Discovery guide

Riskiest Assumption Examples

16 real-world assumptions across the four types — each phrased as a testable belief, with the smallest test that would settle it and a success criterion decided up front.

A warning before the list: your riskiest assumption can’t be copied from a website — “riskiest” is a property of your evidence, not of the assumption. Use these examples to calibrate what a sharp, testable belief looks like, then map your own by importance × evidence (or grab the printable kit and do it on paper in half an hour).

Desirability Do customers actually want this?

The most common killer. Teams rarely doubt demand out loud — it feels disloyal to the idea — so desirability assumptions go untested straight into the roadmap. If the whole bet collapses when nobody cares, this is your riskiest layer.

B2B SaaS · retention

We believe busy team admins will read a weekly activity digest.

Smallest test: Fake-door: add a “Weekly digest” toggle in settings and measure opt-ins for two weeks.

Success criterion: ≥ 20% of active admins turn it on.

Services · fintech

We believe freelancers will hand over invoice chasing to someone else.

Smallest test: Concierge: offer to chase invoices manually for 20 freelancers this month.

Success criterion: ≥ 5 say yes and actually send you an overdue invoice.

E-commerce

We believe shoppers will trust a size recommendation based on their past returns.

Smallest test: Fake-door “What’s my size?” button on the product page, with a one-question survey behind it.

Success criterion: ≥ 15% of viewers click, and survey answers name trust, not curiosity.

Dev tools · AI

We believe developers will let an AI agent open pull requests in their repository.

Smallest test: Ten interviews where you ask them to connect a sandbox repo on the spot — not “would you?”, but “will you now?”.

Success criterion: ≥ 5 of 10 connect the repo in the session.

Usability Can people figure out how to use it?

Rarely the riskiest at the idea stage — you can usually fix usability with iteration. It becomes the linchpin when the whole value depends on unaided self-serve: onboarding flows, imports, anything where a human helping defeats the point.

Martech

We believe non-technical marketers can build a three-step email journey without reading docs.

Smallest test: Prototype sessions with five marketers; no hints, think-aloud.

Success criterion: ≥ 4 of 5 complete it unaided in under ten minutes.

Logistics · mobile

We believe drivers will complete proof-of-delivery photos one-handed at the door.

Smallest test: Field test with five drivers on real routes for one day.

Success criterion: ≥ 80% of deliveries photographed first try, no support calls.

Data import

We believe finance teams can map their own CSV columns without a support call.

Smallest test: Usability test with five real customer files — theirs, not your tidy sample.

Success criterion: ≥ 4 of 5 finish the mapping unaided.

Feature discoverability

We believe users will find export where we put it, inside the share menu.

Smallest test: First-click test on a static screenshot with ten users.

Success criterion: ≥ 8 of 10 first-click the share menu.

Feasibility Can we build and run it, with our team and constraints?

The type engineers surface naturally — and the one type where the test is often a build. The discipline is timeboxing: a spike is a throwaway experiment with an end date, not the first sprint of the feature wearing a lab coat.

Search · scale

We believe we can return search results in under 300 ms at ten times today’s data volume.

Smallest test: Two-day spike: load a synthetic 10× corpus into a prototype index and benchmark.

Success criterion: p95 latency < 300 ms on realistic queries.

Open banking

We believe the bank API delivers transactions within 24 hours of posting.

Smallest test: Pull real data for three pilot accounts for one week and measure the lag.

Success criterion: ≥ 95% of transactions visible within 24h — measured, not promised in the API docs.

AI feature

We believe LLM summaries of support tickets are accurate enough to ship without human review.

Smallest test: Offline eval: summarize 100 historical tickets, have two support leads grade them blind.

Success criterion: ≤ 2% of summaries contain a harmful error (wrong customer, wrong promise, wrong status).

Enterprise onboarding

We believe a new tenant can be onboarded without a migration engineer in the loop.

Smallest test: Throwaway self-serve import prototype, run on two friendly customers’ real exports.

Success criterion: Both imports complete without an engineer touching the data.

Viability Should we build it — does it work for the business?

The type founders test last and regret first. Price, margin, and cost-to-serve assumptions feel like “later problems” — until a validated, desirable, usable product turns out to be a bad business. Viability tests are mostly conversations and spreadsheets, which is exactly why skipping them is inexcusable.

B2B SaaS · pricing

We believe teams will accept per-seat pricing rather than a flat team price.

Smallest test: Quote per-seat in the next ten sales conversations and count the pushback.

Success criterion: ≥ 6 of 10 accept the framing without negotiating toward flat.

Go-to-market

We believe we can acquire customers under $200 CAC with content alone, no paid spend.

Smallest test: Spreadsheet first (traffic × conversion × close rate), then one 90-day organic cohort.

Success criterion: The math closes at realistic conversion rates before you commit the quarter.

Marketplace · supply side

We believe restaurants will accept a 10% take rate.

Smallest test: Fifteen supply-side conversations with a real one-page agreement, not a hypothetical.

Success criterion: ≥ 8 sign a letter of intent at 10%.

Self-serve economics

We believe support load stays under one ticket per customer per month at self-serve pricing.

Smallest test: Measure the pilot cohort for a month and price the support minutes against the plan margin.

Success criterion: Support cost < 15% of plan revenue.

Writing your own

What the 16 above have in common — and what to copy when you write yours.

One belief per card

“Admins will find and love the digest” is two assumptions wearing a trench coat. Split it — each gets its own test.

Name a specific actor and behaviour

“We believe [who] will [do what]” is falsifiable. “Users want more engagement” is not — nothing observable could prove it wrong.

Write the success criterion first

Decide what “validated” means before the data arrives, or you will argue any result into a pass.

Prefer behaviour over opinion

Connecting a repo beats saying “sounds useful.” Every type has a behavioural version of its test — reach for it.

Find yours in 30 minutes

The printable kit walks you through the inventory, the importance × evidence grid, and a fillable test card. No email required.

Get the kit

Frequently Asked Questions

Common questions about riskiest assumptions

Give every assumption a home and a status

In Outcomify, assumptions are first-class nodes under the solutions that depend on them — untested, validated, or invalidated, with the evidence attached.