LaunchDetect

Published (UTC)

Back to Launch Watch

Which research questions fit a public launch evidence archive?

A public collection of launch evidence can help you investigate a specific record. It cannot answer every question that happens to mention launches. The difference is often visible before you open the first image: does the question ask about the material actually available, or about a larger population the collection has not defined?

Start with a bounded question and a list of the evidence it would require. This guide offers a question-fit worksheet for researchers, students and editors. It does not assume that LaunchDetect’s /launches/ pages form a complete historical archive, or that every launch has a public record.

Choose the question before choosing the page

Consider the difference between these two proposed projects:

  • “Which sources support the statements on this selected evidence page?”
  • “What fraction of all launches worldwide did the service detect?”

The first can begin with the page and its cited material. The second needs an independently defined set of eligible events, a period, inclusion rules and a way to account for missing observations and results. A collection of conveniently available pages supplies neither the whole population nor the missing cases merely by existing.

This is a scope test, not a judgment about the collection’s quality. A small set can be excellent material for a source-reading exercise while remaining insufficient for a global performance estimate.

LaunchDetect’s pilot page describes published records as distinguishing observations, schedule information and modeled estimates. It also describes agreeing on launches and a period for a historical replay. Those are the company’s descriptions of its offering, not independent validation of a detector or proof of public-record completeness.

Three questions that can fit, with conditions

1. What does one traceable record actually support?

Choose a specific page, record when you read it, and list the claims you want to use. Ask which are supported by an observation, which come from another source, and which describe a model. If the necessary source cannot be recovered, that gap can be a legitimate finding of the reading exercise.

A useful output is a short account of the record’s evidence boundaries. It might say that a source supports the reported event, while a separate image interpretation remains unverified. This question concerns the selected record; it does not quietly expand to every record on the service.

2. Which documented observation times can be compared?

Select records because they meet a written rule, not because their timestamps look convenient. Check the meaning and precision of each time. A comparison of documented observation times is different from a study of first-alert delivery. The latter needs publication and receipt history as well.

The output may be a list of comparable cases plus a list that cannot be compared. Keep both. If missing cases are removed without explanation, the reader cannot tell which question the remaining sample answers.

3. How does a claim travel through its sources?

Start with one assertion and follow its references. Distinguish an originating account, a downstream copy and a separately produced observation or analysis. The research output is a source chain with unresolved dependencies marked. It is not simply a count of websites that repeat the assertion.

For the detailed claim-review step after selecting a question, use our launch-claim verification guide.

A question-fit worksheet

The following is an original planning tool. “Fits” means the listed evidence could answer the bounded question, not that a particular page has already passed review.

Matching a research question to its minimum evidence
Proposed question Minimum evidence to look for Fit decision
Which sources support one selected record? Identified page/version, exact claims, recoverable references Suitable for a source audit when those materials are available
How do documented observation times differ in my selected set? Stated selection rule and compatible time definitions Suitable only for the selected, comparable cases, with gaps reported
Where did one reported claim originate? Attributed statements and links between sources Suitable for a source-chain study; shared origins must stay visible
Did the service observe every launch worldwide? Independent eligible-event list, coverage evidence and all outcomes The public collection alone is insufficient
What was every payload on a launch? Appropriate mission or payload documentation An image or event label alone is insufficient
How accurate is the detector? Defined task, reference standard, eligible sample and error definitions A set of positive examples alone is insufficient

Before collecting data, add one sentence to your notes: “My unit of analysis is ____.” It could be a page, a claim, an observation or an event. If that unit changes halfway through the project, revise the question explicitly.

A historical NASA example of narrowing the question

NASA’s February 8, 2024 PACE launch release reports that the spacecraft launched aboard a Falcon 9 from Space Launch Complex 40 in Florida at 1:33 a.m. EST that day. The release also records a later correction to the mission’s expanded name.

That source can anchor the question, “What event did NASA report, and what correction is visible in the release?” It does not, by itself, answer whether a particular weather-satellite image depicts the launch, whether LaunchDetect published a corresponding record, or how quickly a detection reached a reader. This article has not established those additional facts for PACE.

The practical move is to divide the project. First document the operator’s historical event account. Then state the separate evidence required for an imagery or delivery question. An unanswered second question does not erase the useful answer to the first.

Write the limit into the result

A strong conclusion can be modest: “We examined the selected public records and traced the available source links; some observation-time comparisons remained unresolved.” Replace “selected” with your actual selection rule, and identify the unresolved cases.

Avoid upgrading that result to a claim about all launches, all observation opportunities or validated accuracy. Those are different studies. If the real goal is to decide whether the service improves a team’s workflow, our pilot scorecard addresses that evaluation. The task here comes one step earlier: asking a research question that the evidence you can obtain is capable of answering.

Sources

Pages checked October 3, 2026. The worksheet is proposed research guidance, not a completed archive audit or performance study.