143 SaaS idea sources, from 1-star reviews to job ads

This is the whole list I work from, in the order I work through it. None of it
is a framework. Every line is a place where somebody has already left evidence
that they have a problem, and the only thing separating a strong line from a
weak one is whether leaving that evidence cost them anything.

Item 143 is the one that outranks the other 142, so read it first.

Reviews and ratings

1. Chrome Web Store reviews. Pull the 1 to 3 star reviews on popular extensions and count the requests that repeat.
2. App Store reviews after a price rise. The week a price goes up, people say out loud exactly what they will stop paying for.
3. G2 against Capterra. Same product, two audiences. The negative reviews differ by segment, and that difference is the brief.
4. Trustpilot negatives, clustered. SaaS, hosting, fintech and B2B each repeat a handful of complaints. Sort them into buckets and count.

Where a product admits its own gaps

5. Cancellation reason screens. The dropdown a product shows you on the way out is its own list of unsolved problems, written by the vendor.
6. Public changelogs. Read them for what never ships, not for what did.
7. Roadmap items parked on "later". A feature that has sat there two years is a product somebody else is allowed to build.
8. Pricing pages. Find the add-on everyone resents paying for separately, and sell that as the whole product.
9. Pricing calculators. If a vendor needs a calculator to explain its price, the category has a simplicity gap.

Automation platforms and spreadsheets

10. Zapier and Make integration gaps. The two services people keep asking to connect and cannot.
11. Zapier template usage. When thousands of people build the same 10 to 20 step automation, that is a product with its spec already written.
12. Google Sheets templates. A complicated shared spreadsheet is an application nobody has built yet.
13. Excel forums. Search "how do I automate" and read only what repeats.

Developer surfaces

14. Stack Overflow repeats. The same technical problem asked thousands of times is a developer tool.
15. GitHub Issues. Feature request, enhancement, TODO and help wanted labels on projects people actually depend on.
16. GitHub Discussions. What people keep asking how to do that the project simply does not do.
17. npm and PyPI. Heavily used, badly documented packages become an easier hosted API.
18. API marketplace gaps. Anything people currently collect by hand on a schedule.
19. API shutdowns. When a public API closes, the demand behind it does not close with it.

Moving data

20. Migration complaints. "I want to move from X to Y but". Migration tooling is a category of its own and almost nobody serves it.
21. CSV import and export pain. Wherever moving data between two products is hard, there is a product sitting in the gap.

Search operators

22. "I hate [tool]". List the products people say it about, unprompted.
23. "alternative to X". High volume with weak alternatives is the single strongest query in this list.
24. "X sucks". Tells you why they are looking, not merely that they are.
25. "X too expensive". Finds the price-sensitive segment, by name.
26. "X complicated". Finds the enterprise product that is waiting for a simple version.
27. "X does not support". Finds the missing integration, platform, country or file format.
28. Autocomplete mining. Sweep [product] alternative, [product] pricing, [product] problem, [product] does not, [product] for [niche].
29. "Best X for Y" results pages. Compare the last-updated dates, review counts and prices of whatever ranks. A stale first page is an open category.
30. Programmatic page gaps. Thousands of searchable combinations with no decent landing page behind any of them.
31. Your own Search Console. High impressions and low clicks is a product request arriving in the shape of a query.

Video and courses

32. YouTube comments under tutorials. "Is there really nothing that does this automatically".
33. "How I do X" videos. A 30 minute manual workflow is a 30 second product.
34. Udemy reviews. Where students say a section was hard, or that they wish a tool existed for it.
35. Amazon business product reviews. Operational problems around physical goods are frequently solvable in software.

Marketplaces and template economies

36. Shopify App Store reviews. What merchants keep asking for that the existing apps do badly.
37. WordPress plugin reviews. Old, insecure or unpleasant plugins have a modern hosted equivalent.
38. Webflow and Framer marketplaces. A heavily reused component is a feature waiting to become a product.
39. Figma Community. The workflow behind a popular UI kit is the product. The kit is not.
40. Canva template trends. Content types people produce over and over can be generated.
41. Notion template sales. A best-selling but complicated template converts directly into software.
42. Airtable and Sheets walkthrough videos. These are internal mini-applications companies built themselves because nothing existed.

Paid human work

43. Freelance marketplace task mining. Find the freelancers doing the identical job every week.
44. Bid counts. A job posted thousands of times is a job worth automating.
45. The 5 to 50 dollar task. Small amounts paid repeatedly are the clearest willingness to pay you will ever find.
46. Virtual assistant task lists. Whatever a VA does on a schedule is a specification.

Vertical workflows

47. Recruiters. CV parsing, candidate screening, interview scheduling.
48. Accountants. Anywhere Excel, email and PDF meet, there is a product.
49. Estate agents. Listings, lead follow-up, valuations, document preparation.
50. Agency client reports. Rebuilt by hand every single month.
51. Agency standard operating procedures. A written repeated process is a product with the spec attached.
52. Consultants. Work currently done across a spreadsheet, a notes app and a text document.
53. Healthcare administration. Appointments, forms, documents, communication. Administration only, never clinical decisions.
54. Construction. Quotes, site reports, photographs, checklists, materials tracking.
55. Restaurant back office. Menus, stock, suppliers, reservations, staff scheduling.

Complaints in public records

56. Local business reviews. "I wish they" and "they are terrible at" describe an industry, not one shop.
57. Maps listings by category. Read hundreds of businesses in one trade and the shared operational problem surfaces on its own.
58. Complaint bureaus. Recurring process failures, already sorted by sector.
59. Government complaint and open data portals. Official complaints arrive categorised for you.
60. Regulation changes. New compliance duties create work that did not exist last quarter and has no incumbent.

New capability

61. New APIs and SDKs. A new primitive makes a vertical product possible roughly a year before anybody builds it.
62. AI wrapper gaps. Everybody has the general tool. Nobody has the one shaped like one specific job.
63. Work that is impossible without AI, rather than merely nicer with it. Extraction, classification, monitoring, automation. Not a chat box.

The shape of a workflow

64. Before and after. One person using seven tools to finish one job.
65. Tool-stack screenshots. People post their stacks constantly. Five tools doing one job is one product.
66. Browser tab mining. List the sites a how-to video keeps switching between, then delete the switching.
67. Email to spreadsheet to chat. The most repeated data-moving process inside any company.
68. PDF to Excel. Relentless in B2B and still overwhelmingly manual.
69. Screenshot to data. Wherever people retype what they can already see.
70. Email attachments. Invoices, orders and CSVs arriving in an inbox and processed by hand.
71. Recurring reports. Anything built every Monday, or on the first of the month.
72. Copy and paste between two systems. That is an integration nobody has written yet.
73. Login-heavy work. Five to ten separate panels to complete one task means an aggregator.

Sentences that give it away

74. "We currently use a spreadsheet".
75. "Our process is". People describe their entire workflow unprompted, for free.
76. "Looking for a tool". Keep every one of these. It is an interview list, not a search result.
77. "Does anyone know software that". That is purchase intent, written down and dated.
78. "We built this internally". Internal tools are unlaunched vertical products.
79. "We use a combination of X and Y and Z". Combine the three.

Dead and sold products

80. Reviews after an acquisition. New owners change prices and remove features, and users say so within days.
81. Shut-down products. A closure does not prove the problem was fake. Founders leave for their own reasons.
82. Abandoned indie projects. Separate a failed idea from failed distribution before drawing any conclusion.
83. "Reason for selling" on marketplace listings. The seller explains the weakness for you.
84. Valuation multiples by category. A low multiple marks a category with a problem worth understanding.
85. App Store "What is New" dates. High downloads, no update in years.
86. Rating against downloads. Heavily installed and poorly rated is the strongest replacement signal there is.

Reading a competitor properly

87. Cluster their 1-star reviews. Sort ten thousand of them into pricing, bugs, UX, missing features, ads, sync.
88. Turn each complaint into its own landing page and let search tell you which one is real.
89. Support documentation volume. A vast help centre is a product that got too complicated.
90. FAQ density. The most-explained topic is exactly where users get stuck.
91. Onboarding videos. Thirty minutes of onboarding is an opening for a three minute product.
92. Demo videos. Watch the steps the user is made to perform, not the features being shown.

Two framings

93. Spreadsheet, then SaaS, then AI. Find the manual solution, study the product that replaced it, then build the version that is ten times simpler.
94. Niche times existing tool. Not CRM. CRM for tattoo artists, for wedding photographers, for recruiters.

Gaps by geography and plumbing

95. Country gaps. Popular in one market, absent in another, and it runs in both directions.
96. Language gaps. An English-only interface in a market that does not work in English.
97. Payment gaps. The SaaS a small business wants and cannot pay for from its own country.
98. Integration gaps. The local or niche service a large product will never connect to.

Workarounds

99. "[product] workaround" and "[product] hack". People documenting how they got past a limit.
100. Browser extensions that patch a missing feature. Each one is a feature the vendor declined to build.
101. Community-written bots. Somebody already built it for free and maintains it badly.

Open source with a business attached

102. Many stars, poor experience. Host it properly.
103. Many stars, difficult installation. Docker plus a CLI plus a config file is a wall in front of every non-developer.
104. Heavy usage, no monetisation.
105. Community-built spreadsheets with thousands of upvotes.
106. "I made a calculator" posts. A viral calculator is version one of something larger.
107. Popular free checkers and generators. The product is the workflow around them, not the tool itself.

Combining signals

108. Keyword first, community second. Find the search volume, then confirm actual people are actually stuck.
109. Traffic, weak competition and a bad incumbent. All three at once is the strongest combination in this list.
110. Constant tutorial production means constant manual work.
111. Viral "here is how I do this" clips. Turn the trick into one button.
112. Newsletter sponsors. Whoever keeps buying the same list has found a segment that pays.

Hiring as a signal

113. Job ads naming a specific tool. That tool is solving something expensive.
114. Job ads saying "manually maintain", "weekly reporting", "data entry", "reconcile", "compile reports".

Who is complaining

115. Employees complaining about operations. Sell the fix to the employer.
116. Founders saying "I still do this in a spreadsheet".
117. Agencies running six to ten tools. One vertical product covering the shared workflow.

Attacking one part of a big product

118. One-feature competitor. Take the single most-loved feature and do only that, ten times better.
119. Removed features. Find the users the incumbent abandoned on purpose.
120. Tier arbitrage. The one enterprise-only feature that small companies also want.
121. Usage limits. 1,000 records, 10 exports, 5 seats. Every limit has a segment stacked up behind it.
122. Export lock-in. Sell the way out.
123. Account sharing. When teams share one login because seats are expensive, the pricing model is the opportunity.
124. White label. Agencies resell the same product repeatedly and want their own name on it.
125. Client portals. Files, tasks, approvals and messages currently spread across four tools.

Processes that lose money

126. Approval loops. Approve, reject, revise exists in every industry on earth.
127. Compliance evidence. Screenshots, PDFs and logs collected by hand for audits.
128. Renewals. Licences, certificates, contracts, insurance, maintenance.
129. Deadline penalties. Being late costs money, and money already lost is the strongest willingness-to-pay signal available.
130. Revenue leakage. Subscriptions, invoices, refunds, usage and billing quietly losing money nobody has counted.
131. Unused software licences.
132. Duplicate software. Three or four tools doing one job inside one company.
133. Human in the loop. AI does 90 percent, a person checks the last 10, and nobody has built the checking part.
134. "We have someone doing this manually". One employee assigned to one task is the clearest automation candidate that exists.

Sizing the problem

135. The 1,000 dollar a month problem. Paying a lot and unhappy about it.
136. The 50 dollar a month problem. Too small for the incumbents to bother with.
137. The 10 minute problem. Ten minutes a day is about forty hours a year.

Cadence

138. Every Monday.
139. Every month. Reports, invoices, reconciliation, analytics.
140. Before every meeting.
141. After every sale. Onboarding, CRM, invoice, email.
142. After every customer leaves. Cancellation, feedback, data export, retention.

The one that outranks the other 142

143. Do not ask people for ideas. Find the workaround they are already paying for. Somebody running a spreadsheet plus an automation platform plus a virtual assistant plus a chatbot plus a freelancer plus five subscriptions to finish one job is a far stronger signal than any number of people telling you an idea sounds useful.

Roughly a third of these produce nothing in a given week. The five that have
never wasted my time are 23, 45, 86, 114 and 143, and they share one property:
the person left the evidence while spending money or hours, not while answering
a question.