Tier method
How the tier scores work
This page describes what our scoring script does today, method 1.0-draft. The weights and limits here are read from the input file the script uses; the point values are set in the script.
Criteria and weights
Each tool gets one score from 0 to 100 for a use case, from 7 criteria. Each criterion is scored from 0 to 100 and counts for its weight below. The weights are the same for every tool.
| Criterion | Weight | What it uses |
|---|---|---|
| Value for money | 38 | Price of the cheapest standard plan, billed monthly, that includes 1,000,000 characters a month. The same volume and billing for every vendor. |
| User sentiment | 16 | Overall star ratings from review platforms we have permission to use (0 of 3 today). |
| Output quality | 10 | Sub-ratings from the same platforms, when shown there. |
| Ease of use | 7 | Sub-ratings from the same platforms, when shown there. |
| Reliability and support | 7 | Sub-ratings from the same platforms, when shown there. |
| Free plan and entry price | 15 | Points from our price data, set out below. |
| Pricing transparency | 7 | Points from our price data, set out below. |
User sentiment, output quality, ease of use, and reliability and support use review platforms. Together they count for at most 40% of the weight (today 40%), and the build fails if the weights go over that or don't add up to 100.
A criterion with no data scores 0; the other weights are never scaled up to fill the gap. On the tier list it shows "not found".
Value for money: for each tool we take the price of its cheapest standard plan (not a premium-quality one), billed monthly, that includes 1,000,000 characters a month. The score is the lowest such price among the tools divided by the tool's price, times 100. Pay-as-you-go rates are not plans, so they don't count. A tool without such a plan in our price data shows "not found" and scores 0, never a guess.
Review data
- The script accepts review data from G2, Capterra and Product Hunt only, and each platform has its own permission switch. A platform's ratings and review counts are used only once its switch is on, which we do only with that platform's written permission. Today 0 of 3 are on.
- While a platform is off, the build fails if its ratings or counts are entered. We may link to its review page so you can read the reviews there.
- When a platform is on, each rating is pulled toward 3.5 stars as if it had 10 extra reviews, so small review counts move the score less. Platforms are then averaged, weighted by the square root of their review counts, and stars become a 0 to 100 score.
- Sub-ratings for output quality, ease of use and reliability count only with 10+ reviews from the last 24 months.
adjusted = (10 × 3.5 + n × rating) ÷ (10 + n)
score = (combined stars − 1) ÷ 4 × 100
Price criteria
Prices come from our dated checks of each vendor's pricing page, and each one links to that page. List prices only: promos, coupons and credits are never treated as prices.
Free plan and entry price (weight 15)
- 40 points for a free plan, or 20 for free credits or a free trial.
- 20 more if there is a free plan, credit or trial and our data shows it allows commercial use.
- Up to 40 more for the lowest monthly-billed paid plan: 40 at $10 or less, 30 at $20 or less, 20 at $30 or less, 10 at $50 or less.
Pricing transparency (weight 7)
- 40 points if our latest check confirmed the listed prices in full (see How we verify), or 20 if only in part.
- 20 if every paid plan shows a public price.
- 20 if our data has the vendor's stated rate per unit of output, such as an API price per 1,000 characters or credits per character.
- 20 if we have 90+ days of price history and no price change in the last 90 days.
Tiers and Provisional
| Tier | Score |
|---|---|
| S | 85 to 100 |
| A | 75 to 84.9 |
| B | 65 to 74.9 |
| C | 55 to 64.9 |
| D | Below 55 |
| Provisional | Not enough evidence yet |
The script marks a tool Provisional, with the reasons and no tier letter, when any of these is true:
- Less than 80% of the weight has data.
- Fewer than 2 permitted review platforms have 10+ reviews each.
- Fewer than 30 reviews from the last 12 months across those platforms.
- The price check is older than 30 days at our latest data check.
A Provisional tool still shows its score so far, where missing data counts 0, unless its price check is too old: then the score is paused until the price is re-checked. Provisional tools are listed A to Z, not by score.
Tools with a tier are ordered by score; ties are ordered A to Z.
Paid links
- We earn a commission when you sign up through a paid link. It never affects our data, scores or order.
- The scoring script reads a fixed list of tool fields, and that list has no paid-link data.
- On every build, the scores are worked out again with each tool's paid-link status removed and with it switched. The build fails if any score or the order changes.
- Every tool gets the same style of "Go to" link. Paid ones are labeled Paid link next to it. See our affiliate disclosure.
Checking our work
- Scores are computed by a script from the input files when the site is built. There is no field to set a score by hand: the build fails if the tier input file has one.
- Each tier list shows the sha256 hash of the script's inputs and of its output. The inputs and output for the voiceover list are at /tiers/voiceover.json.
- The date used for score checks is our latest data check, , not the build date.