DEV Community

Cover image for A free submission is not a live listing: lessons from our 43 site audit
RadarMakers
RadarMakers

Posted on

A free submission is not a live listing: lessons from our 43 site audit

A submission form can show a success message while the product remains absent from the directory. If your launch tracker marks both events as "done", it loses the distinction that matters to someone waiting for a public page.

We ran into this while organizing our own directory research for RadarMakers, a free launch directory for indie makers. The live report records checks of 43 startup directories and launch sites between October 24 and October 28, 2026.

Four free forms accepted an entry. The other 39 checks did not lead to a submission. None of those four submitted entries was confirmed published in the recorded notes.

That last sentence does not say the entries were rejected. It does not establish their status today. It says exactly what the evidence supports at the point recorded.

The free startup directory submission report keeps the dates, outcomes and individual notes visible. We read its live version on October 29 for this article. The observations below come from that report rather than from new submissions made today.

Start with what happened, not a free or paid label

The report groups the 43 checks by what happened in our process:

Recorded outcome Sites
A free form accepted our submission 4
Skipped over badge or link back terms 12
Payment was the reason we skipped 9
Skipped because of an access condition 16
Other notes 2
Total 43

These are outcome groups for this particular report. They are not a market survey of all startup directories.

The access condition group includes circumstances such as a captcha, a bot check, a required personal name or a sign in route we chose not to use. Our choice to stop does not prove that a person following a different process could not submit for free.

The badge group also needs its row notes. Depending on the site, the condition affected submission itself or the kind of outbound link a listing could receive. Those are different decisions for a maker.

A useful table keeps the grouping simple without discarding the explanation that makes each row interpretable.

An accepted entry still has a publication question

The four accepted submissions in the report are DevPages, Minimal Gallery, Tiny Startups and TOOOLS.design. Each was awaiting review in the recorded notes.

There is enough evidence to say we submitted. There is not enough in those notes to say a page was published.

For a launch tracker, those observations should live in separate fields. A single "done" checkbox cannot explain whether someone inspected a form, submitted an entry or found a public listing.

You can see why the difference matters when scheduling work. An accepted entry may mean the next step is waiting. A confirmed public page may mean the next step is checking the URL and recording what it contains. A blocked form may require a different submission method or a decision to stop.

None of those states needs to be presented as failure. Each needs an accurate next action.

Keep observations separate from decisions

Consider the access checks. A note that a form displayed a captcha is an observation. A note that we stopped there is a decision. A claim that the directory has no free submission route is a broader conclusion, and the first two notes do not establish it.

That separation is useful beyond directory research. It applies to integration testing, software evaluations and any workflow where your process has limits that another user might not share.

One way to preserve it is to store the visible condition and your response separately. Here is a possible tracking shape, not the schema used by RadarMakers:

type DirectoryCheck = {
  checkedOn: string;
  observedCondition: string;
  ourDecision: "submitted" | "skipped";
  publicationEvidence: "confirmed" | "not confirmed";
  publicListingUrl?: string;
};
Enter fullscreen mode Exit fullscreen mode

The point is the separation, not these exact field names. A record can say "submitted" while publication remains "not confirmed". A skipped record can describe the condition without labelling the site unavailable to everybody.

If you later confirm a listing, record the evidence and the date of that confirmation. Do not erase the earlier observation or replace its date with the day you edited the report.

A second count needs a second explanation

The same report also summarizes the reasons recorded for the 39 skips:

Recorded reason Sites
Badge or link back condition 12
Captcha or bot check 12
Paid submission 9
Needs a named person 2
Single identity provider sign in only 2
Nofollow on the free tier 1
Submission page unavailable 1
Total skipped 39

The second table explains the skipped checks. It is not another set of 39 sites to add to the 43.

Its categories also answer a different question. The first table tells the reader where our process ended. The second explains the recorded reason for stopping.

The nofollow entry is particularly important to frame honestly. It records our selection decision, not a rule that a nofollow listing is useless. Link attributes and referral value are different subjects. This report alone does not tell us how many readers a listing would send.

Put the limitations beside the result

RadarMakers is itself a launch directory, so we have an interest in the subject. The report describes our own constraints next to its results.

It also excludes entries where the free status was unclear or the evidence did not meet its scope. That means the table cannot answer "How many free directories exist?" It can answer "What happened in these recorded checks?"

The practical editing step is to keep those boundaries near the count. A headline number without the selection rule can look more comprehensive than the underlying records.

Before publishing a similar tracker, check whether a reader can distinguish the inspected path, your decision and the evidence of publication. Keep pending work pending. Keep unknown facts unknown. A smaller claim that someone can verify is more useful than a completed checkbox that hides several different events.

Recorded outcomes in the 43 directory checks.

Chart source: RadarMakers live report snapshot captured October 29, 2026.

Recorded reasons for 39 checks that did not lead to a submission.

Chart source: RadarMakers live report snapshot captured October 29, 2026.

Make every record reproducible

A useful audit row should allow a second researcher to follow the same path and understand why the original person stopped. Store the exact page inspected, the date, and a concise observation. Avoid relying on memory or on a general home page link. A directory can have a public editorial policy, a vendor portal, a paid launch page, and a separate suggestion form. If the record does not say which route was checked, the next person may repeat the wrong task.

The October report covers observations made over a defined five-day window and was captured on October 29. That date boundary matters because a page or policy can change. A reader should not interpret an October observation as a claim about the form today. Keep the original check date even when a later editor revises the row. Add a second dated observation instead of silently replacing history.

Separate four facts: which route was opened, what was visible, what decision the researcher made, and what evidence later showed a public listing. Those facts often arrive at different times. A button can accept a submission without proving publication. A receipt can confirm a form was sent, not that an editor approved it. A live page can confirm publication, but not necessarily that the entry came from the form being studied. The record should state only what each piece of evidence supports.

Use statuses that preserve uncertainty

A tracker should include unknown and pending as real states. If a submission is awaiting review, record that state and the date it was last checked. If no listing URL has been confirmed, leave that field blank rather than guessing. If a page was inaccessible in the browser, distinguish that observation from a claim that the whole site is unavailable.

This vocabulary helps a team avoid false precision. "Submitted" describes an action. "Accepted by the form" describes an interface response. "Approved" describes an editorial decision only when the site explicitly communicates it. "Live" requires a public destination that can be opened. A status label should be tied to evidence and should not silently move from one meaning to another.

Unknown does not mean useless. It points to a follow-up question. The team can revisit a route, ask the editor, or decide that verification is not worth more time. Record the next action only when someone is responsible for doing it. An unattended reminder can become another form of misleading certainty if a pending status is never revisited.

Read the two tables as different views

The audit has 43 checked options and 39 checks that did not lead to a submission. Those totals describe different views of the same research. The first outcome grouping covers all checks; the second groups reasons among the skips. A reader must not add those groups together, since the latter is a subset rather than another population.

This denominator discipline is useful whenever a process log becomes a chart. Put the population in the title or note, name the period, and define the unit. Say whether a row represents a directory, a route, or an attempt. If a directory has several routes, a directory count and an attempt count are not interchangeable. If categories overlap, say so. If each record has exactly one primary reason, explain how that rule was applied.

The counts in this report do not estimate the total number of free directories. The sites were selected for this project, and the researcher used their own criteria for proceeding. A different maker could accept a badge requirement or complete a verification step that this team chose not to use. The chart describes recorded decisions, not objective properties of every possible submission route.

Keep the reason attached to the next action

A reason code should help someone decide what to do. Payment may mean stop under a no-cost policy. A badge condition may require a branding decision. A captcha may call for a manual browser check. A nofollow attribute may be acceptable if referral traffic is the goal. An unavailable page may deserve a later retry. A single rejected label erases these distinctions and makes the backlog harder to act on.

Keep a controlled category and a short plain-language observation. The category supports sorting. The observation explains the evidence. Do not let the category imply more than the observation. If a form displayed a bot check and the researcher stopped, that is narrower than saying the directory blocks submissions. If a policy states that a badge is required for the free tier, quote or link to the policy that supports the statement.

When a case does not fit the current categories, do not force it into the nearest one. Mark it as other, preserve the evidence, and review whether the taxonomy needs another category. A taxonomy should make genuine patterns visible without changing what an individual row means.

Establish a safe stop rule

Research can become repetitive when a team has no condition for pausing. Write a stop rule before opening another form. For example, a team may decide not to proceed when a route requires payment or an on-site badge. It may defer a route that cannot be reached through its available browser environment. These are the team's operating choices, not claims that the directory has no other route.

A stop rule should allow new information to reopen a decision. A later free editorial route is a new observation. A temporary outage that clears is a new check. Preserve the earlier date and add the new evidence. That way, the history explains both the original decision and why the current status changed.

Also decide what follow-up is worth doing. A submitted entry may merit a check for a public URL. A route that required a condition the project will not accept may need no further work. If there is no person or date attached to a follow-up, do not leave a vague pending item that looks like a commitment.

Measure the work you actually want to improve

This report records checks, decisions, and submissions. It does not measure time spent, referral visits, signups, or revenue. It cannot identify which directory was most valuable. If a future audit needs to answer those questions, add a measurement plan before starting rather than inferring outcomes from the submission count.

A future process could record research time, time to complete a form, confirmed listing URL, and a tagged destination link where the site's rules allow it. A web analytics report could then show visits attributed to that link, with the limitations of attribution and consent kept clear. Visits alone would still not prove that a directory caused an eventual purchase or that the traffic matched the intended audience.

Every additional field has a cost. It needs a definition, a source, and a maintenance owner. If a column does not support a submission, follow-up, or investment decision, it may be clutter. A small dataset with maintained definitions is more valuable than a large spreadsheet whose meanings differ between rows.

Write captions that survive sharing

Charts are often copied into a slide or a social post without the surrounding article. Give each one a self-contained caption with the source, capture date, unit, and denominator. Do not rely on a color legend alone to distinguish outcomes. Direct labels make the graphic understandable when viewed at a smaller size or without color.

The first graphic here describes the recorded outcome groups across 43 checks. The second describes the reason groups for the 39 skips. A person who sees either image alone can still identify the source snapshot and what was counted. That small amount of context prevents a process log from being mistaken for a market-wide benchmark.

If a visual is updated after a later audit, change the date and data together. Do not leave an old caption under a new count. Keep the underlying source file beside the exported graphic so another editor can inspect each value. When a value cannot be tied to a row or source, remove it until it can.

Turn the audit into a repeatable workflow

Before research begins, define inclusion criteria and the meaning of each status. During the work, capture the route and visible terms. After a submission, save its confirmation and set a verification step. At the end of a session, review incomplete rows and either assign a next action or close them with a reason. This sequence is simple enough for a small team and clear enough for a handoff.

A second researcher should be able to reproduce the summary from the row-level data. If they cannot, the table needs clearer categories or a note about exclusions. The audit should not depend on the author remembering which sites were evaluated or why a particular policy mattered.

The practical conclusion is modest. The recorded process distinguishes free submissions sent from checks that did not proceed and from public listings that were not confirmed in the notes. That is enough to show why a launch tracker needs more than a single done checkbox. A careful record preserves the action, the evidence, the uncertainty, and the next decision without claiming more than it knows.

Hand off the evidence with the summary

A useful handoff includes both the short result and the files that support it. Keep the dated JSON snapshot with the article, name the capture date in the visual, and make sure the table values can be traced back to individual rows. If an editor changes a category label, they should document how the source rows were regrouped. Otherwise a later reader may be unable to reproduce the chart from the saved evidence.

Before publishing an update, compare the chart totals with the row count and inspect outliers. The first view must reconcile to the audit population; the skip-reason view must reconcile only to skipped attempts. Check the language against the underlying notes and remove any claim that turns a local decision into a universal rule. This simple review protects the reader and makes the next audit easier to compare.

Preserve row identity when the audit changes

Assign every route a stable row identifier so a later update can be compared with the observation that came before it. A row key should refer to the route or attempt, not encode an assumed outcome. If a directory changes its form or separates a launch submission from a vendor profile, record the new route as a new observation and connect it to the same directory record. This makes it possible to revise the current recommendation without erasing the evidence behind an earlier decision.

Keep a small change note when the taxonomy itself changes. If a team moves a route from an access category to a submission-unavailable category, it should say which evidence led to that reclassification. Otherwise a future comparison may treat a change in labels as a change in site behavior. Stable row identity and dated classification notes make a modest audit more transparent than a polished chart with no trail back to its source.

Top comments (1)

Collapse
 
amineoua profile image
Amine •

Submitted is not live is a distinction every maker needs to hear. I like that you kept unknown and pending as real states instead of forcing them into done. That is the same trap with launch checklists, a form can say success while the page is still nowhere to be found. Did any of the four accepted ones actually go live, or are they still waiting for a public page?