Startup Opportunity Window Benchmark: Evergreen vs Time-Sensitive Models
Download the evergreen-versus-window classification for 406 curated startup cases across 13 business-model categories.
This benchmark finds a near-even split between evergreen and time-sensitive startup models in a descriptive convenience sample of 406 records: 225 are labeled evergreen (55.4%), while 181 are labeled as having an opportunity window (44.6%). The result is a category snapshot, not a prediction of startup success.
ProvenStartups is the Organization author and publisher of this analysis.
Where should you start?
What does the benchmark show?
The benchmark compares startup categories using an editorial binary timing label:
- ·Evergreen: the opportunity is described as relatively durable.
- ·Window: the opportunity is associated with a more time-sensitive opening.
The labels do not measure company survival, revenue duration, market longevity, or realized outcomes. They describe how each opportunity is classified in the source dataset.
Across all 406 records, evergreen models represent a modest majority. That overall result hides substantial variation by category. Digital Publishing has the highest evergreen share at 80.6%. Cautionary Tale has the highest window share at 73.7%, followed by AI Content at 72.0%.
The benchmark therefore suggests that timing orientation varies more by category than the overall split alone would imply.
How are the categories distributed?
The table below reports each category’s total records, evergreen records, window records, and evergreen percentage.
| Category | Total | Evergreen | Window | Evergreen |
|---|---|---|---|---|
| AI Content | 25 | 7 | 18 | 28.0% |
| AI E-commerce | 10 | 3 | 7 | 30.0% |
| AI Service | 48 | 19 | 29 | 39.6% |
| AI Website | 15 | 8 | 7 | 53.3% |
| Consumer App | 60 | 34 | 26 | 56.7% |
| Digital Publishing | 31 | 25 | 6 | 80.6% |
| Directory Site | 12 | 9 | 3 | 75.0% |
| Ecosystem Tool | 9 | 4 | 5 | 44.4% |
| Platform Plugin | 14 | 9 | 5 | 64.3% |
| SaaS | 93 | 58 | 35 | 62.4% |
| Scale Reference | 36 | 25 | 11 | 69.4% |
| Simple Tool | 15 | 11 | 4 | 73.3% |
| Cautionary Tale | 38 | 10 | 28 | 26.3% |
| Overall | 406 | 225 | 181 | 55.4% |
SaaS is the largest category, with 93 records. Consumer App follows with 60, while AI Service contains 48. These larger groups contribute materially to the overall result, but they do not determine the pattern for smaller categories.
Among categories with an evergreen majority, Digital Publishing leads at 80.6%, followed by Directory Site at 75.0%, Simple Tool at 73.3%, Scale Reference at 69.4%, Platform Plugin at 64.3%, SaaS at 62.4%, Consumer App at 56.7%, and AI Website at 53.3%.
The remaining categories have more window-labeled records than evergreen records: Ecosystem Tool at 44.4% evergreen, AI Service at 39.6%, AI E-commerce at 30.0%, AI Content at 28.0%, and Cautionary Tale at 26.3%.
What does “evergreen” mean in this dataset?
“Evergreen” is a descriptive timing category, not a claim that a business will remain viable indefinitely. It indicates that the opportunity was labeled as less dependent on a narrow or temporary market opening.
The category should be read comparatively. For example, Digital Publishing has 25 evergreen records and 6 window records, producing an evergreen share of 80.6%. That does not establish that digital publishing startups are more likely to succeed. It only shows how the records in this snapshot were classified.
The same principle applies to the window label. Cautionary Tale has 28 window records out of 38, or 73.7%. This indicates a strong concentration of time-sensitive labels within that category; it does not measure failure risk or predict when a company will stop operating.
How was the benchmark calculated?
The snapshot date is 2026-09-22, and the sample size is n=406.
The calculation groups records by category and timing prefix. For each category, the benchmark counts:
- 1.Total records.
- 2.Records labeled evergreen.
- 3.Records labeled window.
- 4.Evergreen records divided by total records.
The overall percentages use the same approach:
- ·Evergreen: 225 ÷ 406 = 55.4%.
- ·Window: 181 ÷ 406 = 44.6%.
The category counts are not weighted to equalize category size. A category with 93 records therefore contributes more observations to the overall totals than a category with 9 records.
The timing field is an editorial binary label. The benchmark does not infer timing from company age, funding, revenue, survival, or other observed events.
What are the main limitations?
This is a descriptive convenience sample, not a success forecast. The results should not be used to rank startup ideas by expected returns or to estimate the probability that a company will succeed.
The sample has unequal category sizes. Some categories contain many more records than others, which affects how strongly they contribute to the overall 55.4% evergreen share. Smaller categories can also show large percentage changes from relatively few records.
The timing classification is editorial and binary. Real opportunities can combine durable and time-sensitive characteristics, but this benchmark assigns each record to one of two timing prefixes. The labels should therefore be treated as a simplifying classification.
The data does not include time-to-event measurements. There is no observed duration showing how long an opportunity remained available, how long a company survived, or when a market condition changed.
The benchmark also does not establish causality. It cannot show that a category causes an opportunity to be evergreen or time-sensitive. It only reports the distribution of labels in the snapshot.
Finally, category names are descriptive groupings. They should not be treated as mutually exclusive economic theories or as a complete taxonomy of startup models.
How can you verify the result?
The category-level data is available in two formats:
The snapshot manifest is available here:
The public files verify the published category rows and manifest. Full regeneration requires the 406 coded source records: group them by category and timing prefix, count evergreen and window records, calculate each category's evergreen percentage, and then calculate the overall totals. Preserve the editorial labels as coded; do not reinterpret them as measured survival outcomes.
For a citation, reference the dataset snapshot dated 2026-09-22 and include the manifest link alongside the CSV or JSON file used.
Where can you explore the broader collection?
You can browse the broader ProvenStartups projects collection and review the organization’s method and process.
For a related category-level analysis, see AI Startup Category Benchmark.
Frequently Asked Questions
Is this benchmark a prediction of startup success?
No. It is a descriptive convenience sample organized by category and an editorial binary timing label. It is explicitly not a success forecast.
Does “evergreen” mean an opportunity will last forever?
No. Evergreen means the record was classified as relatively durable compared with a time-sensitive opportunity. The label does not measure market duration or company survival.
Why do category percentages differ from the overall percentage?
Category sizes are unequal. The overall percentage is calculated from all 406 records, so larger categories contribute more records to the total than smaller categories.
Can the data prove that one category causes better outcomes?
No. The dataset contains no causal analysis and no time-to-event measurements. It reports label distributions in the 2026-09-22 snapshot.