Free backtest
Your meter, replayed
Send us a year of meter data. We'll show you what Qyvern would have caught on your site, and what it would have been worth.
What we're trying to do
Find out if it works on your site
Our forecast has only been tested on public buildings. The fair way to test it on yours is to go back to the start of your data and forecast one hour at a time, using only what was known at that moment. Then we compare every forecast with what your meter actually recorded, and work out what acting on it would have saved.
How it works
Four steps
-
Send a meter export
Ideally 12 months of 15-minute data, as a CSV, with the units stated.
Shorter files still run, at a lower confidence tier. Under five weeks, we won't quote an accuracy number.
-
Give us your demand charge
It's on your bill, in dollars per kW. Off-grid? Give us your fuel cost instead.
The replayed dollar figure is built on demand charges today, so a fuel figure is labelled as a projection.
-
We run it walk-forward
The model only ever sees the past. It trains, forecasts the next hour, steps forward and repeats across your whole file.
-
You get the results
- Forecast accuracy against a persistence baseline
- The peaks it would have warned about
- A dollar figure, replayed on your own meter
- A confidence level, set by how much history you sent
- Every assumption, written down
Nothing moves on site: the plan is replayed against the load your meter recorded. Time-of-use demand windows and ratchet clauses aren't modelled, and the report says which assumptions the number rests on.
Your data
What happens to your data
Your results only
Your file is used to produce your backtest and nothing else.
Nothing connects
You send a file. Nothing connects to your systems, and we never log in to them.
Deleted on request
Ask, and we delete it.
NDA available
Ask for one before you send anything.
Contact
Request a backtest
Or ask about a design partnership, or anything else. We reply by email.