for research labs, transfer offices, venture funds, family offices, and builders
Learn how to tell if an idea is worth building. On your own ideas.
Phazur Labs · Program
The method works on the situation, not the science. Different fields stall in different places, and each place gives a warning first.
01Where
Products, platforms, and internal tools where a roadmap decision costs a year.
Common stall — the second customer, when what worked for the first turns out to have been consulting.
See how it runs for softwareDevices, instruments, and systems moving from working unit to production.
Common stall — the hundredth unit, when the process that made one was never designed.
See how it runs for hardwareFormulations, packaging, and new lines heading toward retail.
Common stall — the second buyer meeting, when terms arrive that the margin won't carry.
See how it runs for consumer goodsDevices, diagnostics, and therapies with a clearance path and a payer at the end.
Common stall — clearance arrives and reimbursement doesn't, so the product is legal and unsellable.
See how it runs for regulated productsLab results, disclosures, and spinouts with a long road to a buyer.
Common stall — the science holds, and no operator exists who can carry it nine years.
See how it runs for deep techOne committee deciding across categories, with no shared way to compare them.
Common stall — the decision goes to whoever presented best that week.
See how it runs across a portfolio02Why
Most teams decide what to build on judgment. Judgment is often good, but it can't be checked, taught, or handed to the person who takes the job next. When the decision is questioned two years later, there is nothing to show.
Dean is a method that leaves something behind. Ten steps, six documents, two tests. You work your own ideas from the first session. There are no case studies and no exercises.
Dean covers the first gate — the decision to commit — and then hands off. It does not replace design, prototyping, or manufacturing engineering. It decides which ideas earn them.
03Placement
Dean
Decide
Is this worth building, and what would prove it isn't?
Stage 2
Design
Requirements, architecture, design for manufacture and assembly.
Stage 3
Prototype
Build it, break it, iterate it.
Stage 4
Verify & validate
Does it meet spec, and does it meet the need?
Stage 5
Make & launch
Scale, quality, market.
Everything that happens before the first real money is committed — the claim, the comparison to prior attempts, the stall points, and the two tests that settle it. It also sets the condition for stopping, before anyone is invested in continuing.
Design work, tooling, manufacturing engineering, or clinical strategy. Those are their own disciplines with their own experts, and Dean's job is to make sure they're pointed at something worth their time.
The two tests are a small verification-and-validation run pulled forward. Verification that the claim holds, validation that a buyer behaves the way you expect — done before the design work rather than after it, when changing your mind is still cheap.
Ten steps, in order. Each is answered in writing, and the writing is what you keep. The first five screen a whole portfolio. All ten assess one idea.
04Method
The claim is written and frozen before anything is searched, so what you find can't quietly change the question you were asking.
"A better battery" is not a claim. "A cell that holds 80% capacity after 800 cycles at a price a mid-market automaker will pay" is a claim, because a result can prove it wrong.
Matching is on the shape of the attempt — same stage of development, same kind of protection, same sort of buyer at the end. A device with one clinical site, a provisional filing, and no distribution partner is compared against everything that went out with that structure — not against other devices in its category.
The set includes what reached market and what didn't, so the comparison isn't stacked in either direction. Four surviving comparisons is the minimum.
Market figures are the worst offenders. A number appears in a vendor report, gets quoted in an article, gets quoted from the article in a deck, and by the fourth hop nobody can name the original or how it was calculated.
Four levels, from filed documents down to estimates. An estimate never becomes a fact by being repeated. Anything resting on the weak end gets flagged, and the flag travels all the way into the room where the decision is made.
Six things get compared: how far along it is, what protects it, who has already funded it, who is running it, who buys at the end, and when it happened.
The adjustments move toward your idea, never the other way. Any comparison needing more than four adjustments gets dropped — but it stays visible in the grid, struck through, with the reason beside it. A grid that adjusts everything into agreement has stopped being evidence.
Fifty stall points are catalogued across demand, access, cost, capital, timing, regulation, execution, and team. Each one has a warning that shows up before it happens, and a place where that warning shows.
Three most likely get named for your idea, with the warning and where to look for it.
Rarely the expensive one. Four weeks of shelf stability at real distribution temperatures beats a national concept study, because it can come back no.
What each result means is written down before the test starts, and not revised after. That single rule is what stops the answer from being renegotiated once the answer is known.
A crowded field shows interest, not proof — the two get confused constantly, and the difference decides your timing.
There is a second effect worth knowing. An idea clear enough to compare cleanly is clear enough for other people to see. That is a finding about how long you have.
If similar attempts took nine years and your team has two, that decides it, no matter how good the idea is. Applies to a researcher who won't leave the lab, a program director split across three other efforts, and a founder who doesn't exist yet.
Assessed separately from the idea, so a team limit never gets written into the file as a technical limit. Those are different problems with different fixes.
Set early, it's a decision. Set late, it's an argument. Signed by whoever actually has authority to stop the work, which is usually not the person who wrote it.
This is the step everyone skips, and it's why most teams can't tell you how often they're right. Without it the other nine steps never improve.
Two numbers come out: where you started, and where you are now.
The first nine steps tell you what's likely. Only a test tells you what's true. Every idea that survives gets two, and both run during the program.
05Tests
A shelf-stability run at real distribution temperatures. A cycle test on the cell. A load test on the system. A paid pilot with one customer. Whatever the cheapest honest version is for your field.
Written down before it starts: what result means go, what result means stop, and what result means the estimate moves. Three outcomes, defined tightly enough that none can be reinterpreted on the day.
Five to eight real buyers or users, given the thing and watched while they try to use it. Not a survey. Not a focus group. You watch what they do and where they get stuck, because what people say they'd pay for and what they reach for are different things.
Five is enough. The pattern of where people get stuck repeats early, and the sixth session mostly confirms what the first five showed.
Run separately and on purpose. A product can work perfectly and still lose at the second retail conversation. A product people love can fail stability testing. Testing one and assuming the other is how a year disappears.
Six documents per idea. Filled in for the ideas you work, blank for everything after. Each one is written to be handed to someone.
06Artifacts
What it is, how far along, what protects it, who's funded it, who's running it, who buys at the end, who tried before, where it breaks, and the test.
The estimate is a range with a marked midpoint, drawn from the comparison set rather than asserted. At the bottom, the source mark on the weakest number the argument depends on. A page that won't show its floor is a brochure.
One row per prior attempt: its structure, its outcome, the adjustments applied, and the source mark behind each figure. The comparisons you threw out stay in the grid, struck through, with the reason written beside them.
That last part is what makes the one page survive scrutiny. A grid showing only what survived tells a reader nothing about how hard you looked.
Two pages, dated, and not revised after the results come in. The second page is the fork: this result proceeds, this result moves the estimate, this result stops.
It turns a request for money into a request for a specific piece of information at a named price.
One row per stall point: the warning that shows first, where it shows, who checks, and how often. A name and a date in every row.
Not a risk register. A risk register lists what might go wrong and gets read once. This gets read on a schedule, which is the reason you find out early instead of at the review.
One page, signed and dated before the evidence arrives. The date is the whole argument. Written before, this is a decision the organization made. Written after, it's a position someone is defending.
The call, the date, the confidence attached, the outcome, and the gap between them. Starts with the three decisions from session one and never closes.
Everything else in the file is an argument about the future. This is the only page that gets more true over time.
Four sessions. You work your own ideas in all of them.
07Teaching
Three decisions your organization already made, where the outcome is known, worked without hindsight. The results get scored. Most teams have never measured how often they were right, and this is the number everything after gets compared to.
Everything currently in front of you, run through the first five steps and sorted three ways: no path, worth a test, worth the full method. You leave with a ranked list and the reasoning written down.
One idea taken through all ten steps, taught by working it in front of you. Both tests get designed here and signed before anyone runs them.
Between sessions — the tests run
A second idea, worked by your team alone. The faculty examines it afterward, hard, in a room with no money on the table. The test results come back during this session.
Three decisions you already made, scored against what actually happened. If the method doesn't catch what your current process missed, you'll know in week one.
Three ways in. Most organizations start with the first and add the third.
08Formats
Most common
Four sessions with three or four other organizations that aren't competing for the same deals.
Single organization
The same four sessions, run for one team, scheduled against your own funding or greenlight cycle.
After the program
Outside eyes on the two steps you can't do on your own work — checking your sources, and hunting weaknesses in your own case.
Six things move, and each one is measurable against where you started.
09Benefits
Weeks to a defensible answer instead of quarters of circling. The screening step sorts a full portfolio in one session, which is usually the first time anyone sees the whole thing ranked.
The cheapest test comes before the expensive commitment. A four-week run that returns no is the least expensive thing in the whole process.
Stop conditions get set before anyone is invested in continuing, so ending something becomes a scheduled event rather than an admission.
Five to eight real users, watched rather than surveyed, before the design work starts — not after launch, when the finding arrives too late to act on.
Everyone compares the same six things in the same order, so a decision stops going to whoever presented best that week.
Two years on, when someone asks why the money went there, the file answers instead of your memory.
10Capability
A full portfolio sorted, with written reasoning behind every placement — including the ones you'll be asked to defend.
A no supported by comparable cases instead of opinion, which is the difference between a decision and a preference.
Find prior attempts of the same shape in a field you haven't worked before, and know when you've found enough.
Named warning signs, assigned to people, checked on a schedule.
Without us. Including writing the criteria before the run, which is the part teams get wrong.
How often your team is right, measured against where you started. Most organizations cannot produce this number.
11Audience
Which programs advance, which get another cycle, and which should have stopped two years ago.
More disclosures than anyone can examine, and every filing decision commits money someone will eventually ask about.
Screening where the science sits outside your expertise and the comparison set isn't obvious.
Where the clearance path is understood and the reimbursement path decides whether any of it was worth it.
Software, hardware, or consumer goods, where a roadmap decision costs a year.
Holding something you haven't built yet, and wanting the argument now rather than the post-mortem later.
Two to four people from the same organization, including one with authority to stop a project. The program runs on live work and can't be attended in pieces.
12Limits
Or the formulation, or the clinical result. Your experts do that, and they're better at it. Dean starts after.
Most of the time the answer is no. The value is in the reasoning that survives afterward.
When nothing comparable exists, the method says so rather than inventing a comparison. That usually points toward more time and non-dilutive money, not a commitment.
Checking your own sources and hunting weaknesses in your own case both ask you to be neutral about your own outcome. Practice doesn't fix that.
13FAQ
Sessions three and four have to be far enough apart for a real test to finish, which is usually two to six weeks depending on your field. Everything else is scheduled against your calendar, not ours.
Institutions working to a funding or greenlight cycle should say so in the application, and the sessions get placed around it.
The application asks nothing confidential. Inside the program, you work your own ideas and control what gets discussed in front of the group. Teams with sensitive portfolios generally take the private program.
Cohort and private programs are priced differently, and institutions working under economic development or federal funding constraints have terms that need naming early. The test itself is a separate cost, sized during session three and always the cheapest version that can return a no.
An idea with no affordable test is an idea whose next step is expensive and irreversible. That is worth knowing before you take it, and it usually means the claim in step one was drawn too wide.
What you leave with is a filled file on a real idea, a test result, and a measured hit rate. Those are checkable. A certificate isn't.
14Apply
Applications are reviewed as they arrive. Groups are small — the peer sessions only work among teams that aren't competing for the same deals, which limits how many we admit in one field.
No confidential material is requested. Every applicant receives a reply with a reason.