The Panoply Score is a number out of 100 for every tool in the directory. It answers one question from the buyer’s side: how good a bet is this tool, on the evidence we can check? It is not a measure of market share, and nobody can pay to change it.
The six parts
| Part | Points | What it measures |
|---|---|---|
| Pricing honesty | 20 | Whether prices are published (8), whether there is a usable free option (5), whether you can buy without a sales call (4), and whether the bill is predictable (3). |
| Adoption evidence | 25 | Public, verifiable usage: GitHub stars, weekly package downloads, Docker pulls, Stack Overflow questions, Wikipedia readership and review volume on G2 and Gartner Peer Insights. Each is placed on a fixed logarithmic scale, and the tool gets the average of the ones it has. |
| Momentum | 15 | Release cadence and commit activity of the code that is the product. Full marks at one release a week and 5,000 commits a quarter. |
| Openness and exit | 15 | Open-source licence (5), self-hosting (4), whether your data stays in open formats or your own warehouse (3), and how hard it is to leave (3). |
| Practitioner sentiment | 10 | Ratings from G2 and Gartner Peer Insights, weighted by review count and pulled towards the average when there are few reviews. |
| Editorial verdict | 15 | Our own judgement after researching the tool, signed and given with a one-line reason on every tool page. |
Rules we hold ourselves to
- Every point has a visible reason. Each tool page shows all six parts, the numbers behind them and links to the sources.
- Missing is not the same as bad. If a part cannot be measured for a tool, it is marked not applicable and the total is rescaled over the parts that apply. A closed-source product is not marked down for having no public code.
- Fixed scales, not rankings against each other. Adoption and momentum use fixed scales, so a tool’s score does not move because we added a competitor.
- Automated activity earns nothing extra. Release cadence is capped at one a week, and releases of a client SDK for a closed product do not count as product momentum.
- Review scores carry a small weight on purpose. They cluster between 4.2 and 4.7 for most tools, and vendors run campaigns to raise them, so they separate tools less than people assume.
- No money changes a score. Sponsors and affiliate partners get no say in any part of it.
What the score favours
It is written from the buyer’s side, so it rewards tools that publish their prices, let you start for free and are easy to leave. A closed, sales-led product with excellent engineering can score lower than a smaller open-source tool. That is deliberate. If lock-in and procurement do not worry you, read the part scores rather than the total.
Current scores
| Tool | Score | Evidence sources |
|---|---|---|
| Metabase | 84 | 7 |
| dbt | 83 | 9 |
| Apache Airflow | 77 | 8 |
| BigQuery | 74 | 7 |
| Airbyte | 72 | 10 |
| Dagster | 72 | 6 |
| Databricks | 69 | 8 |
| Snowflake | 60 | 9 |
| Fivetran | 57 | 4 |
| Hightouch | 55 | 3 |
| Monte Carlo | 40 | 4 |
| Looker | 39 | 6 |
How often it changes
Public signals are collected again every month, and prices are re-checked whenever a vendor changes its pricing page. Each tool page shows the date its score was computed and the method version. Method changes are noted here.
v1.0, 23 September 2026: first version. Same day: Fivetran’s openness input corrected after our research pass found its Hybrid Deployment option (score 54 to 57).
Disagree with a score?
Tell us which input is wrong and where the correct figure is published. Vendors can correct facts; they cannot buy points. Write to us from the About page.