Keep five types of information separate
Checkable fact
What to record: A direct observation about a specific product, market and date, with its source.
What it does not establish: That one page represents an entire market.
Publisher claim
What to record: Attribute the statement to the brand or seller and preserve its limits.
What it does not establish: That the claim has been independently verified.
Third-party estimate
What to record: Keep the sample, time window, unit and estimation basis.
What it does not establish: Actual marketplace orders or future sales.
Your assumption
What to record: State who made it, the scenario and what needs testing.
What it does not establish: That a calculated result is realized profit.
Unknown or conflicting
What to record: Record the gap or disagreement and the next check.
What it does not establish: That a missing signal means zero demand.
First, make sure you are researching the same product
One keyword can surface kits, finished products, spare parts and different pack sizes. Record the model, variant, use, customer group, materials, specifications, package contents and target marketplace before selecting comparable products.
A similar image, brand name or category is not enough to establish a match. If identity is uncertain, pause price and sales comparisons until the evidence is stronger.
Trace each important claim to a source
A report may combine third-party commerce data, public web pages and AI synthesis. The platforms, fields and dates actually covered depend on the evidence available in that particular report. An accessible page establishes only what could be seen at that time.
Keep the direct link, observation date, product or sample, and the exact claim it supports. A website homepage, search page or brand name alone is not a sufficiently specific citation. Third-party sales estimates must not be presented as official transaction records.
When a source fails, identify what is missing
If a source times out, a field is missing or a report indicates degraded data, first identify the affected inputs. Unavailable data is not zero sales. A search page or fallback lead is not observed evidence of demand.
Even if a report remains readable, that does not establish complete live data coverage. Keep affected conclusions unknown until sources are restored or checked again. Do not calculate profit from failed or missing inputs.
A score is not a probability of success
SeaClaw combines demand, competition and comparable sales and margin inputs through a rule-based screening score bounded from 0 to 100. Inputs can still contain estimates and judgments. A score does not turn them into observed facts, and non-comparable dimensions should not produce confident conclusions.
A score of 77 does not mean a 77% chance of success. Sample quality, timing and data coverage may differ between reports. Read the component scores, contradictions and uncertainty together instead of ranking products by the total alone.
Define the costs before interpreting profit
Specify the selling price, currency, per-unit or per-pack basis and order quantity. Then identify whether sourcing, inbound freight, fulfillment, marketplace fees, returns, advertising and taxes are included. Changing the quote or delivery terms can change the result.
Gross margin, contribution profit and net profit use different definitions. Until your own quotes and fulfillment conditions are known, achievable profit remains unknown. This template supplies no standard fee or tax rates and does not decide whether you should buy stock or run ads.
Dates and reviews mean different things
A generation date tells you when a report was produced. A source date describes the underlying information. An editorial revision date records a substantive content change. Opening an old page does not make it a new observation.
A human review needs a real reviewer, date and stated scope. Without that record, leave it pending. This method is a standard for reading evidence; it does not certify that every historical report has already met that standard.
Finish with a check you can actually perform
State the decision you need to make and identify the missing information most likely to change it. Your next step might be checking a model, obtaining a comparable supplier quote, testing a sample or gathering feedback from intended buyers.
Give the action an owner, completion criteria and a date. Revisit the decision when new evidence arrives: continue, change direction or stop. Generating a report, copying a template or seeing a high score is not the end of product validation.
See the distinction in a published case
The DIY miniature house case preserves conflicting specifications and separates historical claims from current evidence. It shows how to record an unknown; it does not establish sales or profitability.
Read the DIY case →Turn the open questions into your next research step
Bring a product link, target market and the question that matters most to the existing SeaClaw workspace. The template is not imported automatically.
These public guides and templates are available in three languages. Workspace, generated-report and export languages depend on the existing product capabilities.