What the launch plan rests on
Marketing advice is full of numbers that nobody can trace. Here every statistic in the plan that we have checked is traced to the page it came from, with what that page actually says, how well we read it, and what kind of claim it is.
Browse the wiki for the same claims by topic, with the steps that rely on them.
Standard
A platform rule, a law or documented behaviour.
Evidence
Something measured, with a method stated or a vendor figure we say is a vendor figure.
Practitioner report
What operators report, not independently measured.
Our take
Our own result or judgement. Never presented as a requirement.
Strength is separate from kind. A strong claim has at least one source read in full; a claim resting only on a post, a thread or a vendor's own page is never rated above moderate. Where the plan used to say something the source does not, the claim says what changed. Rejected numbers are kept too, so the same statistic is not re-argued.
How the pages were read: Sources were read by AI research agents using a tool that can summarise a page, and re-read by a second agent. 20 of 42 claims were then re-checked directly against their pages by Claude in a browser (12 confirmed, 6 partly, 2 not confirmed and corrected). No person has independently checked them yet. Check an exact number against its source before you rely on it. Some pages could not be opened at all (Reddit's own wiki blocks automated readers and archives); those are listed as cited, not read.
Record a 60–90 second demo video
Evidence · weak
A 2017 Popsters analysis reported by The Next Web, of products from Product Hunt’s home feed (32,657 in the pool, some filtered out, data collected August to October 2017), found that projects with videos receive more upvotes and rank better. It gives no multiplier.
Article opened by an independent reader, but the data is from 2017 and the multiplier in our plan came from an unsourced vendor blog.
Changed in the plan: Plan text corrected 2026-11-10: the 2.7x had no source; other posts give 2.3x and 4.2x.
True as of 2026-11-10; to be re-checked by 2027-11-10.
One H1 per page, sequential heading hierarchy
Evidence · moderate
AirOps (2025; 900 ChatGPT queries, 12,000+ URLs) found a single H1 on 87% of cited pages against 64% of Google page-one results, and sequential headings on 68.7% of cited URLs against 23.9% of Google top results. It is a correlation.
Dated, sized study with the exact figures, but vendor-run and correlational, and it compares cited pages with Google leaders.
Changed in the plan: Plan text corrected 2026-11-10: the 2.8x had no method behind it.
- Structuring Content for LLMs (read in full)
True as of 2026-11-10; to be re-checked by 2027-11-10.
Get ready to be quoted by answer engines
Evidence · weak
G2’s March 2026 survey of 1,076 B2B software buyers found 51% start software research with an AI chatbot more often than with Google, up from 29% in 2025. It counts where buyers start, not the share of research done there, and reports no click figure.
Self-reported survey run by a review marketplace with a stake in the result; it does not measure the quantity our plan stated.
Changed in the plan: Plan text corrected 2026-11-10: it had said 30-50% of tool research happens in an assistant that never sends a click. No study gives that; this is the closest real figure.
- The Answer Economy (G2 AI search insight report) (read in full)
True as of 2026-11-10; to be re-checked by 2027-05-10.
Publish llms.txt
Standard · strong
Google Search Central says there are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary; the fundamentals of SEO apply. It does not name llms.txt.
Official Google documentation, read by a second reader and re-checked directly.
Changed in the plan: The step already says llms.txt is cheap insurance and not a ranking lever; this is the source for that sentence.
- Google: AI features and your website (read in full)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Add a consent banner that defaults to declining
Handle sales tax and VAT
Standard · strong
The European Commission says electronic services sold to private customers are taxed in the customer’s country, with a EUR 10,000 threshold for EU-established sellers, and that one registration in the One-Stop Shop can cover the VAT due across the EU.
The Commission’s own guidance, read directly; it does not cover non-EU sellers, and the claim says so by scope.
- Cross-border VAT (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
State the refund and cancellation policy
Standard · moderate
EU consumers can cancel an online purchase within 14 days without a reason, except digital content they started downloading or streaming after expressly agreeing to give up that right.
The Commission’s consumer page, but only the relevant section was extracted, not the whole page.
Changed in the plan: Reworded 2026-11-11: the step now states the 14-day right and the express waiver for instant access, and labels the conversion line as our view.
- Returns and the right of withdrawal (partly read)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Own your channels before you need them
Standard · strong
Google requires every sender to Gmail accounts to authenticate with SPF or DKIM, have valid PTR records, use TLS and keep the spam rate in Postmaster Tools below 0.3%. Senders of more than 5,000 messages a day must also have SPF, DKIM and DMARC, From-domain alignment and one-click unsubscribe on marketing mail. Google advises aiming below 0.1% and raising volume slowly, starting with engaged recipients and avoiding spikes.
Google’s own help page, read directly twice (by this batch and by the plays program).
Changed in the plan: A founder sending 25 a day is far below the bulk threshold, so only the all-sender rules bind; setting up DMARC anyway costs nothing and is what Yahoo and Outlook expect of bulk senders. The slow ramp in the send step is what Google itself advises.
- Google Workspace Admin Help: email sender guidelines (read in full)
True as of 2026-11-11; to be re-checked by 2027-02-11.
Standard · strong
Yahoo requires all senders to use SPF or DKIM, and bulk senders to use both, publish a DMARC policy of at least p=none that passes, support one-click unsubscribe, honour unsubscribes within 2 days and keep the spam rate below 0.3%. It does not say what volume makes a sender bulk.
Yahoo’s own sender hub, read directly; the only gap is the undefined bulk threshold, which the claim states.
- Sender Best Practices (read in full)
True as of 2026-11-11; to be re-checked by 2027-02-11.
Standard · strong
Microsoft requires domains sending more than 5,000 emails a day to Outlook.com, Hotmail and Live addresses to pass SPF and DKIM and publish DMARC of at least p=none aligned with one of them; from May 5, 2025 non-compliant mail goes to Junk, with rejection to follow. It also recommends removing invalid addresses regularly to cut bounces and complaints.
Microsoft’s own announcement, read directly; the rejection date is still open on the page.
True as of 2026-11-11; to be re-checked by 2027-02-11.
Wire the transactional email sender
Standard · strong
Google requires every sender to Gmail accounts to authenticate with SPF or DKIM, have valid PTR records, use TLS and keep the spam rate in Postmaster Tools below 0.3%. Senders of more than 5,000 messages a day must also have SPF, DKIM and DMARC, From-domain alignment and one-click unsubscribe on marketing mail. Google advises aiming below 0.1% and raising volume slowly, starting with engaged recipients and avoiding spikes.
Google’s own help page, read directly twice (by this batch and by the plays program).
Changed in the plan: A founder sending 25 a day is far below the bulk threshold, so only the all-sender rules bind; setting up DMARC anyway costs nothing and is what Yahoo and Outlook expect of bulk senders. The slow ramp in the send step is what Google itself advises.
- Google Workspace Admin Help: email sender guidelines (read in full)
True as of 2026-11-11; to be re-checked by 2027-02-11.
List 25 people from your own network
Practitioner report · moderate
YC’s Max Kolysh, compiling answers from dozens of YC founders, says customers one to three almost always came from the founder’s personal network with basically no counterexamples, and that prospecting tools start to matter only at 10 to 20 customers.
Many founders reporting the same pattern, read in full, but collected informally with no count or method.
- How to Get Your First 10 Customers (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
List 25 with a live trigger event
Standard · strong
A US company that sells securities under Regulation D must file a Form D on EDGAR within 15 days after the first sale, which is why a new filing is an early public sign of a raise.
The SEC’s own page, read directly. That filings are public is not stated on that page.
- Filing a Form D Notice (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Standard · strong
Greenhouse’s Job Board API needs no authentication for any GET request, and Lever says every published posting is publicly viewable through its postings API, with a key needed only to submit applications.
Each vendor’s own API documentation, read directly.
Changed in the plan: Ashby, also named by the step, was not checked in this batch.
- Job Board API (read in full)
- lever/postings-api (README) (read in full)
True as of 2026-11-11; to be re-checked by 2027-02-11.
List 25 from the rosters your buyers’ industry already publishes
Standard · strong
The UK ICO says marketing email may go to companies, LLPs and other corporate bodies without consent, as long as you say who you are and give an address to opt out, but sole traders and some partnerships count as individuals and need consent.
The UK regulator’s own guidance, read directly; it notes the guidance is under review.
Changed in the plan: Reworded 2026-11-11: the rosters step now says UK sole traders need consent first, so call or write.
- Electronic mail marketing (Guide to PECR) (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Do the outreach yourself
Practitioner report · moderate
YC’s Max Kolysh, compiling answers from dozens of YC founders, says customers one to three almost always came from the founder’s personal network with basically no counterexamples, and that prospecting tools start to matter only at 10 to 20 customers.
Many founders reporting the same pattern, read in full, but collected informally with no count or method.
- How to Get Your First 10 Customers (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Work the warm list first
Practitioner report · moderate
YC’s Max Kolysh, compiling answers from dozens of YC founders, says customers one to three almost always came from the founder’s personal network with basically no counterexamples, and that prospecting tools start to matter only at 10 to 20 customers.
Many founders reporting the same pattern, read in full, but collected informally with no count or method.
- How to Get Your First 10 Customers (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Ask each for two introductions
Practitioner report · moderate
YC’s Max Kolysh, compiling answers from dozens of YC founders, says customers one to three almost always came from the founder’s personal network with basically no counterexamples, and that prospecting tools start to matter only at 10 to 20 customers.
Many founders reporting the same pattern, read in full, but collected informally with no count or method.
- How to Get Your First 10 Customers (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Run cold outreach that is specific
Standard · strong
The FTC says CAN-SPAM makes no exception for business-to-business email: a commercial message needs accurate headers, a subject that matches the content, disclosure that it is an ad, a valid postal address and a clear opt-out honoured within 10 business days, and each violating email can cost up to $53,088.
The regulator’s own compliance guide to a US federal law, read directly.
Changed in the plan: Reworded 2026-11-11: the sequence step now requires a postal address and an opt-out in every email to US recipients.
- CAN-SPAM Act: A Compliance Guide for Business (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Standard · strong
The UK ICO says marketing email may go to companies, LLPs and other corporate bodies without consent, as long as you say who you are and give an address to opt out, but sole traders and some partnerships count as individuals and need consent.
The UK regulator’s own guidance, read directly; it notes the guidance is under review.
Changed in the plan: Reworded 2026-11-11: the rosters step now says UK sole traders need consent first, so call or write.
- Electronic mail marketing (Guide to PECR) (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Write a four-email sequence
Standard · strong
The FTC says CAN-SPAM makes no exception for business-to-business email: a commercial message needs accurate headers, a subject that matches the content, disclosure that it is an ad, a valid postal address and a clear opt-out honoured within 10 business days, and each violating email can cost up to $53,088.
The regulator’s own compliance guide to a US federal law, read directly.
Changed in the plan: Reworded 2026-11-11: the sequence step now requires a postal address and an opt-out in every email to US recipients.
- CAN-SPAM Act: A Compliance Guide for Business (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Practitioner report · weak
The same YC talk advises keeping outreach under 75 words with one clear call to action, and following up three or four times over a couple of weeks.
One practitioner’s advice drawn from founder stories, read in full, with no data on word count or follow-up.
Changed in the plan: The sequence step now uses this talk’s under 75 words (it said 90 before 2026-11-11).
- How to Get Your First 10 Customers (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Send 25 a day for five days
Standard · strong
Google requires every sender to Gmail accounts to authenticate with SPF or DKIM, have valid PTR records, use TLS and keep the spam rate in Postmaster Tools below 0.3%. Senders of more than 5,000 messages a day must also have SPF, DKIM and DMARC, From-domain alignment and one-click unsubscribe on marketing mail. Google advises aiming below 0.1% and raising volume slowly, starting with engaged recipients and avoiding spikes.
Google’s own help page, read directly twice (by this batch and by the plays program).
Changed in the plan: A founder sending 25 a day is far below the bulk threshold, so only the all-sender rules bind; setting up DMARC anyway costs nothing and is what Yahoo and Outlook expect of bulk senders. The slow ramp in the send step is what Google itself advises.
- Google Workspace Admin Help: email sender guidelines (read in full)
True as of 2026-11-11; to be re-checked by 2027-02-11.
Standard · strong
Yahoo requires all senders to use SPF or DKIM, and bulk senders to use both, publish a DMARC policy of at least p=none that passes, support one-click unsubscribe, honour unsubscribes within 2 days and keep the spam rate below 0.3%. It does not say what volume makes a sender bulk.
Yahoo’s own sender hub, read directly; the only gap is the undefined bulk threshold, which the claim states.
- Sender Best Practices (read in full)
True as of 2026-11-11; to be re-checked by 2027-02-11.
Standard · strong
A Google Workspace user can send 2,000 messages a day (500 on a trial account) to at most 3,000 external recipients a day; past a limit, sending is blocked for up to 24 hours.
Google’s own admin help page, read directly.
Changed in the plan: The cap is not what limits a founder at 25 a day; reputation is. The step’s slow ramp is about a new domain’s standing, which no provider publishes a number for.
- Gmail sending limits in Google Workspace (read in full)
True as of 2026-11-11; to be re-checked by 2027-02-11.
DM 25 on the platform they actually use
Standard · strong
LinkedIn publishes no number for its invitation limit: an invitation restriction typically lasts one week, applies to Premium members too and cannot be bought off, and is most often caused by inviting too many people you do not know or too many too fast. Free members can add a personal note to only five connection requests a month.
Two LinkedIn Help pages, read directly; the absence of a number is what they say.
Changed in the plan: Reworded 2026-11-11: the DM step now says to connect without a note and send the message once they accept.
- Invitation limit reached (read in full)
- Connecting With Other Members - Best Practices (read in full)
True as of 2026-11-11; to be re-checked by 2027-02-11.
Answer every reply the same day, and give the inbox one owner
Evidence · weak
The 2007 InsideSales.com/MIT study of over 15,000 web-form leads at six companies found the odds of reaching a lead by phone fell 100 times, and of qualifying it 21 times, when the first call came at 30 minutes rather than 5. It measured phone calls to inbound leads, not replies to cold email, and did not study closes.
One vendor dataset analysed by an academic, not peer reviewed, from 2007 and about a different situation; the HBR follow-up could not be read past its paywall.
Changed in the plan: Supports the direction of “intent decays by the hour”, not a number for email replies; the steps should not quote 21 times as if it were about email.
- The InsideSales.com/MIT Lead Response Management Study (read in full)
- The Short Life of Online Sales Leads (partly read)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Verify every address before the first send goes out
Standard · strong
Microsoft requires domains sending more than 5,000 emails a day to Outlook.com, Hotmail and Live addresses to pass SPF and DKIM and publish DMARC of at least p=none aligned with one of them; from May 5, 2025 non-compliant mail goes to Junk, with rejection to follow. It also recommends removing invalid addresses regularly to cut bounces and complaints.
Microsoft’s own announcement, read directly; the rejection date is still open on the page.
True as of 2026-11-11; to be re-checked by 2027-02-11.
Answer ten real questions with no link
Standard · moderate
Reddit’s official Reddiquette page says to feel free to post links to your own content within reason, and calls the 9:1 ratio a widely used rule of thumb: only one in ten of your submissions should be your own content.
Official Reddit Help page, read directly in Chrome, but it calls the ratio a rule of thumb, not a rule, and individual communities set their own.
- Reddiquette (read in full)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Talk to every single user
Practitioner report · moderate
YC’s Gustaf Alströmer advises keeping your idea out of a user interview until the end or entirely, asking how they do the task today, what is hardest and what they already do about it, never asking whether they would use your product or which features they want, and recording or taking notes; YC’s earlier version of the talk was based on The Mom Test, though its author later said he rarely uses the method.
Two practitioner sources read in full, advice rather than data; one author has since distanced himself from the method.
- How to talk to users (read in full)
- Startup School - How to talk to users (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Offer every signup a 15-minute call
Practitioner report · moderate
YC’s Gustaf Alströmer advises keeping your idea out of a user interview until the end or entirely, asking how they do the task today, what is hardest and what they already do about it, never asking whether they would use your product or which features they want, and recording or taking notes; YC’s earlier version of the talk was based on The Mom Test, though its author later said he rarely uses the method.
Two practitioner sources read in full, advice rather than data; one author has since distanced himself from the method.
- How to talk to users (read in full)
- Startup School - How to talk to users (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Run ten problem interviews and record them
Practitioner report · moderate
YC’s Gustaf Alströmer advises keeping your idea out of a user interview until the end or entirely, asking how they do the task today, what is hardest and what they already do about it, never asking whether they would use your product or which features they want, and recording or taking notes; YC’s earlier version of the talk was based on The Mom Test, though its author later said he rarely uses the method.
Two practitioner sources read in full, advice rather than data; one author has since distanced himself from the method.
- How to talk to users (read in full)
- Startup School - How to talk to users (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Treat the first-run sequence as a permanent test surface
Evidence · weak
Superwall’s case study says Cal AI ran 123 iOS experiments with 160 paywall designs and 424 variants, for a 31% relative lift in trial-to-paid over 12 months.
A vendor’s own marketing page, a relative lift, no independent check, and Cal AI itself was not confirmed as a source.
Changed in the plan: Plan text corrected 2026-11-10: it had credited the figures to Cal AI.
- Cal AI and Superwall case study (read in full)
True as of 2026-11-10; to be re-checked by 2027-11-10.
Send a payment link within the hour of a yes
Evidence · weak
The 2007 InsideSales.com/MIT study of over 15,000 web-form leads at six companies found the odds of reaching a lead by phone fell 100 times, and of qualifying it 21 times, when the first call came at 30 minutes rather than 5. It measured phone calls to inbound leads, not replies to cold email, and did not study closes.
One vendor dataset analysed by an academic, not peer reviewed, from 2007 and about a different situation; the HBR follow-up could not be read past its paywall.
Changed in the plan: Supports the direction of “intent decays by the hour”, not a number for email replies; the steps should not quote 21 times as if it were about email.
- The InsideSales.com/MIT Lead Response Management Study (read in full)
- The Short Life of Online Sales Leads (partly read)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Chase every open deal three times over two weeks
Practitioner report · weak
The same YC talk advises keeping outreach under 75 words with one clear call to action, and following up three or four times over a couple of weeks.
One practitioner’s advice drawn from founder stories, read in full, with no data on word count or follow-up.
Changed in the plan: The sequence step now uses this talk’s under 75 words (it said 90 before 2026-11-11).
- How to Get Your First 10 Customers (read in full)
True as of 2026-11-11; to be re-checked by 2027-11-11.
Line up 20 users willing to review you later
Evidence · weak
G2’s June 2020 What’s New page says review requests sent through Review Automation see a 40% open rate and an 11% conversion rate, with no method stated.
A vendor’s marketing page from 2020 about automated requests; personal asks may do better but no data was found.
Changed in the plan: Plan text corrected 2026-11-10: it had expected roughly 50% conversion from 20 asks.
- What’s New, G2? June 2020 (read in full)
True as of 2026-11-10; to be re-checked by 2027-05-10.
Create the Capterra listing
Evidence · moderate
G2 announced on 2026-01-29 that it had formally agreed to buy Capterra, Software Advice and GetApp from Gartner. Completion on 2026-02-05 is reported from Gartner’s SEC filing, which could not be re-read directly.
G2’s release was re-read directly; the completion date rests on a filing page that returned almost no text.
Changed in the plan: Plan text corrected 2026-11-10: it had said G2-owned since 2026 as settled fact.
- G2 to acquire Capterra, Software Advice and GetApp from Gartner (read in full)
- Gartner SEC filing: completion of the Digital Markets sale (read in full)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Submit to There’s An AI For That
Standard · weak
There’s An AI For That’s official page lists $437 for a listing with a newsletter mention and $99 a month for a highlight, with $49 as the base listing price; two readings disagreed on exactly which options the $49 covers. It shows no free option.
Read by two earlier readers who disagreed on what the $49 covers, and not reproducible in a direct read, where the prices did not render.
Changed in the plan: Plan text updated 2026-11-10: the 2026-09-02 check read $347 for the newsletter bundle and the page now says $437, so the price has moved; the $49 scope is ambiguous.
- Get featured on There’s An AI For That (read in full)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Submit to Futurepedia
Evidence · weak
Futurepedia’s listing and pricing pages returned 404 on 2026-11-10 and its homepage no longer shows the 700,000+ audience figure an earlier reading recorded, so no price could be confirmed from the site.
Homepage only; the price rests on a third-party review.
Changed in the plan: Plan text corrected 2026-11-10: it had stated $497, a sold-out $247 tier and 400k monthly professionals as fact.
- Futurepedia homepage (partly read)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Submit to Toolify.ai
Evidence · weak
Toolify’s submit page advertises a dofollow backlink and a listing that lasts forever. Third-party reviews dated up to October 2026 report $99 one-time, listed within 48 hours and an advertised minimum of six dofollow links; older pages say $49.
The primary page returned 403, prices differ between sources, and the six-link count is unverified.
Changed in the plan: Plan text corrected 2026-11-10: it had said the six links are contractual; at most they are advertised.
- Toolify submit page (not opened) and third-party reviews (read via a summary)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Submit to Future Tools
Standard · moderate
Future Tools’ submit page says Matt will review the submission and shows no fee, sponsored or featured option. Free and by hand are inferred from that.
Primary page read by two readers; the page does not state free, by hand or a review time.
- Submit a Tool - Future Tools (read in full)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Submit to the Linux Foundation MCP Registry
Standard · strong
The official MCP registry takes a server.json through mcp-publisher. Names under io.github.* need GitHub login; a domain name is proved by a DNS TXT record or a /.well-known/mcp-registry-auth file. The description is capped at 100 characters in the schema, and published versions are immutable, so each version can be published once. The registry is described as official and in preview. Anthropic announced on 9 Dec 2025 that MCP moved to the Agentic AI Foundation under the Linux Foundation.
Official docs, schema and the announcement, read by two readers.
Changed in the plan: Plan text corrected 2026-11-10: it had called the registry canonical and said PulseMCP and Glama read it; the docs say official, in preview, and do not say Glama reads it.
- The MCP Registry: authentication, FAQ and server.json schema (read in full)
- Donating the Model Context Protocol and establishing the Agentic AI Foundation (read in full)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Run the launch moment
Standard · strong
Product Hunt’s FAQ says never to ask for upvotes and to share the link and request feedback instead; asking for upvotes is listed under spam and can get an account flagged. No source gives a conversion multiplier for either approach.
Official help page and guidelines, read directly; the multiplier we had stated has no source.
Changed in the plan: Plan text corrected 2026-11-10: it had said feedback requests convert about 3x better and avoid the filter.
- Launch FAQ and myths (read in full)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Day −10: line up a hunter, optional
Standard · strong
Product Hunt’s Launch FAQ says there is no advantage to having someone else post a product and that who posts does not affect ranking.
Official help page, read directly.
Changed in the plan: Plan text corrected 2026-11-10: it had said a hunter is worth roughly 15% of day-one momentum.
- Launch FAQ and myths (read in full)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Day −7: line up a customer to comment
Standard · moderate
Product Hunt says a product’s position depends on its points, comments, time since launch and other signals, and its FAQ says upvotes from established, active members count for more. It says nothing official about maker responsiveness or a 2026 algorithm change.
Official pages read directly, but the weighting and the responsiveness claim we had stated are not in them.
Changed in the plan: Plan text corrected 2026-11-10: it had said the 2026 algorithm weights comment quality and responsiveness over raw upvotes, and that one comment is worth fifty upvotes.
- 1. How Product Hunt works (read in full)
- Launch FAQ and myths (read in full)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Reply to every comment within 30 minutes
Standard · moderate
Product Hunt says a product’s position depends on its points, comments, time since launch and other signals, and its FAQ says upvotes from established, active members count for more. It says nothing official about maker responsiveness or a 2026 algorithm change.
Official pages read directly, but the weighting and the responsiveness claim we had stated are not in them.
Changed in the plan: Plan text corrected 2026-11-10: it had said the 2026 algorithm weights comment quality and responsiveness over raw upvotes, and that one comment is worth fifty upvotes.
- 1. How Product Hunt works (read in full)
- Launch FAQ and myths (read in full)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Ask for feedback, never for upvotes
Standard · strong
Product Hunt’s FAQ says never to ask for upvotes and to share the link and request feedback instead; asking for upvotes is listed under spam and can get an account flagged. No source gives a conversion multiplier for either approach.
Official help page and guidelines, read directly; the multiplier we had stated has no source.
Changed in the plan: Plan text corrected 2026-11-10: it had said feedback requests convert about 3x better and avoid the filter.
- Launch FAQ and myths (read in full)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Follow up once after five days
Evidence · weak
G2’s June 2020 What’s New page says review requests sent through Review Automation see a 40% open rate and an 11% conversion rate, with no method stated.
A vendor’s marketing page from 2020 about automated requests; personal asks may do better but no data was found.
Changed in the plan: Plan text corrected 2026-11-10: it had expected roughly 50% conversion from 20 asks.
- What’s New, G2? June 2020 (read in full)
True as of 2026-11-10; to be re-checked by 2027-05-10.
Publish the Play listing before you decide iOS is the market
Practitioner report · weak
In a 2026 Starter Story interview, Journable’s co-founder reported $125,000 over the past 28 days (about $80K MRR) with about 80% of users and revenue from Android. Journable is an AI calorie counter.
One founder’s self-reported figures, read through a transcript summary, and a 28-day figure is not a stable monthly number.
Changed in the plan: Plan text corrected 2026-11-10: it had said it crossed $100k/month; the interview gives $125k over 28 days and about $80K MRR.
True as of 2026-11-10; to be re-checked by 2027-05-10.
Tag every store link so an install can be traced to the post that earned it
Standard · strong
Apple says campaigns appear in App Analytics only after at least 24 hours since launch, and that each metric shows only once it reaches a minimum threshold of 5 in the selected date range; a campaign’s data appears once it has first-time downloads from at least five users.
Official Apple help page, re-checked directly; a 24-hour attribution window from the click, reported by two earlier readers, could not be extracted.
True as of 2026-11-10; to be re-checked by 2027-02-10.
Standard · strong
The Play Console lets you filter store-listing conversion data by UTM source and campaign, taken from parameters on links to the store listing.
Official Google help page, re-checked directly. An earlier note that UTM source needs the ads and referrals filter was not found on the page.
- Understand and grow your app’s user base - Play Console Help (read in full)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Test a free tier against an extended trial
Evidence · moderate
RevenueCat’s 2026 data reports a median day-35 conversion of 10.7% for hard paywalls against 2.1% for freemium, about five times. Its benchmarks page calls the figure trial-to-paid in the text and download-to-paid in a caption; its separate post defines it as download-to-paid. One-year retention of yearly subscribers is 27% against 28%.
Vendor first-party dataset read by two readers; the metric label is inconsistent between its own pages.
Changed in the plan: Plan text corrected 2026-11-10: it had said trial-to-paid, and implied retention was the same for every paywall.
- Subscription app trends and benchmarks 2026 (State of Subscription Apps) (read in full)
- Hard paywall vs freemium (read in full)
True as of 2026-11-10; to be re-checked by 2027-11-10.
Evidence · moderate
RevenueCat reports a median 42.5% trial-to-paid for trials of 17 to 32 days against 25.5% for 3-day trials, and calls this 70% better; the arithmetic gives about 67%.
Same vendor dataset, re-read directly; the comparison group is 3-day trials on the page.
True as of 2026-11-10; to be re-checked by 2027-11-10.
Generate /integrations/[partner] pages
Build proper dunning
Evidence · weak
Stripe’s billing page says $8.2B in failed payments were recovered in 2025 and $14 recovered for every $1 spent on its tools, and lists Smart Retries, expired card updates, recovery and retention automations and cancellation surveys. It states no share of involuntary churn recovered. Two earlier readings of a 55% average recovery rate could not be found on the page.
A vendor’s own landing-page totals with no stated method, and a figure two readers reported was not on the page when it was checked directly.
Changed in the plan: Plan text corrected 2026-11-10, twice: the first version had 30-50% of involuntary churn with no primary source; the second had a 55% that the page does not carry. It now quotes the totals the page does state.
- Stripe Billing (read in full)
True as of 2026-11-10; to be re-checked by 2027-02-10.
Raise the budget in steps and re-read the payback at each one
Standard · strong
Meta says ad sets usually exit the learning phase after about 50 results in the week after the last significant edit. A budget or bid change may be significant depending on its size; Meta publishes no threshold.
Official Meta pages, consistent across three, read as raw HTML.
Changed in the plan: Do not state a budget-change percentage; Meta gives none.
True as of 2026-11-10; to be re-checked by 2027-02-10.
Send every ad to a page whose first screen repeats its promise
Standard · strong
Google Ads calls landing page experience a component of Quality Score. It advises matching the page to the ad and keywords and putting important information toward the top of the page.
Official Google Ads help page, read by two readers; a rating scale reported by the first reader was not found by the second, so it is left out.
Changed in the plan: The phrase first screen is our paraphrase of toward the top of the page. The above-average rating scale is not on this page and is not claimed.
True as of 2026-11-10; to be re-checked by 2027-02-10.
How the library is written: editorial standards.