brainstorming-saas-ideas

Use when the user wants new SaaS, web app, Android app, iOS app, browser extension or micro-SaaS ideas, a product to build next, a unique startup idea, a way to cash in on a trend or platform shift, a

SKILL.md

GitHub

Brainstorming SaaS & App Ideas

Research real pain and real money, generate many ideas privately (≥ 25), push them through an originality engine so they are not the same ideas everyone else gets, cut hard, red-team the survivors, and present the top 3 as idea cards a builder can act on this week.

Core principles

  1. Unique, but anchored in proven pain. A novel idea nobody needs is worthless; a needed idea everybody already built is a race to the bottom. The target is the overlap: a painful, paid-for problem, attacked from an angle almost nobody else is taking.
  2. Evidence over opinion. Every "people want this" claim points to a source. Unverified claims are labeled [assumption].
  3. Money proof beats interest proof. Someone already paying (for a tool, a freelancer, a spreadsheet consultant, their own time) outranks someone saying "cool idea".
  4. Distribution is part of the idea. An idea without a named path to the first 100 users is not finished.
  5. Present 3, not 30. The rest stay private.

This skill is self-contained: the research method, originality engine, hard gates, scoring rubric, idea card template and validation playbook are all below.


Step 0 — Founder profile (get this first)

If the user hasn't stated their constraints, ask one bundled question (never a questionnaire spread over many turns):

"To make the ideas fit you: (1) your skills/stack and team size, (2) budget and hours per week, (3) target platforms (SaaS/web, Android, iOS, extension), (4) B2B, B2C or either, (5) any industry you know from the inside or audience you already have, (6) goal: side income, full-time indie income, or venture-scale, (7) which country you're in (affects payments and app stores)?"

If the user says "just give me ideas", assume: solo developer, small budget (< $500), ~15 h/week, web-first, goal = indie income, and state these assumptions at the top.

Record the profile as a short block and reuse it for the whole conversation:

Founder profile: skills · team · budget · hours/week · platforms · B2B/B2C · insider domains · audience/distribution assets · goal · country/payment rails

Unfair advantages matter most. Insider knowledge of an industry, an existing audience, access to a community, a language/market others ignore. Ideas built on an unfair advantage are automatically more unique, because other people cannot reproduce them.

Payment rails check. If the founder's country lacks direct Stripe support, plan around a merchant of record (e.g. Paddle, Lemon Squeezy) or app-store billing, and note it on the card.


Step 1 — Live market research

If web search/browsing is available, use it: do at least 8–12 searches across different source types below before generating. If no live tools are available, say so, work from knowledge, and mark every market claim [unverified — check before building].

Signal type Where to look What you're extracting
Pain (raw complaints) Reddit (niche subs, "is there a tool that…", "I hate…", "how do you manage…"), Hacker News "Ask HN", Indie Hackers, X/Threads, niche forums, Facebook groups, Discord communities Exact quotes, frequency, who is complaining, the trigger moment
Review mining 1–3 star reviews on G2, Capterra, Trustpilot, Google Play, App Store, Chrome Web Store, Shopify App Store for the top 3 products in a category Recurring complaints = the wedge. "Too expensive", "too complex", "missing X", "support is dead", "removed feature Y"
Money proof Upwork/Fiverr gigs bought repeatedly, job posts for a manual task, paid templates/Notion/Excel sold on Gumroad, consultants selling the same service Service → software opportunities; proven willingness to pay and a price anchor
Demand Google Trends, keyword tools (if available), "X alternative" / "X vs Y" searches, YouTube tutorial view counts for a workaround Rising vs flat vs dying demand
Shift signals ("why now") New APIs/model capabilities, new OS features (iOS/Android releases), platform policy changes, new regulations and deadlines, incumbent price hikes, shutdowns, acquisitions, sunset announcements A window that just opened and incumbents haven't reacted to
Launch & revenue data Product Hunt, app store top charts by category, Acquire.com / Flippa listings with revenue, indie founders' public revenue posts What small teams actually make money with, and at what price

Evidence log (keep privately, cite on cards): claim — source — date. "Why now" evidence should be from the last 90 days; pain evidence can be older if it's still recurring.

Seeds from the user (optional): an industry, a problem they have, a product they hate, an unfinished project. If the user brings their own idea, run Remix mode (see end of file).


Step 2 — Generate privately (≥ 25 raw ideas)

Write each raw idea as one line in this exact form, which forces it to be concrete:

[Who] struggles with [pain] when [trigger moment]. Today they [workaround]. We [solution]. They pay because [money reason].

If you can't fill every slot, the idea is too vague — sharpen or drop it.

2a. Generation lenses (use at least 8 of these 14)

# Lens How to apply
1 Complaint → product Turn a recurring complaint from Step 1 directly into a tool
2 Unbundle the incumbent Take one feature of a bloated, expensive tool and do it 10× better for one niche
3 Service → software Productize a freelance gig people pay for repeatedly
4 Kill the spreadsheet / WhatsApp group Find a workflow people run in Excel, Google Sheets, Notion, WhatsApp, paper or email, and replace it with a purpose-built tool
5 Verticalize a horizontal tool Generic tool → specific industry (scheduling for physiotherapists, CRM for wedding photographers)
6 Platform shift A capability that just became possible or cheap: new AI model abilities, on-device AI, new iOS/Android APIs (widgets, Live Activities, App Intents, Health/fitness data, NFC), new browser APIs
7 Regulation / deadline A new law, compliance requirement, tax rule or reporting deadline that forces a group to change behavior
8 Borrowed distribution Build inside an ecosystem with built-in discovery: Chrome extension, Shopify app, WordPress plugin, Figma plugin, Slack/Teams app, Notion integration, Zapier/Make app, Telegram/WhatsApp bot
9 Geo / language arbitrage Proven product category that is missing, overpriced or badly localized in a specific country, language or currency
10 Picks & shovels Tools for the people riding a rising trend (creators, AI builders, sellers on a new marketplace)
11 Incumbent neglect A product that raised prices, got acquired, shut down a free tier, or ignores a segment — offer the lifeboat
12 Mobile-native moment Something only a phone does well: camera, location, notifications at the right moment, offline use, quick capture, sensors
13 Boring but paid Unsexy industries (logistics, trades, accounting, clinics, property management, agriculture, small manufacturing) where software is outdated and buyers have budgets
14 Second-order effect Not the trend itself, but the problem the trend creates (e.g. more AI-generated content → need for verification, provenance, dedup, review tools)

2b. Required mix in the raw list

  • ≥ 3 B2B ideas (businesses pay more and churn less)
  • ≥ 2 mobile-native ideas (Android and/or iOS where the phone is essential, not just a wrapper)
  • ≥ 2 borrowed-distribution ideas (extension, plugin, marketplace app)
  • ≥ 2 ideas built on the founder's unfair advantage (insider domain, audience, language, location)
  • ≥ 2 "boring but paid" ideas
  • ≥ 2 "twist on a current riser" ideas (a rising product/category from Step 1 + a clear differentiation)
  • ≥ 1 "smallest possible" idea: shippable and chargeable in ~7 days, mainly to get the founder to first revenue fast
  • ≥ 1 "bold bet": harder, but with a much bigger ceiling

Step 3 — Originality engine (make the ideas unique)

Run every raw idea through this step. The goal is that if five other people asked an AI for SaaS ideas, none of them would get these.

3a. Cliché filter — banned unless there is a sharp, specific twist

These are the ideas everyone generates first. They are not allowed in the final 3 unless the card explains a wedge that a skeptic would accept in one sentence:

  • Generic AI chatbot / "ChatGPT for X" with no proprietary data, workflow or distribution
  • AI writing assistant, AI resume/cover-letter builder, AI caption generator
  • To-do list, habit tracker, notes app, journaling app, Pomodoro timer
  • Expense tracker, budget app, generic personal finance dashboard
  • Social media scheduler, link-in-bio page, generic URL shortener
  • Recipe app, meal planner, generic fitness/workout tracker
  • Generic CRM, generic project management tool, generic invoice generator
  • Flashcards / generic language learning app
  • "Uber for X", "Tinder for X", "Airbnb for X" (two-sided marketplaces without a cold-start plan)

Test for a cliché: if the idea can be described in three words that also describe 100+ existing apps, it's a cliché.

3b. Uniqueness techniques (apply at least 4 to the strongest raw ideas)

Technique How it works Example of the move
Collision Combine two distant domains: a mechanism from field A applied to pain in field B "Air-traffic-control style board" × "home-renovation contractors juggling 6 sites"
Inversion Flip the default assumption of the category (remove the main feature, reverse who pays, make it slower/offline/private on purpose) A "no-dashboard" analytics tool that only sends one sentence a week
Extreme niche Narrow the customer until it sounds almost too small, then check it can still pay Not "gyms", but "climbing gyms managing route setting and wall rotation"
Trigger-moment design Build around one precise moment in someone's day/week/year, not a general category The 10 minutes after a client meeting when a consultant must log notes, tasks and billable time
Insider workflow Use a step only practitioners know about (ask: "what do people in this job do that outsiders never see?") The re-check a pharmacist does before handing over a prescription
Output, not tool Sell the finished result instead of software to produce it ("done-for-you" automated) Not "SEO tool" but "a weekly ready-to-publish fix list, applied automatically"
Constraint forcing Force the idea to work under a hard limit: offline-only, one-screen, voice-only, works on a 2G connection, no login, under 30 seconds A field-inspection app that works fully offline and syncs later
Time travel Ask "what will be obvious in 3 years?" and build the first decent version of that today Tools for managing fleets of AI agents inside small businesses
Ugly data Find valuable data trapped in PDFs, photos, screenshots, voice notes, faxes, emails — and structure it Turning photographed handwritten delivery slips into a searchable ledger
Local reality Design for conditions mainstream products ignore: local payment methods, mixed languages, power/network cuts, cash economies, shared devices Shop inventory app that tracks customers buying on credit ("on tab") and works on low-end Android
Anti-incumbent Take the incumbent's #1 complaint from reviews and make its opposite your entire identity "The project tool with zero setup — it builds itself from your email"
Community-born Build for a community the founder is already inside, where they can observe pain daily A tool for a hobby/profession forum the founder actively posts in

3c. Uniqueness check (for each candidate that survives)

  1. Search for clones. Search "[core idea] app", "[core idea] tool", "[category] for [niche]", "[incumbent] alternative". Record the 2–3 closest competitors.
  2. Classify:
    • Blue — no direct competitor, but adjacent proof of pain/money exists → great, verify demand is real
    • Wedge — competitors exist, but there is a specific, defensible difference (niche, platform, price, workflow, locale) → good
    • Clone — competitors already do the same for the same people → cut, or reshape with a technique from 3b
    • Empty — no competitors and no evidence anyone has the pain → probably no market; cut unless strong evidence emerges
  3. The five-people test: "Would five other smart people, given the same generic prompt, land on this idea?" If yes, push it further with another technique or cut it.
  4. "Why hasn't this been done?" Give an honest answer. Good answers: "too small for VC-funded companies", "needs insider knowledge", "the API only launched recently", "the market speaks a language big players ignore". Bad answer: "nobody thought of it" (almost never true).

Step 4 — Platform fit

Pick the platform the problem demands, not the one that's trendy. Note the choice and reason on the card.

Platform Best when Monetization norm Watch out for
SaaS (web, B2B) Desktop work, teams, integrations, admin dashboards Monthly/annual subscription, seat or usage pricing Longer sales cycle; needs onboarding and trust (security page, support)
Web app / PWA (B2C or prosumer) Fast to ship, SEO-driven discovery, no store review Freemium, one-time purchase, credits Harder retention without push notifications; payments via merchant of record
Android Large global/emerging-market audience, low-end devices, local business tools Ads + subscription/IAP; lower ARPU, higher volume Play Console fee is one-time; new personal developer accounts must run a closed test with a minimum number of testers before production (check current Play rules); Google Play billing required for digital goods
iOS Users with higher willingness to pay, health/productivity/prosumer, US/EU/JP markets Subscriptions with free trial; higher ARPU Needs a Mac and a yearly developer fee; App Review rejections; in-app purchase rules for digital goods; store commission (lower for small-business programs)
Browser extension Work happens inside existing sites (Gmail, LinkedIn, Sheets, marketplaces) Freemium + subscription Platform policy changes; store review; depending on host-site DOM
Marketplace plugin (Shopify, WordPress, Figma, Slack…) The buyer already lives in that ecosystem Subscription billed through the platform Revenue share; platform can copy you; tied to one ecosystem

Cross-platform rule: start with one platform for the MVP. Add the second only after retention is proven. Note a cross-platform stack (e.g. React Native / Flutter / PWA) if the founder needs both stores later.


Step 5 — Hard gates (cut immediately, no scoring)

Drop any idea that fails one of these:

  1. IP or ToS violation: uses someone's trademark/brand, clones a specific product, or depends on scraping that violates a site's terms.
  2. Licensed/regulated without a license: medical diagnosis, legal advice, lending, investment advice, insurance, gambling, handling payments/funds custody — unless the founder has the license or the idea clearly stays on the tooling side.
  3. Over budget or skills: needs hardware, large data acquisition, a big team or tech the founder can't learn within the MVP window.
  4. No reachable buyer: can't name a specific channel or community where the first 100 users are.
  5. No money proof anywhere: nobody pays for any solution to this pain (tool, freelancer, template, staff time).
  6. Cold-start trap: two-sided marketplace or social network without a plan to seed one side alone.
  7. Single-platform kill switch: the product only works while one API/platform tolerates it, with no fallback.
  8. Runway mismatch: time to first revenue is longer than the founder can sustain.
  9. Sales mismatch: requires enterprise sales cycles (security reviews, procurement) for a solo, part-time founder.
  10. Unfixable cliché: fails Step 3a with no defensible wedge.
  11. Harmful by design: deceptive, surveillance of people without consent, spam tooling, academic cheating, fake reviews, dark-pattern subscription traps.

Step 6 — Score (11 criteria, 1–5 each)

Criterion 1 = 3 = 5 =
Pain intensity ★ Nice to have Annoying, recurring Costs money/time daily; people actively search for fixes
Willingness to pay ★ Users expect free Some pay for alternatives Clear money proof; buyers have budget; easy to charge $20+/mo (B2B) or $5+/mo (B2C)
Reachability ★ "Social media and ads" One plausible channel Named communities/marketplaces/search terms where buyers gather; founder has access
Originality ★ Cliché / clone Clear wedge vs competitors Blue-ocean angle with adjacent proof of demand; passes the five-people test
Founder fit Unfamiliar domain, no advantage Some familiarity Insider knowledge, existing audience, or unique access
Build simplicity > 3 months to MVP 4–8 weeks ≤ 2 weeks to a chargeable MVP
Competition gap Strong incumbents loved by users Incumbents with visible complaints Incumbents absent, neglectful or overpriced for this niche
Timing (why now) No change in years Gradual tailwind Fresh shift in the last 90 days (API, regulation, price hike, shutdown)
Retention / frequency One-time use Weekly use Daily use or embedded in a business workflow; data lock-in
Unit economics Low price, high CAC, high AI/infra cost Workable High margin, low CAC channel, AI/infra cost < 20% of price
Defensibility Trivial to copy Some switching cost Proprietary data, workflow lock-in, community, integrations, brand in a niche

★ = critical. Any ★ criterion scoring ≤ 2 cuts the idea, regardless of total.

Total = /55. Keep the top candidates by total. Break ties with Willingness to pay, then Reachability, then Build simplicity.

Be harsh. If most ideas score 4–5 everywhere, you're grading too kindly — re-score with the evidence log open.


Step 7 — Red team the finalists (top ~5)

For each, write privately:

  1. Pre-mortem: "It's 12 months later and this failed. The top 3 reasons are…"
  2. Skeptic's one-liner: the strongest objection an experienced founder would make.
  3. Rebuttal or reshape: answer the objection with evidence, or change the idea (narrower niche, different platform, different pricing) until it holds.
  4. Copy risk: "If the biggest incumbent added this feature next quarter, what's left?"

Drop any finalist whose objection you can't answer.


Step 8 — Pick a diverse 3

The final three must differ from each other in at least two of: customer type (B2B/B2C), platform, lens used, risk profile. Aim for this shape:

  • Safe bet — strongest evidence, fastest path to revenue
  • Sharp wedge — most unique angle with solid proof of pain
  • Bold bet — highest ceiling, more risk

Never present three variants of one idea.


Step 9 — Present

  1. Assumptions line (if any profile items were assumed).
  2. Comparison table: idea × 11 scores × total, one row each.
  3. One idea card per idea using the template below.
  4. Recommendation: which one to build first and why (in 2–3 sentences, tied to the founder profile).
  5. One question: which idea to deep-dive / validate / turn into a PRD or build plan — or whether to generate a new batch in a different direction.

If a file system is available, save the set as reports/YYYY-MM-DD-app-ideas.md. Otherwise present it in chat.


Idea card template

### Idea: {working title} — {one-line pitch}

| Field | Content |
|---|---|
| **One-liner** | [Who] struggles with [pain] when [trigger]. We [solution]. |
| **Ideal customer (ICP)** | Specific role/type, company size or life situation, where they are, what they already pay for |
| **Pain + evidence** | 2–3 short paraphrased complaints or data points, each with `(Source, Month YYYY)` |
| **Current workaround** | What they do today (tool, spreadsheet, freelancer, nothing) and what it costs them |
| **Why now** | The fresh shift that opens the window, with `(Source, Month YYYY)`; estimated window length |
| **Originality** | Blue / Wedge classification, uniqueness technique used, answer to "why hasn't this been done?" |
| **Closest competitors** | 2–3 named products + their main weakness (from reviews) |
| **Wedge** | The one thing we do that they don't, provable in a 30-second demo |
| **Core workflow & aha moment** | 3–5 steps from signup to first value; aha moment in < 5 minutes |
| **Platform** | SaaS / web / Android / iOS / extension / plugin — and why this one first |
| **Pricing** | Model (subscription/usage/one-time/freemium), price points, price anchor (what they pay today) |
| **Unit economics (napkin)** | Price · est. infra/AI cost per user · gross margin · likely CAC channel · rough payback |
| **First 100 users** | Specific communities, marketplaces, search keywords, partners, or outreach list — with the first message to post/send |
| **Retention loop** | Why they come back weekly/daily; what data or habit locks them in |
| **MVP scope** | ≤ 5 must-have features, stack suggestion, target build time |
| **Not in MVP** | Features deliberately left out |
| **Validation test (≤ 7 days)** | Test type + exact success threshold before writing serious code |
| **Kill criteria** | The metrics and thresholds that mean "stop or pivot" after launch |
| **Risks** | Top 2, each with a mitigation |
| **Moat over time** | What becomes hard to copy after 6–12 months |
| **Expansion path** | Next niche, next platform, or upsell after product-market fit |
| **Scores** | Pain · WTP · Reach · Originality · Fit · Simplicity · Gap · Timing · Retention · Economics · Defensibility = total/55 |

Validation playbook (use on the card's "Validation test")

Pick the cheapest test that answers the riskiest question. Thresholds below are rules of thumb to start from, not industry laws — adjust to the channel and say so.

Riskiest question Test Rough success signal
Do they have this pain? 10 problem interviews, asking about past behavior, not hypothetical future ("When did this last happen? What did you do? What did it cost?") ≥ 6/10 describe the pain unprompted and already use a workaround
Will they want this solution? Landing page + waitlist, traffic from the "first 100 users" channels Strong waitlist conversion from targeted traffic; people reply to the confirmation email with their use case
Will they pay? Pre-sale / paid pilot / "founding member" price, or a fake-door "Upgrade" button At least a few real payments or signed pilots from strangers
Can we deliver value? Concierge MVP: deliver the result manually behind a simple form Users come back for a second round; manual hours are clearly automatable
Can we reach them cheaply? Post in 3 communities, cold-message 30 ICP prospects, or run a tiny ad test A repeatable channel produces signups at a cost the price can support

Interview rule: never pitch during problem interviews; compliments are not data.


Pricing heuristics

  • B2B: price on value saved, not cost to build. Small-business tools often sit in the tens of dollars per month; workflow-critical tools can charge more. Offer annual billing at a discount.
  • B2C mobile: subscription with a free trial converts best for ongoing value; one-time purchase for utilities; ads only for high-volume, low-intent apps.
  • AI-powered products: cap usage per tier or use credits so heavy users can't make a plan unprofitable. Estimate AI cost per active user on the card.
  • Regional pricing: lower local prices for emerging markets via app-store price tiers or purchasing-power pricing.
  • Don't underprice. A price that's too low attracts the worst customers and kills the budget for acquisition.

Modes

  • Fresh batch (default): full Steps 0–9.
  • Quick mode (user wants something fast or no web tools): skip deep research, generate ≥ 15, still run the cliché filter, gates and scoring; mark evidence [unverified].
  • Remix mode (user brings their own idea): score it honestly with the rubric, run the uniqueness check and red team, then present the original plus 2 sharper variants (narrower niche, different platform, or a technique from 3b) as cards.
  • Direction mode (user says "more like #2" or "only B2B" or "only Android"): keep the profile, regenerate within the constraint, present a new top 3.
  • Deep dive (after picking one): expand that card into a mini plan — 10 interview questions, landing page copy, first-100-users outreach list, MVP feature spec, 4-week build plan, launch checklist.

Honesty rules

  • Never invent statistics, revenue figures, quotes, sources or competitor names. If you can't verify, write [assumption] or [unverified].
  • Paraphrase user complaints; don't fabricate them as quotes.
  • Name real competitors found in research; if none were found, say "no direct competitor found in N searches" rather than claiming none exist.
  • State platform rules, fees and policies as "check current" when unsure — they change often.

Common mistakes

Mistake Fix
Listing 15 ideas Present 3. The rest stay private.
Generic ideas everyone gets ("AI tool for X", habit tracker) Run the cliché filter and the five-people test; apply uniqueness techniques until the idea is specific to a person and a moment
Unique but nobody needs it Every card shows pain evidence and money proof; "Empty" ideas are cut
"Why now" without evidence Cite the shift (API launch, price hike, regulation, shutdown) with source and date
"Market: everyone" Name one ICP specific enough to find in a single community
"We'll market on social media" Name the exact communities, keywords, marketplaces and the first post/message
Building for two platforms at once One platform for MVP; justify the choice
Forgetting AI/infra cost Napkin unit economics on every card
Ignoring the founder Score Founder fit honestly; prefer ideas built on their unfair advantage
Copying a product 1:1 Name what's kept vs changed and the wedge a 30-second demo can show
Validation = "launch and see" A ≤ 7-day test with a success threshold before serious code