All articles Advanced signals

Device Fingerprinting Pricing: What You Pay Per Call

Developer working at night at a dark desk, screen glow in electric blue, abstract interface shapes without text

Device fingerprinting pricing confuses teams because providers bill a unit that never appears in the product: the identification. You ship an agent, it returns a visitorId, and a counter increments somewhere. Everything else — your bill, your forecast, your conversation with finance — follows from how often that happens and what the provider counts as one call.

This guide breaks the cost model into parts you can budget: what counts as one identification, how per-call and per-MAU plans behave under the same traffic, and the two engineering habits that shrink the number — caching visitorId on your server and identifying only where a decision is made. Where you see arithmetic below, treat it as a worksheet: plug in your traffic and your own rate.

What counts as one identification

One identification is normally one successful response that returns a visitorId for a browser session — not a page view, not a user, not a month of activity, but a single request that produced a result.

That matters because the trigger lives in your code. A snippet in a global layout fires on every marketing page view, prefetch, and QA reload. A call inside your signup and login handlers tracks the events you actually care about.

Read the rate card for these five things

  • Do requests that error, time out, or return no visitorId count?
  • Does a retry count twice?
  • Do Smart Signals bill separately from plain identification?
  • Is there a monthly floor or a seat fee?
  • What happens above your plan: overage rate, throttling, or a hard stop?

Per-call and per-MAU behave differently on the same traffic

Risk analyst in a dark office pointing at a single blue node on a dark abstract panel

Per-call

You pay for what you run. Cost tracks product usage: a signup spike costs more, a quiet month costs less. The risk is instrumentation — a loop, an uncached widget, or a bot hammering your login form all become line items. Per-call rewards teams that review their own code.

Per-MAU

Cost tracks the size of your user base, not the number of decisions. Forecasts are simple. The drawback: the bill grows even when you identify nobody, and it grows fastest across free-tier and trial populations — the accounts you most want to inspect and the ones that may never convert. For products with a large free base, per-MAU charges you for the abuse you are trying to stop.

If identification fires on a handful of high-value events (signup, login, checkout, trial start), per-call gives you a lever. If it fires on nearly every authenticated request, fix the instrumentation before you pick a model.

Do the arithmetic with your own traffic

Fill in five lines for one product surface:

  1. Monthly sessions on the identifying flows.
  2. Calls per session today, read from network logs rather than guessed.
  3. Cache hit rate you expect after the next section.
  4. Billable calls = sessions × calls per session × (1 − cache hit rate).
  5. Monthly cost = billable calls × your quoted per-call rate.

The shape of it, with illustrative inputs: 40,000 monthly sessions, the agent mounted on 6 pages per session, no cache — 240,000 calls. Add a 90% cache hit rate and roughly 24,000 calls reach the meter. Same funnel, same signals on the decisions, a tenth of the volume. Substitute your own rate.

Session count comes from analytics, calls per session from a code review, and cache hit rate from a week of measurement.

Cache the visitorId on your server

The cheapest identification is the one you already made. A visitorId stays stable for a browser profile, so store it once and reuse it.

Product engineer sketching two abstract line shapes on a glass wall in a dim meeting room

  • Return the visitorId in an HttpOnly cookie scoped to your domain.
  • Store it server-side beside the session or account, with the timestamp of the last identification.
  • Read the cached value on later requests, and re-identify when the cookie is missing, the device is new, or the action is sensitive: password change, payout change, new card.
  • Keep the secret key on the server; it never ships to the browser.
// pseudo-code, server-side
const { visitorId, identifiedAt } = req.session;
const stale = Date.now() - identifiedAt > CACHE_MS;
if (!visitorId || stale) {
  const res = await identify(req);      // one billable call
  req.session.visitorId = res.visitorId;
  req.session.identifiedAt = Date.now();
}
decide(req.session.visitorId, req.body.action);

Set CACHE_MS from your risk policy, not your price list: hours for read-only pages, minutes for money movement. Treat a cleared cookie as a new device in your rules instead of assuming continuity.

Identify where a decision happens

Write down your decisions — allow, challenge, block — and put identification on those paths only: signup, login, password reset, trial activation, promo redemption, checkout, payout change.

The usual waste: a global layout script that fires on blog and pricing pages, client-side and server-side identification for the same event, a call per form field, webhooks and health checks booting the agent, and no rate limit on the signup endpoint — so a bot script bills you for every attempt it makes. A per-IP and per-device ceiling placed before the call is a cost control and a risk control at once.

How Prynt’s plans fit this model

Prynt bills the identification, and the free tier covers 5,000 identifications per month without Smart Signals — enough to instrument one funnel and measure your real cache hit rate before you commit. Pro is $29/month (100,000 identifications with Smart Signals) and Scale is $99/month (250,000). Verification runs server-side with your secret key, and a request you answer from a cached visitorId never has to leave your infrastructure. Your metered volume should track decisions, not page views.

Developer tracing a fiber cable to a glowing blue port in a dark server aisle

Try it free at pryntid.com: run identification on signup and login for a week, read the count, then set your cache and your policy.

FAQ

Does every page view cost an identification?

Only if your code calls identification there. Move the call into the event handlers and marketing traffic stops touching the meter.

Can I make decisions with a cached visitorId?

Not for everything. Repeat evaluations of a known browser profile are fine; re-identify when the cookie is missing, on a new device, or before money movement and credential changes.

How do I budget for a launch spike?

Model expected sessions × calls per session × (1 − cache hit rate), add margin for retries and bots, rate-limit the identifying endpoints, and alert when daily calls exceed your model by a set percentage.

Try it free

Prynt is device intelligence with a free plan: 5,000 identifications a month, visitor IDs with confidence and bot detection — no credit card. Upgrade to Pro for VPN, proxy, tampering and behavioral signals.

Keep reading