Fighting Multi-Accounting: The Platform View

It helps to look at the problem from the other side: understanding what a platform sees explains why some measures work and others do not.

The signals a platform sees

Four groups, and they are not equal.

Network. The exit address, its ownership (hosting or home provider), geolocation, and the activity history of that address. Determined instantly from public data - see residential versus datacentre.

Environment. Browser and system characteristics: resolution, fonts, graphics rendering quirks, language, time zone. Together they form a stable fingerprint - see how fingerprinting works.

Behaviour. Action speed, active hours, paths through the site, the character of movement and input. It requires accumulated data and is hard to fake.

Data. Email, phone, payment details, delivery address, name. Matches here are a direct link.

What is genuinely caught

In descending order of ease:

Data matches. One card across five accounts is an ordinary database query requiring no analytics. The cheapest and most reliable method, which is why it is the primary one.
A shared address. Also cheap and obvious, but easily explained by a shared office or provider network, so not proof on its own.
Matching fingerprints. A fully identical fingerprint across several accounts is unlikely by chance.
An anomalous fingerprint. The paradox: an environment that is too unique is more suspicious than a typical one. A combination of characteristics no ordinary user has is itself a signal.
Behavioural patterns. Registrations at even intervals, identical paths, actions faster than a human.

The key point: decisions are made on the combination. One signal gives a probability; three overlapping ones give confidence.

What is not caught

Honestly about the limits:

Carefully separated accounts with different data, addresses, environments and behaviour. Automated systems do not find those; it takes a person and a reason to look.
A small number of accounts. Mass schemes are visible by scale; two or three accounts do not reach that scale.
Accounts behaving like ordinary users. If the activity is indistinguishable from normal, there is nothing to distinguish.

Hence the conclusion for platforms: complete prevention is impossible and is not the goal. The goal is making duplicates cost more than they yield. Measures that turn away honest users cost more than the problem does.

What follows for both sides

For a platform:

Start with the data. Matching payment and contact details tell you more than sophisticated technical analytics, at lower cost.
Do not ban on one signal. A shared address is also an office, a dormitory, a mobile carrier.
Look at consequences rather than the fact. Several accounts are not inherently harmful; specific behaviour is.
Leave a recovery path. False positives are inevitable, and without an appeals process you lose honest users.

For anyone running several accounts:

Data matters more than technique. Perfect technical setup with one card across all of them is pointless - see multi-accounting rules.
A typical environment beats a unique one. The goal is blending in, not hiding.
Behaviour outweighs fingerprint over time.
Scale is the main risk. The more accounts, the higher the chance of landing in a manual review sample.

What actually triggers bans in practice: why accounts get banned. The overview is in the proxy guide.

FAQ

How do platforms detect multi-accounting?

By combining signals rather than relying on one: network data, environment fingerprint, behaviour and, above all, matches in the data - payment details, phones, delivery addresses. No single signal is proof, but overlapping ones are telling.

What is easiest to catch?

Data matches: one card, one phone, one delivery address across different accounts. That requires no analytics at all, just a database query. Technical signals are harder, produce probabilities, and therefore serve as supporting evidence.

Can multi-accounting be prevented entirely?

No, and that is not the goal. The goal is making duplicate accounts cost more than they are worth. Absolute prevention would require registration checks strict enough to turn away honest users, which costs more than the problem.