Startup Opportunity Window × Evidence Matrix
A reproducible cross-tab of 1,012 curated startup records by editorial timing label and evidence class.
ProvenStartups is the Organization author and publisher. Snapshot date: 2026-09-30.
This asset cross-tabulates 1,012 curated startup records across two separately assigned editorial dimensions: timing and evidence. The timing labels are Evergreen and Window of Opportunity. The evidence classes are third-party verified, founder-reported, creator-relayed, and unproven.
The labels organize this collection. They do not predict market duration, urgency, success, claim truth, or causation.
What does the matrix show?
The snapshot contains 705 Evergreen records, or 69.7% of the full set, and 307 Window of Opportunity records, or 30.3%. No record is in the Other or unstated timing class.
| Editorial timing | Verified | Founder-reported | Creator-relayed | Unproven | Total |
|---|---|---|---|---|---|
| Evergreen | 39 | 389 | 259 | 18 | 705 |
| Window of Opportunity | 26 | 148 | 106 | 27 | 307 |
| Other or unstated | 0 | 0 | 0 | 0 | 0 |
| Total | 65 | 537 | 365 | 45 | 1,012 |
Within the Evergreen row, the shares are 5.5% verified, 55.2% founder-reported, 36.7% creator-relayed, and 2.6% unproven. Within the Window row, they are 8.5%, 48.2%, 34.5%, and 8.8% respectively.
The matrix also reports column shares. Evergreen contains 60.0% of all verified records, 72.4% of founder-reported records, 71.0% of creator-relayed records, and 40.0% of unproven records. The Window row contains the remaining 40.0%, 27.6%, 29.0%, and 60.0%.
Row and column percentages answer different questions. A 5.5% figure describes the verified share of Evergreen records; 60.0% describes the Evergreen share of all verified records. The downloads retain both denominators to prevent those interpretations from being mixed.
How was the cross-tab produced?
Each published record already has one timing label and one evidence class. The generator normalizes those labels, counts every timing-by-evidence intersection, and attaches three denominators: the timing row, the evidence column, and the full 1,012-record snapshot.
The public output contains 12 rows: three timing classes multiplied by four evidence classes. The zero-count Other or unstated rows remain in the output so the classification space is explicit.
The public files contain aggregate results. Full regeneration requires the private 1,012-record collection.
Why is the matrix independently useful?
An individual founder interview or project source can support a specific company claim. It cannot establish the joint distribution of timing and evidence labels across more than a thousand consistently coded records.
This asset supplies that derived result. Researchers can cite the 307-record Window cohort and its evidence composition without manually tallying the directory. The fixed snapshot and explicit denominators also make later comparisons possible.
That added value is aggregate only. A matrix cell does not validate the underlying startup claim. Use Projects and follow its source when discussing an individual company, revenue figure, or market.
How can readers verify the result?
- 1.Open the manifest and confirm the 1,012-record source count and 2026-09-30 snapshot.
- 2.Confirm that the CSV or JSON contains 12 rows.
- 3.Add the Evergreen cells:
39 + 389 + 259 + 18 = 705. - 4.Add the Window cells:
26 + 148 + 106 + 27 = 307. - 5.Confirm the evidence columns total 65, 537, 365, and 45.
- 6.Confirm
705 + 307 = 1,012and recalculate the 69.7% and 30.3% headline shares. - 7.Keep row shares separate from evidence-column shares.
Counts are the primary reconciliation values; percentages are rounded to one decimal place.
How should the result be interpreted?
The matrix describes the editorial composition of this dataset. It makes visible that founder-reported records are the largest evidence group in both timing cohorts, while the unproven share is larger in the Window cohort than in the Evergreen cohort.
That observation is not an explanation. The data do not show that a short-lived market causes weaker evidence, that an unproven claim makes a market temporary, or that one cohort will perform better. Selection and coding choices can affect the distribution.
Practical uses include documenting a research sample, comparing later snapshots, identifying segments for source review, and making explicit which denominator supports a percentage. How it works provides more context on the evidence framework.
What are the limitations?
The collection is curated and nonrepresentative. Its 69.7% versus 30.3% timing split is not a market estimate.
Evergreen does not mean permanent. Window of Opportunity does not establish a closing date or a guarantee of urgency. Both are editorial labels based on the record’s opportunity framing.
Evidence classes are also editorial categories. They support transparency, but they are not universal truth scores and cannot replace examination of an upstream source.
The matrix contains no outcome data, causal model, confidence intervals, or test of whether the observed distributions differ beyond this collection. It should not be used to predict revenue, durability, or startup success.
How should this matrix be cited?
Use the title, publisher, denominator, date, and exact file URL:
> ProvenStartups, “Startup Opportunity Window × Evidence Matrix,” 1,012-record snapshot dated 2026-09-30, https://provenstartups.com/datasets/2026-09-30/startup-opportunity-window-evidence-matrix.csv.
Use the JSON URL when the machine-readable JSON is the analyzed version. Cite upstream source material separately for individual startup claims.
What is the practical takeaway?
This snapshot is mostly Evergreen by its editorial labels: 705 of 1,012 records. The 307-record Window cohort has a different evidence composition, but the matrix does not say why.
Its value is reproducible classification, not prediction. Use it for aggregate statements, preserve the correct denominator, and return to project-level sources before drawing conclusions about any individual opportunity.