IGNITE EXTENDED / LESSON 06
Can constraints create useful product ideas?
A five-template idea sheet and two testable concepts.
Systematic ideation starts with the components and resources already present in a product or its immediate environment. The template specifies a transformation. The team then considers what benefit the transformed form could provide and to whom.
This reverses the familiar sequence of identifying a need and searching freely for solutions. The unfamiliar form can interrupt habitual thinking, but novelty provides no evidence of demand. Evaluate the resulting idea for customer benefit, technical feasibility and the economic work required to deliver it.
Subtraction removes a component or function while requiring the remaining product to remain useful. Division separates components physically, functionally or in a preserving way, where each part keeps the character of the whole. Multiplication introduces another instance of a component and changes an attribute of that instance.
The distinction matters because a vague instruction to “improve the product” often produces a familiar addition. A template gives the team a specific operation to attempt. Work through the operation before judging the result, then examine the new form for a plausible benefit.
These are different starting points for generating possibilities. Both still require a real customer need and a feasible delivery mechanism before the idea becomes a business proposal.
Begin with a user problem, then explore possible responses.
Systematically change a product, then ask which real needs the new form might serve.
Illustrative application: a need-first process starts with one-handed drinking while carrying equipment. A form-first process removes the bottle’s separate cap, visualizes the new opening and then asks which situations benefit. The structural change itself does not prove that those users want it.
The template method below makes the form-first process repeatable while keeping validation as a later task.
Form-first ideation deliberately changes an existing product before searching for a use. Keep the structural change separate from the later claim that a customer wants it.
List components, functions and nearby resources. For a bottle: container, cap, handle, insulation and the user’s hand. This inventory defines the system in which the transformation will occur.
Apply one specified template. Task unification assigns an additional job to an existing component. Example: make the handle also measure a serving, rather than adding an unrelated digital service.
Describe or sketch the resulting object precisely. Show where the measuring marks sit, how the handle fills and how the user pours. A label such as “smart bottle” is too vague.
Investigate a plausible beneficiary and situation. Does a person preparing drinks away from a kitchen need a measuring tool? Test handling, hygiene and willingness to carry it before claiming value.
The template generates a candidate. Customer evidence and technical constraints decide whether that candidate deserves development.
Five claims: templates are a structured approach to idea generation, derived from patterns in past successful innovations, reproducible when searching for new ideas, usable for products, services and advertising, and product-focused so that function follows form.
The last line is the key reversal: you change the product first and look for the need second. That makes the method fast and generative, but it also means every output is a hypothesis about demand, never evidence of it.
The 2005 iPod shuffle, a music player with no display, and its ads: 'Life is random' and 'Enjoy uncertainty'.
Apple removed a component every competitor considered essential and then found the need the new form served: cheap, tiny, random play during exercise. The ad copy turns the missing feature into the benefit, which is what a well-run subtraction does.
List the internal components the maker controls; identify each one's function; remove or reduce one; visualize the result; search for needs the new product could meet.
Steps 4 and 5 are where teams cheat. Visualize the stripped-down product literally before asking who wants it, and do not quietly restore the removed component's function with a new part.
Google's reCAPTCHA page, 'Stop a bot. Improve a map.', with distorted words from scanned books and house numbers from Street View.
The user's effort, already spent proving they are human, is assigned a second job: transcribing text machines could not read. No new resource is added; an existing one does two tasks, which is why the idea was cheap to scale.
Six steps: list internal components; identify their functions; list external components in the environment or packaging; move an internal function to an external component (and drop the obsolete part); move an external function inside; ask what new functions external components could perform.
The internal/external split is the practical engine. Listerine PocketPaks pushes a function out to saliva; the Yerka bike pulls the lock's function into the frame. For a service like Relay, external components include patients, front-desk staff and the clinic's EHR.
Task unification assigns an additional job to an existing resource. The resource may be a component, a person or something in the immediate environment. The question is whether that resource can perform the new function without undermining its existing purpose.
Attribute dependency creates or changes a relationship between attributes. The relationship can be continuous or triggered by a threshold. The source graphs show why drawing the relationship helps: the graph makes the proposed behavior specific enough to discuss and test.
Economic value concerns money or resources. Functional value concerns the ability to perform a task. Psychological value concerns an experience such as confidence, identity or reassurance. A single offering can create several forms of value.
The customer’s situation determines their importance. A technically superior feature may be irrelevant when the buyer values simplicity or reduced perceived risk. Psychological value also appears in business purchases because people anticipate consequences for their work, reputation and relationships.
Illustrative design relationship: a display changes as contents become hotter.
Physical division (cut it), functional division (separate components with different functions) and preserving division (each part keeps the character of the whole).
The three types produce different questions. Physical and functional division ask what improves when parts are apart; preserving division asks what a smaller unit of the whole could do on its own.
A close-up of a razor cartridge with several parallel blades.
Adding copies is only half the template: the extra blades are set at different positions so each does a slightly different job. Copies that are identical to the original usually produce plain addition, not an idea.
Two graphs of color against temperature: a flat line (no dependence) and a step that drops above a threshold (color depends on temperature).
The graph forces precision: which variable drives which, and whether the relationship is continuous or a threshold. 'Make it smarter' cannot be drawn; 'send to more patients as the slot gets closer' can.
A grid with ringing tone, ringing volume, speaker volume, screen display and vibrating mode as columns, and the same five plus ambient noise, signal strength and the user's calendar as rows; the diagonal is blocked out.
Each empty cell is a candidate dependency: ringing volume that rises with ambient noise, or vibrate-only during calendar meetings, both of which later shipped. The external rows are where most useful dependencies come from.
Subtraction for complex products or cost control; task unification for cost control; division and multiplication for simple quantitative improvements; attribute dependency useful generally but hard.
Use this as a selection guide, not a ranking. A cost-constrained startup gets most from subtraction and task unification, because both remove or reuse resources rather than adding them.
A fictional team adds a second compartment because “more storage means more value.” A template exercise asks a sharper question: who benefits, why must the compartments differ, and what new cleaning or carrying cost appears?
What would make multiplication useful rather than merely additional?
Wharton researchers compared 200 product ideas from MBA students with 200 from GPT-4 for the same brief. A blind consumer survey gave the AI ideas a higher average purchase probability (47% versus 40%), and GPT-4 produced its 200 ideas in about 15 minutes. Volume is no longer the bottleneck; screening is.
In a 2024 Science Advances experiment with 300 writers and 600 evaluators, access to AI suggestions raised individual novelty and usefulness ratings but made the resulting stories about 10.7% more similar to one another. A team that asks the same model the same open question will converge on the same ideas as its competitors.
The practical response is to give the model a specific operation, not an open brief: list Relay's components and ask for subtraction of each one, or for dependencies between hours-to-slot and every controllable attribute. The structure that disciplines human ideation also forces varied machine output, and the human job shifts to visualizing, screening and testing.
Sources: Wharton Knowledge: Is ChatGPT a Better Entrepreneur Than Most? (Girotra, Meincke, Terwiesch, Ulrich) · University of Exeter: AI found to boost individual creativity but results in less varied content (Doshi and Hauser, Science Advances 2024) · Goldenberg, Mazursky and Solomon, The Fundamental Templates of Quality Ads, Marketing Science 18(3), 1999
Choose a familiar object: a bottle, backpack or desk lamp. List its components and the function of each.
Create one concept with each template. Select two, identify a plausible user need and a concrete cost or drawback. Choose one assumption to test.
Relay's Wizard-of-Oz pilot runs at 3 clinics: Maya texts waitlisted patients from a shared phone and logs results in a spreadsheet. A typical clinic loses about 42 slots a week to late cancellations and no-shows (illustrative), and Relay's target is to refill 40% of them, against 10-15% for front-desk phoning. Before Daniel builds v1, the team wants a wider concept set than 'automate what Maya does'. Relay's components: the waitlist, the outbound SMS, the patient's reply, the reminder message the clinic already sends, the front-desk staff, the EHR schedule and the cancellation event itself.
Produce a template idea sheet for Relay: one concept per template (five), each visualized concretely. Select three, and for each state the beneficiary, the benefit hypothesis, the new cost or risk, and one cheap test Maya can run in the existing pilot.
Bolting on a new chatbot or app and calling it task unification defeats the method. The template only counts if an existing component or nearby resource (the reminder text, the patient, the front desk) takes on the new job.
Teams remove the waitlist, then reintroduce it as an opt-in 'move-up list'. If the removed component's function comes back under a new name, you have not examined the stripped-down form.
Multiplying a component without changing an attribute of the copy (two identical reminders instead of one) yields more of the same. Ask why the copy must differ and what that difference does for the user.
A five-template idea sheet for Relay with three screened concepts (move-up offers, reply-to-cancel reminders, lead-time-dependent batch size), each with a benefit hypothesis, cost and pilot test, and a recommendation on which to test first.
Next lesson: Market segmentation & target choice
Check your understanding
Question 1
Relay's team proposes dropping the waitlist entirely and instead offering any opened slot to patients already booked later that week. Which template produced this concept?
An essential component, the waitlist, is removed and the remaining system (existing bookings) must still deliver the function. It is not task unification, because no existing resource takes on a second job; the existing bookings simply become the candidate pool.
Question 2
A clinic's tablet check-in kiosk, already used to confirm insurance details, is set up to also collect a short pain-and-function questionnaire before each visit. Which template does this illustrate?
An existing resource (the kiosk and the patient's waiting time) is given an additional task. Multiplication would require adding a second, modified kiosk; here nothing is added.
Question 3
After a subtraction exercise, a team has a striking concept and says: 'Nobody has built this, so the market is wide open.' What is the right next step?
Templates generate form, not demand. Novelty can mean an untapped need or a need that does not exist. The concept becomes a candidate only after a plausible beneficiary is named and the benefit and new burden are tested with them.
Question 4
Relay drafts an attribute-dependency concept: 'make outreach smarter as the slot approaches.' What change makes it a testable dependency?
A dependency must specify the driving attribute (hours to slot), the dependent attribute (batch size) and whether the link is a threshold or continuous. 'Smarter' and 'add AI' name no relationship that can be drawn or tested.
Question 5
Illustrative: a clinic has 42 empty slots a week and Relay fills 40% of them. A concept raises the fill rate to 50%. At $110 per visit, how much extra weekly revenue does the concept add over the current 40%?
Extra visits = 42 x (0.50 - 0.40) = 4.2; 4.2 x $110 = $462. $1,848 is the revenue from the 40% fill itself (16.8 x $110), and $4,620 treats all 42 slots as gained.
Question 6
Relay adds a second waitlist called 'flex list' for patients who will accept any therapist and any time, alongside the normal list of patients with specific preferences. Which template is this?
The original list stays intact and a second instance is added that differs in one attribute (preference strictness). Division would separate the existing list into parts; nothing is separated here, a modified copy is added.
Workbook
Module output: An evidence log, problem frame, prototype, segment choice, positioning and route to market.
Sources
Original pages and full documents are in the Library edition.