How we define, collect, and normalize benchmark data

1. Overview & philosophy

BenchmarkHQ aggregates SaaS benchmark data from 15+ published industry research sources and normalizes it into a single, filterable dataset segmented by ARR band, GTM motion, ACV, and vertical.

We are not a primary data collector. We do not survey companies directly. Every data point in our system traces back to a named external source — you can always see which sources inform each metric. Our value is in synthesis: taking data that exists across dozens of reports, normalizing the definitions, resolving conflicts, and making it filterable in ways no single source allows.

We target the $1–20M ARR B2B SaaS window because most free public benchmarks (VC-published, analyst reports) target $10M–$100M+ ARR companies. The dynamics at $1–20M ARR are meaningfully different — higher churn rates, longer CAC payback periods, lower NRR — and conflating them with mature-company benchmarks leads to bad target-setting.

2. Data sources

The following sources contribute data to BenchmarkHQ reports. Each source has a different focus, methodology, and company sample — which is why synthesis is necessary.

3. Data inclusion criteria

Not all data from source reports is included in BenchmarkHQ. We apply the following inclusion rules to ensure data quality and relevance.

Company type

Only B2B SaaS companies are included. Consumer SaaS, marketplace businesses, hardware/software hybrids, and transactional businesses (even if software-based) are excluded. When source reports don't segment by business model, we apply our best estimate of the B2B SaaS portion based on the source's described methodology.

ARR band eligibility

Data points are only included in an ARR band if the source explicitly segments by ARR range or provides sufficient disaggregation to infer band-level benchmarks. We do not extrapolate overall benchmarks (e.g., "all ARR ranges") into specific bands.

Sample size minimum

We require a minimum sample size of n ≥ 25 for a data point to be reported. Data points with n < 25 are suppressed and marked as "insufficient sample." We report sample sizes in all exports so you can weight data points appropriately.

Recency

For our quarterly reports, we include data published within the past 18 months. Older data is retained in our historical archive but not included in current benchmark calculations. This prevents stale data from diluting current benchmarks.

4. Metric definitions & formulas

Different sources define the same metric differently. Our definitions are documented below. When a source uses a different definition, we note how we adjusted its data to conform to our standard.

Core formulas are shown here; additional metric notes appear in the glossary, report footnotes, and export metadata.

5. Normalization & reconciliation

When multiple sources report the same metric for the same ARR band, they often produce different results. This section explains how we handle conflicts.

Weighted averaging

When sources agree within ±5 percentage points (or ±5 units for non-percentage metrics), we report a weighted average. Weights are assigned based on: (1) sample size (larger sample → higher weight), (2) recency (newer data → higher weight), and (3) methodology similarity (how closely the source's definition matches our standard).

Conflict resolution

When sources disagree by more than ±5 percentage points, we apply the following resolution hierarchy:

1. Definition mismatch check — If the conflict stems from definitional differences (e.g., one source includes involuntary churn, another doesn't), we adjust the outlier source to match our standard definition and re-evaluate.
2. Sample bias check — If one source has a materially different company profile (e.g., heavy enterprise bias vs. SMB-heavy), we apply a correction or flag the source as a separate data point.
3. Credibility weighting — If unresolvable, we weight toward the source with the larger sample size and more transparent methodology. We document which source was de-weighted and why in the report footnotes.
4. Disclosure — If sources remain in conflict after the above steps, we report both values with a note explaining the discrepancy rather than presenting a false consensus.

Percentile reporting

We report p25 (bottom quartile), p50 (median), and p75 (top quartile) rather than averages. This matters because SaaS metric distributions are typically right-skewed — the mean is pulled up by outliers and misrepresents the typical company's experience. The median is a more reliable "what does a normal company look like" signal.

6. Update cadence

BenchmarkHQ publishes a new benchmark report each quarter. Here's how the update cycle works:

Annual sources (survey-based reports) are typically published between January and April each year. When a major annual source publishes new data, we issue a supplemental update to our database, noting which benchmarks changed and by how much.

7. Limitations & caveats

We believe transparency about limitations is more valuable than projecting false confidence. Read these carefully before using the data to make decisions.

We are not a primary data source

BenchmarkHQ does not collect data directly from companies. We depend on the accuracy and representativeness of our source reports. If a source has a selection bias (e.g., companies that use a particular billing platform tend to have higher NRR), that bias may be present in our data.

Survivorship bias

Most benchmark sources survey or measure companies that are still operating. Companies that churned or shut down between data collection and publication are typically excluded. This means benchmarks likely overstate how well "typical" companies do, especially at early ARR stages where failure rates are higher.

Geographic concentration

The majority of our data sources are US-centric. Non-US SaaS companies face different pricing environments, CAC structures, and growth dynamics. We label data as "US-weighted" where relevant and note when a source has meaningful international representation.

Definition drift

Metric definitions evolve over time. What counts as "CAC" has shifted as attribution models have matured. We document our current definitions but acknowledge that year-over-year comparisons may reflect definitional changes as much as actual performance shifts.

Not financial advice

Benchmark data describes what companies have achieved historically, under varying market conditions. It is not a guarantee of what's achievable or appropriate for your company. Use it as directional context, not as a hard target. Discuss with your investors and advisors before setting formal goals.

8. Source taxonomy

Sources in the BenchmarkHQ dataset fall into three distinct categories. Understanding the difference matters for interpreting metric confidence and freshness. Every benchmark displayed in reports and exports is tagged with its source type.

Tier 1 — Current recurring sources

These sources publish on a regular cadence (annual or quarterly) and are actively updated in our database each cycle. They form the backbone of every benchmark report.

Tier 2 — Historical benchmark series

These sources have published in the past and contributed data to earlier reports. Their data is preserved in our historical trend database to enable QoQ and YoY comparisons, but they are not actively updated in new cycles. Historical data from these sources is labeled with the year it was collected.

Tier 3 — Reference & comparison sources

These sources are used for spot-checks, cross-referencing, and contextual notes in reports — but are not primary data contributors to calculated benchmarks. They appear in methodology notes and source disclosures when relevant.