Subscription sharing is the quiet reason a publisher’s subscriber count and reader count stop matching. For subscription sharing, publishers usually reach for a hard device limit first, and the result is angry emails from families plus very little effect on the accounts being resold. The teams that handle it well treat it as a measurement problem before an enforcement problem: they count reads per device and per account, separate a household from a resale ring, and price the difference.
The difficulty is that sharing covers three different behaviors. A subscriber on a phone, a tablet and a work laptop is normal. A family passing one login around four devices is against most terms but easy to convert. One account serving dozens of readers in different cities is resale, and it costs you seats you never sold. A cookie-based meter cannot tell these apart, and an IP address barely helps: mobile carriers put thousands of readers behind one address.
So the order is: define the buckets, instrument reads, review the pattern, then decide per event. Prynt Control exists for the last step, and it is only as good as the two before it.
Define the buckets before you enforce anything
Write down what you will do in each case, because the policy decides which signals matter.
- Household. A few devices, mostly one network, human hours. Decision: allow, maybe offer a second seat.
- Casual sharing. More devices than seats, mixed networks, a plausible amount of reading per device. Decision: step-up, then a seat offer.
- Resale. Many devices, little overlap between them, logins spread across time zones, the same articles read repeatedly. Decision: block the login and contact the account holder.
Prynt Control answers each event — signup, free_trial, bonus, login, checkout, extra_seat, sms, paywall — with allow, step-up or block, a reason code and a suggested step (captcha, verify_email, card_or_passkey). That structure keeps the policy in one place instead of scattering if statements through the paywall. Control is its own plan, and it ships in mirror mode first.
Meter reads per device and per account
The paywall meter in Control counts per device and per account, and an incognito window does not reset it. Most “we can’t see sharing” stories are really “our meter lives in local storage.” Device-level identity is the foundation; see how our device fingerprinting works and what it does not claim.

Log, for every read you might later act on:
- the account ID and the plan’s seat count;
- the device identity attached to the request;
- a timestamp and the article or section;
- the network signals you already receive (VPN, proxy, datacenter, Tor).
With a week of that table you can see how many accounts sit above one device, how far above, and whether those devices look like one person’s day or a distribution channel.
Separate a household from a resale ring
No single signal separates them, and any product claiming otherwise is selling certainty it does not have. Combine them and read the pattern:
- Device count over time. Families add devices slowly and keep them; resale adds in bursts and drops them.
- Network spread. Household devices cluster on one or two networks; resale devices arrive from many, often in different countries, and the VPN and proxy signals around them are noisier.
- Concurrency. Two articles at once on one tablet is normal; twenty simultaneous reads from twenty networks is not.
- Payment and email reuse. A card or address that appears across many accounts says more than any browser detail.
- Behavior after a limit. Households write in; resale accounts create a new login within minutes.
Two or three together justify a decision. One alone justifies watching.
Decide per event, not per account
Decisions belong at the moment of access, so the reader gets a next step rather than a locked door. In pseudo-code:

const result = await control.decide({
useCase: "paywall",
accountId: account.id,
seats: account.seats
});
if (result.decision === "allow") {
serveArticle();
} else if (result.decision === "step-up") {
// result.suggestedStep: captcha | verify_email | card_or_passkey
showStep(result.suggestedStep, result.reason);
} else {
showSeatOffer(result.reason);
}
Prynt never blocks on its own. It returns a decision and a reason code and your code applies it, because only you know whether a fourth device is a violation or a sale waiting to happen.
Turn sharing into extra seats
When an account regularly goes over its seats but the behavior looks domestic, the honest answer is not a block; it is a second seat at a fair price. Send the extra_seat use case instead of the paywall one, step up to card_or_passkey, and show the price per seat next to the list of active devices. The reader removes the device they no longer use, or pays for the extra.
Two rules keep it from feeling punitive: ask once per billing period, and never step up a device the account has used for months. New devices are the ones worth a question.
Watch it before you enforce it
Control starts in a 14-day mirror period. Nothing is blocked, and every decision is logged with its reason code. Use those two weeks to read the reasons, tune thresholds and check how many of your largest accounts you would have stepped up. A few false positives in week one are normal; a policy that fires on your most valuable subscribers is a pricing problem wearing a security costume.

After that, review weekly: step-ups, how many became seat purchases, and the complaint rate. If step-ups rise and purchases do not, the offer is wrong, not the detection.
FAQ
Does an incognito window reset the paywall meter?
No. Control meters per device and per account, so a private window does not wipe the count. Readers who test the limit will find it holds.
What is the difference between step-up and block?
step-up asks the reader to prove something — captcha, verify_email or card_or_passkey — and lets them continue. block denies the event. Both arrive with a reason code, and your code decides which one it applies.
Can we tell a family from a reseller without false positives?
Not with certainty, and you should not try. Combine device count, network spread, concurrency and payment reuse, then choose a step-up path that a household can pass easily and a reseller cannot. Keep the block path for repeated, clear patterns.
Pick one account tier, log the meter for two weeks and read the reasons before you change anything. The docs show the request and response shapes for each use case, and you can see the whole suite at pryntid.com.
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.