Pricing
Our catalogue and cost base, how the price floor works and why it sits where it sits.
Our base catalogue
| Plan | ₽/mo | $/mo | Quota | Speed cap |
|---|---|---|---|---|
| base | 299 | 3.99 | 300 GB | 25 Mbit/s |
| pro | 799 | 9.99 | 800 GB | 50 Mbit/s |
| dedicated | 2,490 | 29.90 | 2,500 GB | 100 Mbit/s |
Annual prices run about two months cheaper than monthly. Your prices are yours; this is the starting point and the base the floor is computed from.
Where the floor comes from
A node costs money, and a plan's traffic is direct cost of goods. The quotas in the catalogue are set so traffic consumes no more than roughly 30% of a plan's price; the floor is the price below which a plan stops paying for the traffic it authorises.
A practical reference point: the average subscriber across our fleet uses about 11 GB a month, so the base plan's quota is an anti-abuse ceiling rather than something an ordinary user runs into. That is exactly why discounting is dangerous: the price drops immediately, and the heavy users show up a month later.
The API will not accept a price below the floor — it returns price_below_floor and the minimum itself, see Your prices. The floor is agreed at onboarding and revisited when the cost base moves.
There are no geo multipliers
The system has no automatic per-region multipliers — do not plan around them. What exists is a separate price for each plan-and-currency pair. That is how you price a rouble market differently from a dollar one.
On churn
We have no churn statistics of our own across partner projects yet, and we are not going to pass other people's benchmarks off as ours. What our own support work does show, and what is worth building into your process:
- The first week decides everything. A user who failed to connect on the first try with nobody to ask does not come back.
- An unannounced charge reads as a scam even when it is exactly per the terms. A reminder a couple of days before renewal is cheap.
- The specific pain in our markets is throttling rather than a blackout: the user "has internet", it is just slow. That is churn too, and the cure is node selection, not support.