How to read ROIpad public research

These pages connect public records with research generated by ROIpad’s Deep Analytics pipeline. They are intended to help you investigate markets and products, not to replace primary-source research.

Three different kinds of information

Evidence and provenance

Public market briefs include only research profiles with at least one HTTP(S) source reference in the research database. References are supplied by the upstream analysis and have not necessarily been independently verified. A link alone does not establish that every claim is supported. Profiles based only on model training data are omitted. Research dates describe when an assessment was generated, not when all underlying events occurred.

How a record is connected to market research

A stored entity classification can establish an analytical market association. Otherwise, we consider unambiguous exact multiword keyword associations and distinctive words shared with the market taxonomy. Ambiguous ties are omitted. These are transparent topic associations, not proof of commercial demand.

Tag-specific business research is shown on entity pages only when its scope shares at least two substantive terms with that entity's supplied context. Market briefs show separate labelled scopes for readers to compare. A broader market assessment is not a claim about an individual product. Source references are retained.

Activity measurements

Week-over-week calculations require fourteen consecutive daily observations. Missing days are not replaced by zero. A zero comparison period has no percentage change. Corpus coverage uses exact extracted keyword matches and distinct public source records; it measures the stored corpus, not demand or independent corroboration. Topic similarity and overlapping terms remain imperfect filters.

Comment sentiment

We examine up to 60 recent, distinct substantive comments from up to 180 stored records for an item. Comments shorter than 40 characters are excluded, repeated text is deduplicated, and long comments are truncated at 1,000 characters. The default eligibility threshold is 20 comments, configurable by the administrator from 10 to 60.

Gemini assigns aspect labels to the supplied text. ROIpad validates that displayed excerpts occur in the comments and calculates the totals itself. A comment may discuss multiple aspects; positive and negative counts are mentions, not percentages of customers. Mixed assessments use the neutral category. Quoted text is public source content and is not an endorsement by ROIpad. Source links may lead to the parent discussion when a comment-level URL is unavailable.

Issue trackers overrepresent problems, launch discussions may contain promotional feedback, and app reviews have selection bias. We do not treat any of these as a representative customer survey. No sentiment inference is shown below the configured sample threshold.

Comparisons

Feedback comparison candidates must share an inferred market, pass a substantive topic-overlap check, and meet the comment threshold. A weakness-based candidate requires at least three negative mentions on the current item and at least three positive mentions, exceeding negative mentions, on the candidate for that same aspect. Similar-pattern candidates share an aspect with at least three positive mentions. These are leads for investigation, not “winners,” verified replacements or claims of comparative performance. Candidates are drawn from a bounded set of recently cached products, so coverage is incomplete.

Freshness and cost controls

Snapshots normally refresh every 30 days through a background job. Visits register missing pages but do not run AI analysis. Old snapshots remain available during refresh or service failures. The displayed timestamp is the snapshot build time; sentiment shows its own analysis time. Unchanged comment samples reuse their analysis. Daily generation limits and bounded worker batches control usage.

Scope and privacy

This public enrichment layer uses public-source records and Deep Analytics market research. It does not publish private customer projects, competitor reports, support integrations or customer conversations.

Interpretation checks

Download counts are activity measures, not unique users. Repository stars are not customers or revenue. SEC offering amounts are proposed offerings, not proof of funds raised. Academic topic overlap does not establish a successful commercial application. Before making a decision, open the linked sources, check dates and confirm that the context fits your use case.