from the seven: looks-like, works-like, Mechanical Turk, scale model, role-play, storyboard, fake door. Then ask what the roughest version is that could still answer the question, and build that one. Right, rapid, rough.
-
Write the test card before you build. Four lines only: the hypothesis in words, the technique, the quantified measure you will take, and the threshold that would give you confidence. Sign the threshold off with the team while the outcome is still unknown.
-
Recruit real people in the real situation. Name where you will find them and how many you need. Pretest the whole thing on two or three people first and fix whatever they misread, exactly as the audit-and-pretest step demands.
-
Run it, and record everything as it happens. Notes, data, photos, sketches, video. Note especially the things nobody planned to measure - the unexpected reaction is usually the finding.
-
Interpret as a team, then act. Separate what you observed from what you think it means, compare the number to the threshold you set, and then confirm, change one variable and re-run, or go back to the problem space. Write the learning down before starting the next loop.
Key terms
Section titled “Key terms”| Term | What it means in plain words |
|---|---|
| Prototyping strategy | The overall direction of the learning process, including a deliberate allowance for failure |
| Prototyping tactics | The organisation of the actual testing activities: a series of choices about approach, fidelity and measurement |
| Problem space | The first half of the timeline: articulating a wide range of problem ideas, then selecting the problem worth solving |
| Solution space | The second half: generating a wide range of solution ideas, then selecting and developing the concept |
| Divergence and convergence | The repeated widening and narrowing that runs through design, happening once for the problem and again for the solution |
| Fidelity | How finished the artefact is, from a rough sketch to an engineered sample; the dial that sets both cost and honesty |
| Right, rapid, rough | The three prototyping principles: focused on the issue, explored quickly and simply, built at the lowest appropriate resolution |
| Looks-like prototype (Pinocchio) | A non-functional object that mimics the concept, used for reactions to direction and desirability |
| Works-like prototype | A functional rig with little or no visual finish, built to test one or more functions and resolve feasibility |
| Mechanical Turk prototype | A prototype in which humans stand in for the machinery, so an expensive function can be simulated today |
| Role-play prototype and bodystorming | Acting out the experience with real bodies, the natural technique for services, processes and environments |
| Storyboarding | Drawings of a process flow, step by step; digital wireframes are the same technique on a screen |
| Fake door, smoke screen, landing page | A front end that looks more finished than it is, used to measure real interest before anything is built |
| Pretotyping | Testing the initial appeal and the actual usage by simulating the core experience with the smallest possible investment, to see whether this is the right thing to build |
| Say test versus do test | A test that collects what people state versus one that collects what people actually do; the second is the stronger evidence |
| Test card and threshold | The four-field plan - hypothesis, technique, quantified measure, and the pass mark agreed before the test is run |
Test yourself
Section titled “Test yourself”- Distinguish prototyping strategy from prototyping tactics in one sentence each, and say what a tactic actually consists of.
- Describe the double movement of divergence and convergence across the two spaces, naming all four stages in order and saying where prototyping and market introduction sit.
- Give three reasons a low-fidelity artefact can be the better test, and two specific dangers of a highly polished one.
- What is the single question pretotyping asks, and how does it differ from the list of questions prototyping asks?
- Name any five of the seven early techniques and give the situation each is best suited to.
- Why must the threshold be defined before the test is run, and what are the legitimate responses to a result that falls below it?
Revision summary
Section titled “Revision summary”Next: Revenue Models & Pricing → - how the idea actually makes money.