DEV Community

Cover image for Which launch sites gave our listing a followed link, and how fast
RadarMakers
RadarMakers

Posted on

Which launch sites gave our listing a followed link, and how fast

Submitting a product to a launch directory leaves several questions unanswered. Was it accepted? Can somebody open its page? What kind of website link does that page contain? When will the actual launch happen?

While launching RadarMakers, a free launch platform for indie makers, we recorded our experience with 19 sites. The log contains 10 followed links, 3 nofollow links, 4 UGC links, and 2 entries marked not live yet. Those labels describe our own listings in October 2026. They are not a measurement of these sites in general or a promise about what another submission will receive.

The useful lesson was less about finding a winner and more about keeping publication, link type, and launch scheduling separate. A listing is a link, not an audience. Our recorded first-week referral total was 9 visits, with limits on what that total can explain.

Read the link and publication fields separately

The table has 19 rows and four values in the link field: 10 follow, 3 nofollow, 4 UGC, and 2 not live yet. The final label is a state rather than a relation, which is why it should not be described as a fourth kind of HTML link. Uneed and Launching Next were unresolved when the record was checked. They do not have a published product link in the dataset to inspect in the same way as the other rows.

This difference matters in any report that counts links. If you combine not live yet with nofollow, you imply that an absent listing has a link relation. If you call all 17 visible pages followed or not followed, you leave out the UGC category. If you count approvals as live, you may move TinyLaunch into the wrong state even though the note describes a scheduled date. Keep the raw fields and present them with labels that match what was recorded.

The array is an account of one project experience in October 2026. It is not a claim about a platform policy in every case. A site can change how it publishes listings, how it labels links, or what a free plan includes. The note attached to TinyLaunch says the observed free-plan link was nofollow and records a conditional badge option for a followed link. The article does not treat that optional arrangement as a result received, and it does not recommend adding a badge to obtain a link.

Followed is one field, not a traffic result

A followed link label says how the link relation was recorded in the source row. It does not say whether a person clicked it, whether the page was indexed, how much authority a search engine assigns, or whether the link affects rankings. Those are separate questions requiring separate evidence. The article uses followed as the code value and plain language label, not as a forecast of SEO gains.

The link may also live on different surfaces. BuildShipper, CLAPSTORM, Peerlist, and BusinessHunt were recorded as same-day followed examples. HereIsMySaaS and Startup Ranked were described as within a day. Lifto required another attempt. LaunchRanked and PeerPush were associated with a future launch schedule or queue. The fact that a listing link was followed does not make these states equivalent.

To study clicks, use analytics records with a defined date window and source field. The separate referral array records five source names and nine visits from October 21 to 27. It is not a join to all 19 directory rows. Only a note for HereIsMySaaS explicitly connects that home-page card with two visits in the measured week. Do not assign remaining visits or zero values to platforms by assumption.

This is why the article distinguishes a link inventory from an attribution report. A link inventory can be checked in a listing page. An attribution report depends on analytics collection, filtering, redirects, referrer behavior, and the defined period. A visitor may arrive after a delay, through a share, or with the referrer removed. The available code constants do not answer those questions for every source.

Publication timing is a separate event

The same-day helper looks at the recorded wait wording. The code explicitly excludes text that begins with accepted the same day, then includes text containing same day or live at once. This yields six rows. Four are also followed, one is nofollow, and one is UGC. PeerPush says the page was public at once but is excluded by the helper because the exact text lacks the included phrases. That is a limitation of a text predicate, and the article states it rather than quietly claiming six includes every phrase a reader might consider immediate.

The 3-day, 2-day, and 5-day observations belong to GreenRocket, BuildVoyage, and SideProjectors. The wait helper matches a number of days after and requires a value above one. It does not infer a number from within a day, overnight, or second try. That makes the count repeatable while leaving some qualitative wait descriptions outside it.

Scheduled calendar dates form another kind of evidence. DirArc, TinyLaunch, LaunchRanked, and HUNT0 name specific dates or a launch week. A queue length without a named date does not match the calendar-date helper. The October record can preserve those dates as what the page said then; a later check is needed before claiming the event occurred.

For an operations timeline, keep at least three event fields: submission received, public page observed, and scheduled launch. Approval may be a fourth. It is tempting to put them in one status column because a short table is easier to scan, but a compressed state hides the point at which a link became available. Store the source wording and a checked date, then render the concise view from those fields.

The row notes help explain implementation friction

The log contains observations beyond status and link relation. BuildShipper used a password account. Peerlist required a profile and project page. PeerPush sent an emailed sign-in code. Shipboard required a confirmation link sent by email. Canopy Launch Directory did not need an account. CLAPSTORM rendered its content with JavaScript, and BusinessHunt rendered the website link with JavaScript. These are notes about our attempt, not generalized requirements for every user.

The practical use of these notes is to plan the work and prioritize a second check. If a directory requires account setup, include that task in the launch checklist. If the listing appears through client-side rendering, inspect the visible page and actual link. Do not convert that implementation observation into a claim about how a search engine processes the page without a crawl test. A screenshot or saved HTML from the check can help a later reviewer understand what was observed.

Some entries also explain review friction. BuildVoyage was reviewed before publication. SideProjectors was approved after five days. Lifto rejected the first attempt because the images did not show the product in use adequately. These notes can guide a better submission, but they do not establish that the same review rule applies to every submission or that changing the images guarantees approval.

Separate facts from next-step advice. The source note is the fact. Preparing a clearer product screenshot next time is a reasonable action. Saying that the screenshot change will produce a faster review would be an unsupported causal claim. A good field note can connect an observation to a hypothesis while labeling the difference.

Keep a small but traceable measurement pipeline

The source file is a typed array. A platform row includes a name, URL, wait string, link label, and note. Helper functions apply explicit rules to selected fields. This is a compact design that makes it possible to build several pages from the same dataset without copying every platform name by hand. It also means a change to the shared source can affect several generated pages, so tests should assert both counts and exact records.

Prefer a single normalized source over separately maintained totals in article copy, route metadata, and charts. A build can derive the array and count, then render a table and chart from the same values. If a manually written article is used, the reviewer should compare its names and totals to the source before accepting it. Add a date to any measurement whose truth can change and preserve the old observation if the report is historical.

Validate each field before filtering. A link label should come from the declared set. A wait field should preserve its source wording. A URL should point to the product or listing being described. Notes should avoid adding a status that conflicts with the primary wait field. A schema or test can catch a misspelled relation, while a human can catch a sentence whose meaning does not match the real platform page.

The count tests should cover edge cases, not only the desired total. Test accepted the same day with a future date, public at once, within a day, a number of days after, a queue with no date, and an explicit scheduled date. Verify the output list by name. A total of six alone could remain unchanged even if one valid record disappears and an invalid one enters.

When the research changes, check the entire presentation path: source row, derived helper, visible table, explanatory prose, metadata, chart, social image, and internal links. Review alt text too. A chart with six rows and a paragraph saying four followed needs a clear unit and a caption that identifies the four subset. This is a small amount of discipline that prevents one stale copy from contradicting the others.

Plan another submission without overreading the sample

For a new product, first decide whether a public holding page is useful before the actual launch. If it is, a platform that publishes a page early may offer a link to share while the product launch is still scheduled. If a coordinated launch day matters more, an unknown queue or a future date may change the plan. The October sample provides examples to inspect, not a guarantee about the next submission.

Before spending time on a directory, read its current free and paid terms. Confirm whether an account, profile, badge, reciprocal link, or approval process is required. Then write down the intended outcome: a public product page, a scheduled launch, a referral source, or feedback from a community. Those goals are different. A page that becomes public quickly can satisfy one and fail another.

Use a short submission record with a URL, date, product description version, image set, account steps, and a dated result. Record when the link was checked and what relation it carried. If there was no response, keep the state as pending rather than marking a rejection. If the domain fails, record the resolution and HTTP evidence separately from editorial review. This protects the research from being recast as a success story after the fact.

For referral measurement, keep a campaign-tagged link only where the platform permits it and do not assume that a source without a tagged visit sent no traffic. Review analytics with a defined window and bot filters. A small visit total is a descriptive measurement, not a customer acquisition number. To measure customers, connect visits to a defined product action and a consent-respecting analytics method.

What this dataset cannot answer

The 19 rows cannot rank all launch directories, prove the impact of link attributes, estimate search rankings, or compare long-term referral performance. The wait descriptions do not give a consistent set of timestamps. The source list does not assign every referral to every listing. The data does not contain acquisition costs or customer revenue from the platforms. A chart should not imply any of those answers.

It can answer a narrower question: what did the source file record when RadarMakers submitted to these named sites in October 2026? It can show the recorded categories, publication states, and separate referral counts. It can also show how the code chooses a subset. The value is transparency about the observation, not an inflated claim of representativeness.

If a future report wants to compare performance, define a consistent denominator and collection window first. Keep all attempted submissions, including the pending ones. Report the number checked, the number with a public page, the number with a link of each observed relation, and visits with the analytics caveats. An outcome should be compared only if the records support the same definition across platforms.

Separate observation, interpretation, and advice

An observation is a row value or a page state that was checked. An interpretation is a conclusion drawn from those records. Advice is an action a maker might take next. Keeping the three layers separate improves trust. For example, the source records ten followed link labels. It does not establish that followed links brought more visitors. A maker may still choose to inspect link relations as part of a technical audit, but should not treat that label as a traffic measurement.

The article should preserve words such as about, within, and not scheduled when they are part of the source. They mark different levels of precision. About 97 days is an approximate queue statement. Within a day is a loose response window. Not scheduled means no launch date had been booked in the source note. Replacing those phrases with exact durations would imply timestamps the dataset does not contain.

The distinction between a public page and a launch event deserves equal care. A coming-soon listing can be publicly viewable before a scheduled launch. That state may help a maker share an address, but it is not evidence that the launch promotion has occurred. A later status check can say whether the event happened, while the October row remains a record of the earlier promise.

Use the record to plan a follow-up measurement

The present data can guide what to collect next. Save the date a submission was sent, the first date a public page was observed, the date of any scheduled event, and the date the outgoing link was inspected. Store the page address and the exact relation value. If the platform has a profile page and a separate product page, identify which one was checked. This makes later updates comparable without pretending the original notes contained more detail.

Then define an attribution period before reading analytics. A visit reported from a source domain can occur on a different day from publication. Campaign tags can help where the publisher permits them, but a referrer field can be absent or changed by redirects. Keep a separate unknown or unattributed state. Do not convert a missing source into zero visits, and do not assume a visit is a unique person.

Connect the visit record to a useful product action only when the analytics system supports it. A click can be followed by a page view, a signup, a product setup step, or no further event. The 2026 record in this article contains the source and visit count only. It does not supply a signup count or a customer outcome. A future study could define those events, but this article should not retroactively claim them.

Design a more maintainable launch log

Keep one canonical row per platform attempt and use stable field names. A second attempt should be a new event attached to the same platform, not a replacement that erases the first. Store who performed the check only where needed for internal work, and keep public copy focused on the method and observation. The article should not expose private account information or personal contact details.

Separate a platform policy from a one-time interface observation. If a site publicly states a link rule, save the policy source and its date. If a link was inspected in one listing, record that concrete result separately. The policy can change, and one listing can differ because of a plan or a special condition. A historical row should say which one it represents.

For the next audit, define status values before collection begins. Useful states might include submitted, needs clarification, approved, public, scheduled, rejected, and unresolved. The link relation belongs in another field. A clear schema prevents a category such as nofollow from being used to describe a listing that has not gone live.

At the end of a check, save evidence sufficient for another person to review the result. That may be a permitted screenshot, a public page address, or the visible link markup and check date. Avoid storing unnecessary personal data. If a platform blocks automated tools, inspect it in a normal browser and note the limitation rather than presenting a robot error as a site failure.

Read the chart without turning it into a ranking

The first chart groups the recorded relation labels and the two not-live states. It is a snapshot of our 19 rows. The categories do not measure platform quality, authority, or expected reach. The not-live rows have a different unit from relation labels, which is why the caption names them as unresolved listing states. A reader should not stack them as if every bar describes an HTML link.

The second chart shows the five referral source counts from a separate analytics array. The bars total nine visits in the October 21 to 27 observation window. Their order is for readability, not a platform ranking. The source list includes domains that are not a one-to-one mapping of the submission rows. The chart does not support a per-platform yield calculation.

When a visualization contains few observations, label every bar and preserve the exact unit. A chart title that says visits is clearer than a title that says traffic. A caption that says the values are not users is more accurate than a generic growth claim. The accompanying table remains available for readers who prefer text or need to compare exact values.

A review checklist for future updates

Before updating the page, verify the source row, the observed date, the meaning of each field, and the relevant helper function. Recompute counts from the same input rather than copying an old total. Review the positive and negative examples. Check that all link relations are described consistently, and that unresolved rows remain visible.

Then test the output. Confirm the chart values, alt text, title, description, canonical, and internal links. Ensure a referral count remains separated from link labels. Confirm that scheduled dates are explicitly historical if the page is a snapshot. Finally, record the check date and preserve the previous observation when the platform state later changes.

The goal is not to make a small log look larger. It is to let another maker understand exactly what was observed and what remains unknown. That is enough to make the report useful without claiming that a directory submission guarantees an audience.

What we actually recorded

The source is frontend/src/lib/launch-platforms.ts. Each row has a platform name, URL, wait description, link label, and note. It is a compact experience log rather than a controlled comparison. The wait fields mix publication events, approvals, queues, and future launch dates because those were the states we encountered.

Here is the link breakdown counted from LAUNCH_PLATFORMS:

Recorded label Entries Platforms
Followed 10 GreenRocket, BuildShipper, CLAPSTORM, Peerlist, BusinessHunt, HereIsMySaaS, Startup Ranked, Lifto, LaunchRanked, PeerPush
Nofollow 3 TinyLaunch, DirArc, Canopy Launch Directory
UGC 4 BuildVoyage, SideProjectors, HUNT0, Shipboard
Not live yet 2 Uneed, Launching Next

These are the file categories. A row marked not live yet is an unresolved listing outcome, rather than another link attribute. Likewise, a recorded followed link does not establish that the platform scheduled promotion has already happened.

October launch record counts: 10 followed, 3 nofollow, 4 UGC, and 2 not live yet.

We keep the followed-link directory list as a way to inspect this experience. For planning, the notes matter alongside the label: a public holding page and a completed launch can both leave you with a link, while representing different stages of the process.

Same day needs an explicit definition

The repository uses a helper to decide which wait descriptions count as same day. Excerpt from frontend/src/lib/launch-platforms.ts:

export function isSameDayListing(wait: string): boolean {
  const text = wait.toLowerCase();
  if (text.startsWith("accepted the same day")) return false;
  return text.includes("same day") || text.includes("live at once");
}
Enter fullscreen mode Exit fullscreen mode

This is a design decision in our reporting code. It classifies recorded wording; it does not calculate elapsed time from submission and publication timestamps. In particular, acceptance followed by a future launch date is excluded. An approval message alone is not enough.

Applying that helper gives 6 same-day entries. BuildShipper, CLAPSTORM, Peerlist, and BusinessHunt had followed links. Canopy Launch Directory had a nofollow link, and Shipboard had a UGC link. The same-day followed-link subset therefore contains 4 entries.

There is an important limitation. PeerPush's wait field says its page was public at once, but that phrase does not match the helper. Its followed link is consequently outside this same-day count. The observation remains in the row; the classification is narrower than all possible descriptions of immediate publication.

HereIsMySaaS and Startup Ranked are described as approved or live within a day. We do not silently translate that into same day. Lifto went live after a resubmission, without a precise elapsed time in the field. Keeping those distinctions makes the count reproducible without pretending the log has timestamps it lacks.

A launch date is not a publication date

GreenRocket's row records a launch 3 days after submission, with its page public beforehand. LaunchRanked records overnight approval and a launch week scheduled for December 21 to 27, 2026, while a coming soon page was already public. Those observations show why a single waiting-time column needs explanatory text.

PeerPush is another example: its public page coexisted with a free launch queue of about 97 days. That approximate queue was what we saw then, not a service commitment. A maker seeking a page to share and a maker coordinating a launch announcement would read the same row differently.

The distinction also applies outside followed links. TinyLaunch accepted our submission the same day but scheduled the launch for November 23, 2026. DirArc had a page public before a November 9, 2026 launch, and HUNT0 had a public product page before its launch week scheduled for March 1, 2027.

Those future dates belong to the October record. They do not tell us whether those launches eventually happened as scheduled. Our design choice is to preserve the recorded state instead of presenting every acceptance as a completed launch.

The notes expose submission work

The notes describe small operational details that the link totals hide. BuildShipper used a password account. Peerlist involved a profile and project page. PeerPush used an emailed sign-in code, while Shipboard required confirmation through an emailed link. Canopy Launch Directory needed no account.

CLAPSTORM's page loaded its content with JavaScript, and BusinessHunt's website link was rendered with JavaScript. These are implementation observations from our experience, not evidence about how a search engine processed either listing. The link label alone cannot answer that question.

Review also varied. BuildVoyage's submission was reviewed before publication, and its recorded wait was 2 days. SideProjectors was approved after 5 days. Lifto needed another attempt because the initial submission did not demonstrate the product in use adequately.

For nofollow entries, context matters too. TinyLaunch's note records nofollow on the free plan and describes a conditional followed-link option for the top 3 showing its badge. We did not record that conditional outcome for our listing. The nofollow listing notes keep the actual outcome visible rather than treating a possible upgrade as something received.

Uneed and Launching Next remained unresolved in this log. Uneed's free waiting line was full, and we saved a draft without booking a free launch. Launching Next offered a paid queue skip that we did not take. Neither row should be counted as a public listing merely because submission work had started.

What the referral total can support

The separate REFERRAL_VISITS array records these sources for the first week:

Recorded source Visits
hereismysaas.com 2
alohafind.com 3
admin.webcatalog.io 2
thelaunchdirectory.com 1
github.com 1

The source calculates the total directly. Excerpt from frontend/src/lib/launch-platforms.ts:

export const REFERRAL_VISIT_TOTAL = REFERRAL_VISITS.reduce((sum, row) => sum + row.visits, 0);
Enter fullscreen mode Exit fullscreen mode

That produces 9 recorded visits. HereIsMySaaS also has a note connecting its home-page card with 2 visits during the measured week.

Referral visits from October 21 to 27: alohafind.com 3, hereismysaas.com 2, admin.webcatalog.io 2, thelaunchdirectory.com 1, and github.com 1.

The referral sources do not match the submission list exactly. We cannot divide this total across the 19 platforms, assign missing sources zero visits, or compare followed and nofollow performance from these arrays. The file supplies neither that mapping nor evidence about signups or customer outcomes.

How we would use this log again

For another launch, we would record acceptance, public availability, and scheduled promotion separately. We would keep link type alongside those states, preserve uncertain waits as uncertain, and inspect the actual listing before marking the submission complete.

That is a proposed workflow, not a measured improvement. The observed result is narrower: our submissions produced different link labels and publication states, while the recorded referral total stayed a separate piece of evidence. Getting listed completed a distribution task. Whether anybody arrived and found the product useful still needed its own evidence.

Sources

The link labels, wait notes, helper functions, and referral values are from frontend/src/lib/launch-platforms.ts, checked against the current repository. The data describes RadarMakers submissions in October 2026. The charts are exact counts from that source. They do not estimate general directory behavior.

Top comments (0)