Fundraising Foundations
Fundraising Benchmarks: Read Medians, Percentiles and Outliers
Read fundraising benchmarks without letting an average set your plan. Work through a fictional dataset and a practical comparison checklist.
What you will learn
- Distinguish a mean from a median and percentile.
- Recognize when populations and periods are not comparable.
- Use a benchmark as context, not a fundraising target.
A fundraising benchmark describes a defined group, not what your company should raise. Before using one, check who is counted, when the data was collected and whether the statistic is a mean, median or percentile. A single unusually large round can move the average without changing the middle company.
Mean and median answer different questions
The mean adds all observations and divides by their count. The median is the middle observation after sorting; for an even count, it averages the two middle observations. NIST’s explanation of location measures shows why extreme values can pull a mean away from the center of most observations.
Neither measure supplies a target by itself. A median can hide wide dispersion, and a mean can be relevant when you want total capital divided across all recorded rounds. State which question you are answering.
Work an original five-round example
Fictional round amounts, in millions of dollars
| Company | Initial dataset | One round changes |
|---|---|---|
| A | $1 | $1 |
| B | $2 | $2 |
| C | $3 | $3 |
| D | $4 | $4 |
| E | $10 | $30 |
| Total | $20 | $40 |
| Mean | $4 | $8 |
| Median | $3 | $3 |
The mean doubles from $4 million to $8 million when the last round changes. The median remains $3 million. No other company raised more. This arithmetic explains why “average round size increased” does not necessarily mean a typical company secured twice as much capital. These are invented teaching numbers, not current market data.
What does a percentile mean?
A percentile describes position within a specified distribution. A 75th-percentile figure is a value associated with roughly three-quarters of observations at or below it, subject to the provider’s calculation convention. With small datasets, interpolation methods can differ. Check the methodology instead of treating a percentile as an exact universal boundary.
Percentiles are not success probabilities. Being above a valuation percentile does not mean a startup has that probability of raising, and a round-size percentile does not establish the capital needed to reach a milestone.
Use a benchmark-reading card
Questions to answer before citing a benchmark
| Check | What to record |
|---|---|
| Population | Stage, sector, geography, company age and inclusion rules |
| Unit | Companies, rounds, dollars, investors or survey respondents |
| Period | Announcement date, close date or observation date |
| Statistic | Mean, median, percentile, total or rate |
| Coverage | Reported-only deals, undisclosed rounds, sample size and missing values |
| Method | Currency conversion, duplicate rounds, outliers and revisions |
| Decision | Exactly how the comparison changes your next action |
Keep unlike groups apart
A seed benchmark containing priced rounds and convertible instruments may not match a priced-only table. Announced capital may differ from funds received. A global number converted at one exchange rate may differ from a local-market series. A later-stage software company is not automatically a useful comparison for a preclinical company.
Do not delete an outlier merely because it is inconvenient. Investigate whether it is a valid observation, a classification mismatch or an error. If you show both an unadjusted and adjusted result, explain the rule.
Apply the lesson to your raise
Use a current benchmark to understand the market, then build your own capital requirement from milestones, costs, timing and scenarios. Keep the Seed Fundraising Comprehensive Guide central for the round strategy and use your financial model to explain the plan.
Practice with the five-round table: remove company E entirely. The remaining mean is $2.5 million and median is $2.5 million. That is a different sample, not proof the first result was wrong. Write the population label beside each calculation.
Carry the context into conversations
Keep the benchmark source, observation period and your own assumptions alongside the investor materials in Finta. When someone asks why your planned raise differs from a market median, answer with the work the capital must accomplish. That makes the comparison useful without letting somebody else’s distribution write your plan.