How to tell a returning customer from a duplicate account
Email matching refuses real customers and misses real duplicates. Why that happens, which basic checks break, and a fair policy for people who come back.
Standing in the kitchen on the 1st, you keep scrolling back up to the total on your model provider's invoice, and it keeps saying $4,216.40. You open your laptop and pull last month's trial list, because that number is higher than it should be.
This is a scenario, so you can check every figure. Say 410 trials started last month and each one cost $6.40 to serve. That's $2,624. Your paying customers used the other $1,592.40. You read all 410 rows by hand, and by 11:30 you're fairly sure about 96 of them. They're the same few people, starting over.
- 96 repeat trials × $6.40 = $614.40 a month
- $614.40 × 12 = $7,372.80 a year
You have 130 customers at $49, so $6,370 came in last month. After this one invoice, $2,153.60 is left for hosting, your tools and your own pay. Without those 96 accounts it would have been $2,768. You don't know it yet, but you've just walked into the returning customer vs duplicate account problem, and it has two ways to lose.
That afternoon you ship the obvious rule: one trial per email address. If an address has had a trial, the form says "This email has already been used." It takes 20 minutes.
The next invoice arrives at 9:04 a.m. on the 1st: $4,069.20. That's $147.20 lower, because the rule refused 23 signups (23 × $6.40). You open the log expecting to feel better.
Seven of the 23 were repeat accounts trying the same address twice, which saved you $44.80. The other 89 repeat accounts simply used a new address, so you're still paying $569.60 for them (89 × $6.40).
The remaining 16 refused addresses had paid you before. One of those people paid $49 a month for 14 months, $686 in all, and left when a client project ended. They came back at 4:17 p.m. last Tuesday with a new project, read "This email has already been used", closed the tab and never wrote to you. There's no ticket and no angry message. There's one row in a log you almost didn't open.
Say half of those 16 would have come back and stayed for a year:
- 8 customers × $49 × 12 months = $4,704 in revenue turned away
- Real repeat accounts stopped by the rule: $44.80
Now you have a decision to make, and you don't like any of the options. You can delete the rule and go back to paying $614.40 a month. You can keep it and go on refusing people who've already trusted you with their money. Or you can cut the trial to 3 days for everyone, including whoever would have been customer 131.
Why email matching can't tell a returning customer from a duplicate account
An email rule asks whether it has seen this string before. You need to know two other things: is this the same person, and did you mean for them to have this twice? A string answers neither, which is why the rule did so badly on your log.
| Who's signing up | Same email as before | Different email |
|---|---|---|
| Someone coming back to use your product | Refused, and you lose a customer | Looks brand new, mostly harmless |
| Someone collecting the benefit again | Caught | Missed, and this is most of the bill |
Email is a weak ID in both directions. Gmail ignores dots in an address, and it delivers mail sent to name+anything@gmail.com to the same inbox, so one mailbox can have many spellings. A shared address like billing@ passes to whoever holds the job next. And someone who changes jobs arrives with a new address and the same habits.
"Duplicate" also covers more than abuse. Some people forget they ever signed up. Plenty keep a work account and a personal one. The difference that holds up is what the person is asking for. A returning customer wants their old work back, or wants to pay. A repeat claimer wants the free allowance again.
So matching harder doesn't solve this. A stricter match refuses more former customers and a looser one misses more repeat accounts, and either way you've only moved the mistake from one column to the other.
The basic checks, and how they treat people who come back
Most teams add these next. Each one does catch something, and you should know exactly what.
Checks on the email address
Email verification catches typos and dead addresses. It only proves someone can open an inbox, so a second real mailbox passes, and a former customer whose code lands in spam doesn't. Unique emails catch the same address twice and nothing else, and they're the rule that refused your 16. Disposable-domain blocks catch known throwaway services until a new one appears. They also turn away people who use privacy forwarding, and someone has to keep that list current forever. Treating Gmail's dots and plus signs as one address catches one inbox written several ways, and does nothing about a second mailbox.
Checks on the connection
One signup per IP address catches repeats from the same connection. It misses every other network, and it refuses the former customer whose colleague started a trial from the same office last week. Campuses and households hit it too, and each one becomes an exception you maintain. VPN blocks catch the ranges on your list, miss the rest and land on customers who work behind a company VPN.
Asking for more at signup
CAPTCHA stops simple scripts, but someone opening a fifth account by hand passes it as easily as the first, and it costs you signups from people using screen readers or privacy tools. A phone requirement slows casual second accounts, bills you for every text and refuses people with a recycled number or no reliable mobile. A card requirement has about the same reach. Anyone with a second card gets through, and plenty of honest people won't hand one over just to try something.
Limits, shorter trials and review
Rate limits catch bursts and miss an account opened every few days. Usage caps stop one account overrunning, and each new account brings a fresh cap. Shorter trials lower your cost per account for everyone, honest users included, and more accounts make up the difference. Manual review catches what a person can spot, and you've already done it: 410 rows at 20 seconds each is more than 2 hours, due again next month, ending in guesses about people who just look odd.
Stack all of that and a former customer coming back after nine months meets a code, a puzzle, a phone prompt and a card form. Someone ready with a fresh mailbox, another number and a second card gets through all four.
A fair returning-customer policy
None of those checks gives you certainty, so write a policy that treats people well when you're wrong. It starts by separating the account from the benefit.
Let anyone have an account. Nobody gets refused at the signup form for having been here before. Your rule only decides whether the free thing comes with it, so a former customer without a second trial can still pay you today.
Write down what "again" means. A trial could be once per person, or once a year. In this scenario a second trial after a year costs you $6.40, and the customer you lost was paying $49 every month.
Let payment history win. If an address has ever paid you, skip the error. Send them to sign in, bring back their work and merge anything they opened by accident. Put a "Been here before? Sign in" link on the form too.
Say no the way you'd say it to a customer. Every refusal should name what they can do next.
Give people a way through. Replies reach a person, and that person can grant the trial. Decide now what you'll do when someone asks for a second one, and write down each exception.
Count both mistakes every month. Add up what repeat accounts cost you, then read every refusal and count how many had paid you before. If a rule refuses more former customers than repeat accounts, turn it off.
What a former customer should see
Welcome back. This address has already had a trial, so we can't start another one automatically. Sign in to pick up where you left off, or reply to this message and we'll take a look.
That policy is fair once you know who's asking. Steps 2 and 3 both begin with "has this person had this before?", and your records can only answer for the address they typed today. The 89 accounts on a new address look like strangers. You could keep adding checks to close that gap, but you've seen what each of them costs.
Evaluating the claim, without building it yourself
AbuseZero exists to evaluate whether a claim like that should go through, and you don't have to build your own abuse prevention system to get there. Where a unique-email rule looks at one string, 140,000+ signals go through multi-layer AI risk analysis that detects linked claims and flags abuse that single checks miss. Those signals are read together, never one at a time, which an IP limit or a domain list can't do. AbuseZero recognizes repeat devices across new accounts, even when email addresses keep changing, and it checks mailbox aliases, disposable domains and mail routing as well. You also get a proprietary algorithm, with AI that checks the claim again. Creating a fresh account isn't enough to erase previous activity.
Your signup form stays open, because the evaluation is about the benefit. When someone claims a trial, your server makes one API call and gets a clear allow or deny, under rules you control. The policy above carries straight over: eligibility can be once per user or once every 365 days, a former customer you want back goes on your allow list, and Observe mode shows what would be denied before anyone is blocked.
Frequently asked questions
Is a duplicate account always abuse?
No. People forget they've signed up, and many keep separate work and personal accounts. It only costs you when the second account collects a benefit you meant to give once. Your one-account-per-person policy should say so.
Should a returning customer get a second free trial?
Usually, after enough time has passed. A second trial a year later is cheap next to a monthly payment. In AbuseZero that's an eligibility setting, so it doesn't need a code change.
Can I find duplicates by matching email addresses?
You'll find the same address used twice. You won't learn whether that's a customer coming back, and you won't see anyone who used a different address. There's more on finding repeat trial users in your database.
Protect your trials and keep your former customers
Start with Builder
Pay as you go. Run your first evaluation today.
See pricing
Builder, Pro, Scale and Enterprise.
Talk to us about Enterprise
High volumes and custom terms.