../

Playbook

The master sheet for getting from "I have an idea" to "real users are using it" in 6–8 weeks. It defines the six phases, the artifact each one hands to the next, the timeboxes, the decision rules, and the specific ways technical founders stall between phases. Each phase has its own sheet: ideation, validation, product design, building the MVP and launch and iterate. Attributed advice lives in the advice sheets; the stack assumed here is Next.js and Postgres.

Why ideas stall

Ideas rarely die because the founder can't code. They die in the gaps between phases, where nobody tells you what the next step is, what "done" looks like, or when to stop. The pattern for technical founders is consistent: the comfortable activity (building, researching, choosing tools) expands to fill all available time, and the uncomfortable one (talking to strangers, asking for money, shipping something embarrassing) never happens.

SymptomUnderlying causeFix
New idea every week, none pursuedendless ideation: evaluating feels productive and carries no risk of rejectiontimebox ideation to 1 week, score with a fixed matrix, commit to the winner for one validation cycle
Months of building, zero users have seen itfear of judgment disguised as "it's not ready yet"; building in secretshow something to one real user this week; public launch date
Rewriting the auth flow for the third timeperfectionism: polishing what you control instead of testing what you don'tfixed time, variable scope; a "not now" list; ship on the date whatever is done
"I don't know what to do next"ambiguity about the next step: no phase model, no exit criteriause the phase table below; every phase ends with a named artifact
MVP spec has 14 featurestoo-big scope: MVP treated as "v1 minus a bit" rather than an experimentone core job, one user type, one happy path; everything else to "not now"
Comparing five ORMs, three hosts and two auth librariestool or stack yak-shaving: tool-shopping is building-adjacent procrastinationuse the stack you already know; once a stack decision is made, it stays closed until launch
Project drifts for monthsno deadline: work expands to fill the time available (Parkinson's law)timebox every phase; announce the launch date publicly
Opinions about users, but no conversationsno user contact: the building feels safer than the marketone user conversation per day, every day, in every phase
Doing a course on Kubernetes "for the MVP"confusing learning with building: learning feels like progress but produces no evidencelearn only what the next deliverable requires, just in time
Waiting for the perfect ideafear of commitment: any real idea has visible flawsan idea only has to be good enough to test; tests are cheap
Launch keeps slipping by "one more feature"one-more-feature syndrome: each feature is a reason not to face usersfeature freeze 3 days before launch; new ideas go to the backlog
Stuck after a lukewarm launchno decision rules: ambiguity about whether to continuepre-commit kill/continue criteria before each phase starts

The phase model

Six phases. Each answers one question, produces one artifact, and has a hard timebox. Default timeboxes assume a solo or two-person technical founder working close to full time; roughly double the timeboxes (never the scope) if you are working evenings and weekends.

PhaseQuestion it answersKey activitiesArtifactsExit criteriaTimeboxBiggest risk
0. IdeationWhich problem is worth testing?problem journal, idea list, scoring, bottom-up sizing, "why now"problem hypothesis one-pager; list of 20 people to talk toone idea chosen; riskiest assumptions ranked; 20 names≤ 1 weekendless ideation; picking a made-up or tarpit idea
1. Problem validationDoes this problem exist, is it painful, will people pay or switch, and can I reach them?10–20 interviews, synthesis, landing page, smoke tests, pre-salesvalidated problem briefpre-set evidence thresholds met (or idea killed)1–2 weeksfalse positives from friends and compliments
2. Solution designWhat is the smallest thing that delivers the must-have outcome?scope cut, user flow, clickable prototype, prototype testsone-page spec; prototype; "not now" list5+ target users complete the core task on the prototype1 weekscope creep; designing for imagined users
3. Build the MVPCan I deliver the core outcome reliably enough for real use?walking skeleton, core flow, analytics, onboarding, payments if testeddeployed MVP with analytics; launch listcore job works end to end in production for a real user2–4 weeks (hard cap 6)polishing; tool-shopping; building in secret
4. Launch to first usersWill real users adopt it, and what breaks?personal onboarding of first users, launch posts, supportlaunch results (activation, retention signals, quotes)first cohort onboarded; data and feedback collected1 weekbig-bang launch; vanity metrics
5. Iterate toward product–market fitWhat must change for users to keep coming back and pay?weekly build–measure–learn cycles, retention cohorts, PMF surveyiteration backlog; weekly metricsretention curve flattening; strong PMF signalsweekly cycles"one more feature" instead of fixing retention

Total target: idea → users in 6–8 weeks. That is phases 0–4. Phase 5 has no fixed end; it runs until the product has found fit or you kill it. The detail for each phase is in its sheet: ideation, validation, product design, building the MVP, launch and iterate.

The pipeline at a glance

 PHASE 0          PHASE 1            PHASE 2           PHASE 3          PHASE 4          PHASE 5
 Ideation   -->   Problem      -->   Solution    -->   Build the  -->   Launch to  -->   Iterate
 (<= 1 wk)        validation         design            MVP              first users      toward PMF
                  (1-2 wks)          (1 wk)            (2-4 wks,        (1 wk)           (weekly)
                                                        cap 6)
    |                 |                  |                 |                |                |
    v                 v                  v                 v                v                v
 Problem          Validated          One-page spec     Deployed MVP     Launch results   Iteration
 hypothesis  -->  problem brief -->  + prototype  -->  + analytics -->  (cohort data, -> backlog +
 + 20 names       (quotes, WTP,      + "not now"       + launch list    quotes, bugs)    weekly metrics
                  channel)           list                                                    |
                                                                                             |
    ^                 ^                                                                      |
    |   kill / pivot  |                        kill / pivot / persevere                      |
    +-----------------+---------------------------------------------------------------------+
 
 Running underneath every phase:
   - 1 user conversation per day          - deploy something every day (from phase 1)
   - Friday review against exit criteria  - public deadline for the launch date

Arrows back to the left are normal. A failed validation that sends you back to ideation in week 2 is the process working, not failing: it saved you 5 weeks of building.

What an MVP is (and is not)

TermSourceDefinitionWhat it emphasises
Minimum viable productcoined by Frank Robinson (SyncDev), 2001"the unique product that maximizes return on risk for both the vendor and the customer" (SyncDev's wording)right-sized: big enough to cause adoption and sales, small enough not to be bloated and risky
Minimum viable productEric Ries, "Minimum Viable Product: a guide" (blog, 3 Aug 2009), popularised in The Lean Startup (2011)"that version of a new product which allows a team to collect the maximum amount of validated learning about customers with the least effort"learning, not a small product; the book frames it as the fastest way round the build–measure–learn loop
Minimum lovable productcommon industry variant; no single clear originatorthe smallest product users would genuinely love, not just tolerateemotional response; retention
SLC: simple, lovable, completeJason Cohen, "Your customers hate MVPs. Make a SLC instead." (A Smart Bear, Aug 2017)simple enough to build fast, lovable enough that people want it, complete enough to do one job fullyrespect for the customer: Cohen argues "'MVP' implies a selfish process, abusing customers so you can 'learn'"

How to reconcile them: Ries's definition decides what you build first; Cohen's decides how it should feel. An MVP is an experiment aimed at your riskiest assumption. It can be a landing page, a spreadsheet you run by hand, or a working app. If it is software that real users rely on, make it SLC: one job, done completely, without broken edges. Cohen's line is the useful test: customers would rather use "v1 of something simple" than "v0.1 of something broken".

An MVP isAn MVP is not
the cheapest test of the riskiest assumptiona smaller version of your eventual product roadmap
scoped to one user type and one core joba platform, a marketplace for everyone, or "phase 1 of 4"
something real users touch, and which produces evidencean internal demo, a pitch deck, or a Figma file no user saw
allowed to be manual behind the scenesrequired to scale, be multi-tenant or be fully automated
deliberately embarrassing in the parts that don't matterbroken in the part that does

Types of MVP

Choose the cheapest type that can produce the evidence you need. Cost is for a technical founder; signal strength follows the ladder in validation: compliments are worth nothing, money is worth the most.

TypeWhat you doTypical costWhat it testsSignal strengthExample or note
Landing page / smoke testpage describing the product with a sign-up or "request access" buttonhours–1 daywhether the pitch attracts the target segmentweak–medium (an email is cheap)Buffer started as a two-page site: pitch, then an email form (Joel Gascoigne, 2010–11)
Fake doora button or pricing tier for something not yet built; clicks are measured, then an honest "coming soon"hoursdemand for a specific feature or price pointweak–mediumBuffer inserted a pricing page between the two pages to see which plan people clicked
Explainer video / demoa video of the product working (possibly faked)1–3 dayswhether the concept is understood and wantedmedium when it drives sign-upsRies describes Dropbox's early demo video driving beta sign-ups (The Lean Startup)
Conciergedeliver the outcome manually, openly, to a handful of customersdays of your time per customerwhether the outcome is valued enough to use and pay forstrong if they pay or keep coming backRies's Food on the Table case (The Lean Startup); see also Paul Graham, "Do Things that Don't Scale" (2013)
Wizard of Ozreal-looking product front end; humans do the work behind it, unseendays–2 weekswhether users behave as if the product existsstrong (real behavior)Zappos is the usual story: shoes photographed in shops and bought at retail when orders came in
Piecemeal / no-codestitch existing tools (forms, spreadsheets, Stripe payment links, email) into a working service1–5 daysthe whole flow, end to end, with real usersstrongtechnical founders skip this out of pride; don't
Single-feature productbuild only the one feature that delivers the core outcome2–4 weeksretention and willingness to pay on real softwarestrongest before scaleBuffer's first build did one thing: queue tweets
Pre-salesell before building: deposits, paid pilots, letters of intentdayswillingness to pay, in moneystrongest (cash)refund promise and delivery date required; see validation

Rule of thumb: test problem risk with interviews, demand risk with a landing page or pre-sale, value risk with concierge or Wizard of Oz, and only then test feasibility and retention with software. Technical founders do this backwards because software is what they are good at.

Handoffs between phases

The single most effective anti-stall device is an explicit artifact at each boundary. You cannot start the next phase without it, and it is the first thing the next phase opens, so no phase starts from a blank page.

From → toArtifact passedMust contain"Done" test
0 → 1Problem hypothesis one-pager (the problem statement)target customer, problem, current alternatives, why now, riskiest assumptions ranked, 20 names to contactcould a stranger book your first interview from it?
1 → 2Validated problem briefwho, problem in their words with quotes, frequency, workaround and its cost, willingness-to-pay evidence, reachable channel, 1–3 must-have outcomesevery claim has an interview or data point behind it
2 → 3One-page spec + clickable prototype + "not now" listone user type, one core job, happy-path flow, success metric, what is explicitly out5 target users completed the core task on the prototype
3 → 4Deployed MVP + analytics + launch listproduction URL, core flow working end to end, activation event tracked, 20–100 named people to invitea real user completed the core job in production without your help
4 → 5Launch resultswho signed up, who activated, who came back, quotes, bugs, channel performancewritten up in one page, with the next hypothesis stated
5 → 5Iteration backlog + weekly metricsranked problems from users, retention by cohort, this week's experimentone experiment shipped and measured per week

Reducing friction between phases

TechniqueHowWhich stall it kills
Decide in advancewrite kill/continue criteria before each phase starts ("if fewer than 5 of 12 interviewees raise the problem unprompted, I go back to ideation")motivated reasoning after the fact; zombie projects
Overlap phasesbuild the landing page during validation; recruit launch users during the build; write launch posts during week 1 of the builddead time at boundaries; launching to nobody
Walking skeleton on day 1deploy a trivially thin end-to-end slice to production on the first day: domain, Next.js app, Postgres, deploy pipeline, one page"deployment is a problem for later"; big-bang integration; tool-shopping
Fixed time, variable scopeset the appetite (the time budget) first, then shape the solution to fit itscope creep; "one more feature"
Public commitments and deadlinesannounce the launch date to people whose opinion you care aboutinfinite polishing; drift
Daily shippingsomething reaches production (or a user) every working daybuilding in secret; fear of judgment
A "not now" listevery idea that isn't in scope goes on a visible list with a date to reconsiderthe guilt that drives scope creep; losing good ideas
Blank-page-free templatesevery phase starts by filling in the previous phase's artifact and this phase's templateambiguity about the next step
One user conversation per daycalendar block; counts in every phase, including the buildno user contact; designing for imagined users
Pre-chosen stackdecide once, write it down, don't revisit before launchyak-shaving

Fixed time, variable scope

Borrowed from Basecamp's Shape Up (Ryan Singer, 2019). Shape Up calls the time budget the appetite: "You can think of the appetite as a time budget for a standard team size." The contrast with estimation is the whole point: "Estimates start with a design and end with a number. Appetites start with a number and end with a design." Shape Up works in six-week cycles, with small-batch projects of one or two weeks. For an MVP: the build appetite is 2–4 weeks and the hard cap is 6. When the cap arrives, you cut scope, not time.

The walking skeleton

A term from Alistair Cockburn: the thinnest possible implementation that exercises the whole architecture end to end. On day 1 of the project, spend at most 2 hours: create the app, connect Postgres, deploy to production on your real domain, add analytics. Every later deploy is then routine, and the landing page for validation has somewhere to live. See building the MVP.

Walking skeleton, day 1 (max 2 hours):
[ ] repo created, pushed
[ ] Next.js app renders one page at your real domain
[ ] Postgres connected; one table; one write, one read
[ ] deploy on push to main
[ ] analytics snippet installed; one event fires
[ ] error tracking installed
[ ] STOP. No auth, no design system, no CI matrix.

The six-week plan

The fast path: 6 weeks from idea to first users. The 8-week variant adds a second validation week and a fourth build week; use it if you are part-time or the interviews are hard to book. Week 1 is planned by the day because that is where most people stall.

WEEK 1 - PHASE 0 IDEATION (and skeleton)
  Mon  AM  Dump every idea and every problem from your problem
           journal into one list (aim for 30+; 100 if you can).
       PM  Walking skeleton deployed (<= 2h). Pick stack: done.
  Tue  AM  Cut the list to 10 using the "tarpit / made-up" filters.
       PM  Score the 10 on the idea matrix. Keep the top 3.
  Wed  AM  For each of the 3: competitors and alternatives scan
           (45 min each), bottom-up size (30 min each).
       PM  Write a draft problem hypothesis for each of the 3.
  Thu  AM  Choose ONE. Rank its riskiest assumptions.
       PM  Write kill/continue criteria for validation.
           Build the list of 20 people to talk to.
  Fri  AM  Send 20 outreach messages. Book >= 5 interviews.
       PM  Landing page v0 live on the skeleton (headline,
           3 bullets, email capture). Friday review.
  OUTPUT: problem hypothesis one-pager + 20 names + 5 bookings
 
WEEK 2 - PHASE 1 PROBLEM VALIDATION
  Daily: 2-3 interviews; synthesise same day; send more outreach
  Mid-week: smoke test running (landing page, community posts)
  Thu: ask for commitment (pilot, pre-order, LOI, intro)
  Fri: synthesis; check against pre-set thresholds
  OUTPUT: validated problem brief  (or kill -> back to week 1)
  8-WEEK VARIANT: add a second validation week here
 
WEEK 3 - PHASE 2 SOLUTION DESIGN
  Mon: one-page spec; cut scope to one user, one job
  Tue: user flow + low-fi screens
  Wed: clickable prototype
  Thu: 5 prototype tests with interviewees from week 2
  Fri: fix spec; "not now" list; build plan by day
  Keep going: 1 conversation/day; start the launch list
  OUTPUT: one-page spec + prototype + not-now list
 
WEEKS 4-5 - PHASE 3 BUILD THE MVP
  Week 4: core flow end to end, ugly, in production
          (data model, the one job, minimal auth)
  Week 5: onboarding, activation event tracked, payments
          if you tested price, error states on the happy path
  Every day: deploy; 1 user conversation; show progress
  Thu wk 5: FEATURE FREEZE
  8-WEEK VARIANT: add a third build week (cap is 6 total)
  OUTPUT: deployed MVP + analytics + launch list (20-100)
 
WEEK 6 - PHASE 4 LAUNCH TO FIRST USERS
  Mon-Tue: personally onboard the first 5-10 users
  Wed: fix what blocked them; invite the next batch
  Thu: public posts to the channels in the problem brief
  Fri: launch results one-pager; next hypothesis
  OUTPUT: launch results -> iteration backlog
 
WEEK 7+ - PHASE 5 ITERATE (weekly cycles, see rhythm below)

Weekly operating rhythm

The rhythm is what keeps the phases moving once the plan meets reality. It applies from week 1 onwards.

WhenWhatTimeOutput
Mondayplan: re-read the current phase's exit criteria; pick 1 goal and at most 3 outcomes for the week30 minwritten weekly goal
Dailyone focused build (or research) block; ship something to production or to a user3–5 ha deploy, a sent message or a finished interview
Dailyone user conversation (interview, onboarding call, support reply, demo)30 minnotes in the synthesis doc
Dailyupdate the "not now" list instead of acting on new ideas5 minshorter scope
Fridaydemo to someone outside your head (co-founder, peer, user)30 minfeedback
Fridayreview: metrics, exit criteria, kill/continue rules; write 5 lines on what you learned30 mindecision: continue, cut scope, change phase, or kill

Metrics to track each week, by phase:

PhaseLeading metric (effort)Lagging metric (evidence)
0–1outreach sent; interviews heldproblem mentioned unprompted; commitments obtained
2prototype sessions runusers who completed the core task unaided
3deploys per weekreal users through the core flow in production
4people personally invitedsign-ups → activated → returned in week 2
5experiments shippedretention by weekly cohort; revenue; PMF survey

Paul Graham's "Startup = Growth" (2012) gives a YC reference point for phase 5: "A good growth rate during YC is 5-7% a week." That describes funded startups aiming for venture scale; treat it as a yardstick, not a pass mark.

Decision rules

Write your rules before each phase starts, in the one-pager, with numbers. The thresholds below are heuristics, not laws: they are starting points for a solo B2B or prosumer product and should be adjusted for your market (a niche enterprise product might need 3 committed customers, a consumer app hundreds of sign-ups). What matters is that you set them in advance and honor them.

PhaseContinue if…Go back or kill if…
0one idea scores clearly ahead and you can name 20 reachable people with the problemyou can't name 10 people who have the problem → pick another idea or another segment
1at least half of 10–15 interviewees describe the problem unprompted and have tried to solve it; at least 3 give a real commitment (time, reputation or money)fewer than about a third describe it unprompted → back to ideation; they describe it but nobody has tried to solve it → problem too mild, kill or re-segment
1smoke test beats the threshold you wrote down before launching itit misses by a wide margin after reaching enough of the target segment → rewrite the pitch once; if it misses again, re-segment or kill
2most prototype testers (say 4 of 5) complete the core task and ask when they can use ittesters complete it but shrug → the outcome is wrong, not the UI; back to the brief
3the core flow works in production by the appetite deadlineat the hard cap (6 weeks) → cut to what works and launch anyway; never extend twice
4a meaningful share of activated users come back in week 2 without being chasedalmost nobody returns → talk to every one of them before writing code
5retention by cohort flattens; PMF survey approaches Sean Ellis's 40% "very disappointed"8–12 weekly cycles with no retention improvement → pivot or kill (see launch and iterate)

Startup canvas

A one-page summary you keep open throughout. It is a cut-down version of the Lean Canvas that Ash Maurya adapted from Alexander Osterwalder's Business Model Canvas in 2010 (see Maurya's Running Lean). Fill it in during ideation, correct it with evidence during validation, and re-read it every Friday.

STARTUP CANVAS                          Version: __  Date: ____
---------------------------------------------------------------
1. PROBLEM (top 1-3, in the customer's words)
   -
2. CUSTOMER (one segment; specific enough to find on LinkedIn
   or in a named community)
   - Who:
   - Early adopter trait (already hacking a workaround):
3. CURRENT ALTERNATIVE (what they do today, and what it costs
   them in time / money / risk)
   -
4. UNIQUE VALUE PROPOSITION (one sentence: outcome, for whom,
   unlike what)
   -
5. RISKIEST ASSUMPTION (the one that kills the idea if false)
   - Assumption:
   - Cheapest test:
6. MVP (type + the one job it does)
   - Type: landing page / concierge / WoZ / single feature / ...
   - Core job:
7. SUCCESS METRIC (one number, with a threshold set now)
   - Metric:            Threshold:           By (date):
8. DEADLINE (public launch date and who you told)
   - Date:              Told:
---------------------------------------------------------------
KILL CRITERIA: I will stop or pivot if ________________________
NOT NOW: ______________________________________________________

Phase gate checklists

Tick every box before moving on. If a box can't be ticked by the end of the timebox, apply the decision rules above.

GATE 0 -> 1  (end of ideation)
[ ] One idea chosen; the others parked with a note on why
[ ] Problem hypothesis one-pager written (ideation template)
[ ] Riskiest assumptions ranked (top 3 named)
[ ] Kill/continue criteria for validation written, with numbers
[ ] 20 named people with the problem; >= 5 interviews booked
[ ] Walking skeleton deployed; landing page v0 live
 
GATE 1 -> 2  (end of validation)
[ ] 10-20 interviews held, notes synthesised
[ ] Problem described unprompted by the pre-set share
[ ] Commitment obtained (time, reputation or money) from >= N
[ ] Current workaround and its cost documented
[ ] Reachable channel identified and tested once
[ ] Validated problem brief written with quotes
 
GATE 2 -> 3  (end of design)
[ ] One-page spec: one user type, one job, one happy path
[ ] Clickable prototype tested with >= 5 target users
[ ] "Not now" list written; appetite (2-4 weeks) fixed
[ ] Activation event and success metric defined
[ ] Build plan by day for week 1 of the build
 
GATE 3 -> 4  (end of build)
[ ] Core job works end to end in production
[ ] A real user completed it without help
[ ] Analytics: sign-up, activation, return events firing
[ ] Payments work (if price is part of the test)
[ ] Launch list of 20-100 named people; posts drafted
[ ] Feature freeze held for >= 2 days
 
GATE 4 -> 5  (end of launch week)
[ ] First cohort personally onboarded
[ ] Launch results one-pager written
[ ] Top 3 problems from users ranked into the backlog
[ ] Weekly metrics dashboard exists (even a spreadsheet)

Anti-patterns

Named so you can catch yourself doing them.

Anti-patternWhat it looks likeWhat it's really avoidingCounter
Endless ideationa Notion database of 80 ideas, none testedcommitment and possible rejection1-week timebox; scoring matrix; pick one
Building in secret"stealth mode" for a to-do appjudgmentshow a user something every week; build in public
Polishing before usersanimations, dark mode, custom design system in week 2finding out the core is unwantednothing cosmetic until 10 users used the core flow
Tool-shoppinga week comparing hosting, ORMs, auth providers, component librariesthe actual problemuse the stack you know; see building the MVP
"One more feature"launch slips weekly as features are addedlaunch-day exposurefeature freeze; "not now" list; public date
Premature scalingmicroservices, queues and Kubernetes for zero usersthe market question, by answering an engineering oneone app, one database, one region
Survey-as-validation"83% of respondents said they'd use it"talking to peopleinterviews about past behavior; commitments
Friends-and-family validationmum, flatmates and ex-colleagues love itstrangersonly count people who match the segment and don't know you
Tutorial hell"learning" the stack you'll need "properly" firstbuildinglearn just in time, for the next deliverable
The big-bang launcheverything rides on one Product Hunt daythe slow grind of onboarding users one by onelaunch to 10 people you know first; public launch later
Pivot-hoppingnew direction every week after one bad conversationsitting with ambiguous dataonly change course at a Friday review, against written rules
Brand before productlogo, name and domain agonised over for daysthe productplaceholder name; 30-minute timebox
Fundraising before evidencepitch deck before a single userbuilding and sellingusers and revenue are the pitch; see y-combinator

Stuck? Do this now

Pick the row that matches, do the action in the next 30 minutes, then return to the phase plan.

If you are…Do this now (≤ 30 min)
unable to choose between ideasscore the top 3 on the ideation matrix; take the winner; set a 1-week validation timebox
without any ideawrite down 10 things that annoyed you at work this month; for each, who else has it
afraid to contact strangerssend 5 outreach messages from the validation templates, now, before you think about it
getting polite, useless interviewsre-read the Mom Test rules; replace every "would you" question with "tell me about the last time"
unsure whether validation "passed"open the kill/continue criteria you wrote; if you didn't write any, write them now and apply them honestly
staring at a 14-feature speccross out everything not needed for one user to complete the core job once; move it to "not now"
stuck choosing a toolpick the one you have shipped with before; if none, the most boring mainstream option; close the tabs
three weeks into the build with nothing deployeddeploy what exists today, even if it's broken; show it to one user tomorrow
polishingask "would a user notice this in their first session?"; if not, stop
scared to launchmessage 5 people from the launch list personally with a link; that is the launch
demoralised after a flat launchcall three users who signed up and didn't return; ask what they were trying to do
drifting with no deadlineput a launch date in writing and send it to three people whose opinion you care about

References