Learn More
How Judoscale Estimates Autoscaling Savings
Judoscale estimates how much you save by autoscaling versus running at your peak scale all day. It’s a ballpark figure — we have no way to know exactly where you’d land if you were scaling manually — but it’s the most honest comparison we can make.
How we calculate it
For each process on each day, we calculate two numbers.
Estimated compute cost
This is what your instances likely cost that day: the price of each instance multiplied by the average scale — how many were running over 24 hours.
We price an instance from whatever we know about it:
- Heroku, Render, and Scalingo — the published list price for that size (dyno type, instance plan, or container size).
- Fly, AWS ECS, Northflank, and custom platforms — an estimate from the vCPU and memory on the process, using typical public rates. Those platforms don’t publish a single price per named size the same way.
- Railway — we don’t estimate dollar costs. Railway billing is usage-based and varies too much for an honest daily figure.
This is the Compute cost column on your usage page.
Peak-scale cost
This is what you would have spent if you ran at your peak scale all day, using the same per-instance price. In other words, if you were manually scaling and wanted to avoid any slowdowns or queue backups, this would be your estimated cost.
Estimated savings
The difference between those two numbers:
Estimated savings = peak-scale cost − estimated compute cost
We only count savings on days with autoscaling activity.
An example
The math is the same on every platform we can price. On Heroku, say your web process runs on Standard-1X dynos ($25/month) and, over a 30-day month, averages 1.5 dynos with a peak of 3:
| Metric | Calculation | Cost |
|---|---|---|
| Estimated compute cost | 1.5 dynos × $25 / 30 days | $1.25 |
| Peak-scale cost | 3 dynos × $25 / 30 days | $2.50 |
| Estimated savings | $2.50 − $1.25 | $1.25 |
That’s one process, one day. Monthly emails add these daily figures up across your processes.
Where to see it
- Your usage page in Judoscale has the daily breakdown for each process — scale, size, autoscale events, compute cost, peak-scale cost, and estimated savings. The CSV download at the bottom of the page matches the table.
- Monthly summary emails roll the daily dollar savings up for the previous month, for platforms we can price.
Why it’s an estimate
We can’t know how you would have scaled manually. Some teams would have stayed at peak 24/7. Others would have done some manual scaling and landed somewhere in between. Peak-scale cost is the conservative comparison: it’s what you’d pay to never risk a slowdown.
These numbers are compute costs only. They don’t subtract what you pay for Judoscale.