Back to Blog

What to do in the first hour of a signup attack

A minute-by-minute signup attack response runbook. Confirm it's real, stop the spend, protect paying customers and know which blocks won't last.

It's 2:14 a.m. and your phone buzzes on the nightstand. You open the dashboard with one eye and see 30 new accounts a minute. A good day for you is 30 signups in total. For about four seconds you're thrilled, and then you scroll the list and nobody in it looks like a person.

There's no signup attack response plan in your docs folder, because until tonight you had no reason to write one. So you do the arithmetic in bed, on your phone. Each signup kicks off a welcome email, a workspace setup and an AI onboarding step, which comes to about $0.40 of work you pay for. (Work out your own number.)

30 signups a minute x $0.40 = $12 a minute
$12 x 60 minutes            = $720 an hour
$720 x 8 hours              = $5,760 if nobody notices until 10:14 a.m.

Say your product brings in $4,000 a month. Eight hours of this eats more than six weeks of revenue, spent on accounts that will never open your app again. That's the contractor you were about to hire, or the ad test you'd planned for December. The signup graph you've been checking every morning now has a spike in it that means nothing.

The queue hurts more than the bill does. Three jobs per fake signup is 90 jobs arriving every minute. Say your workers clear 60. Your paying customers' jobs are in that same line, behind the fake ones, and the line grows by 30 jobs a minute.

By 3:14 a.m.:  1,800 jobs waiting, a 30-minute delay
By 10:14 a.m.: 14,400 jobs waiting, a 4-hour delay

At 10:14 a.m., one of your best customers starts a job and watches a spinner. It finishes around 2:14 p.m. Their first message asks if you're down. The next one asks about a refund, and ten refunds at $49 is another $490 gone. You answer them while trying to decide whether to shut off the free plan, the same free plan that brought you nearly every customer you have.

An illustrative first hour at $12 a minute, if nothing changes:

Minute 0Phone buzzesMinute 5$60 goneMinute 10$120 goneMinute 20$240 goneMinute 40$480 goneMinute 60$720 gone
Minute 0Phone buzzesMinute 5$60 goneMinute 10$120 goneMinute 20$240 goneMinute 40$480 goneMinute 60$720 gone

That's a scenario. This one wasn't:

From our early beta

A team whose cloud bill normally ran $200 to $300 a month woke up to a $12,000 bill. Automated registrations had triggered thousands of expensive serverless functions, database operations and background jobs overnight.

That bill is 40 to 60 times a normal month. If this is happening to you right now, go straight to the runbook.

Why signup attacks get expensive so fast

Your app was built to give a new person a great first five minutes, so all of that work starts the moment the form is submitted. A script can submit the form in a fraction of a second and it never stays for the five minutes. You pay for them anyway. Some waves are after your free credits, which is credit farming. Others just hammer the form. For the next hour the reason doesn't matter.

The obvious fix is to block whoever's doing it. At 2:14 a.m. that's the wrong place to start. You don't yet know who's doing it, every block you can reach for can land on a real customer, and the whole time you're writing the rule you're paying $12 a minute. So the order matters: find out if it's real, cap the cost, protect the people who pay you, and only then start blocking.

NoYesYesYesNoNoFree creditsBackground jobsProvider usageSignups far above normal?Keep watchingLaunch or press today?New accounts using theproduct?Real traffic, protect capacityWhat costs most perminute?Pause free grantsPause costly jobsSet a hard capProtect paying customers
NoYesYesYesNoNoFree creditsBackground jobsProvider usageSignups far above normal?Keep watchingLaunch or press today?New accounts using theproduct?Real traffic, protect capacityWhat costs most perminute?Pause free grantsPause costly jobsSet a hard capProtect paying customers

The first hour, minute by minute

Minutes 0 to 5: confirm it's real. Don't change anything yet. Open these side by side:

  • signups per minute, against your normal for this hour of the night;
  • spend at your AI, cloud and email providers;
  • where the traffic comes from in your signup logs, by country, referrer and email domain.

A launch looks like a spike too, but those accounts go on to use the product. If you still can't tell after five minutes, treat it as real. A false alarm costs you some delayed welcome emails. Waiting costs $12 a minute.

Minutes 5 to 10: stop the spend. In this order:

  1. Pause free grants with a feature flag. People can still sign up, they just don't get credits or a trial yet. No flag? An environment variable and a redeploy will do.
  2. Set a hard spending cap at each provider. Read the setting carefully, because some budgets only send an alert and keep charging. Set the cap above what paying customers need for the next day.
  3. Pause the costly background jobs, starting with the AI step. Pause them, don't drop them. Real people signed up tonight too.

Then watch the spend for a minute and make sure it actually falls.

Minutes 10 to 20: protect paying customers. Give their jobs their own queue, or a priority that puts them first. One worker that reads only paid-plan jobs is enough for tonight. Keep their rate limits and quotas exactly as they were, and exempt signed-in paying accounts from anything you turn on next.

At minute 10 the backlog is about 300 jobs, a five-minute delay, so this is the cheap moment to fix it. Sign in to a paid account and run the thing your customers run.

Minutes 20 to 40: apply basic blocks. Turn on one at a time, and repeat your paid-account test after each.

BlockWhat it catches tonightWho it can hurtHow long it holds
Rate limit on the signup routeHeavy volume from a few sourcesOffices, campuses and mobile networks sharing one addressOnly while the traffic stays concentrated
Block the disposable domains in useThrowaway addresses on those domainsPrivacy-minded real usersUntil the next domain. The list is usually stale within days
Temporary CAPTCHAPlenty of basic scriptsPeople using screen readers, old phones or slow connectionsSome automation gets through from day one, and it costs you real signups
Block obvious IP rangesTraffic from those ranges tonightAnyone else on them, including VPN users and company networksOften only hours

Write down everything you turn on. You'll have to take most of it off again.

Minutes 40 to 60: keep the evidence and tell people.

  • Export first. Save signup logs, job records and provider usage for the whole window, before any of them expire.
  • Don't mass-delete yet. Flag the accounts and leave their free grants off. Some real people signed up tonight, and you can't sort them out at 3 a.m.
  • Tell your team, even if your team is one co-founder: what happened, what's paused, who's watching until morning.
  • Tell affected customers plainly. Say what they saw, what you did and what happens next.
  • Write down what happened while you remember it: the timeline, what you switched off, what it cost and what's still paused.

A message you can send

Early this morning a wave of fake signups slowed down background jobs, and some of yours were delayed. They're running again now, and paying customers are first in the queue. We'll update you when everything's back to normal. Sorry for the wait.

Only add "your data wasn't affected" if you've checked.

If the pause at minute 10 held, tonight cost about $120. The same night with nobody awake costs $5,760. That's a $5,640 difference, from switching off three things in ten minutes.

Don't do this at 3 a.m.

Tempting moveWhy to wait
Delete all the new accountsYou lose your evidence and some real customers with it
Block all VPNsIt would stop some of tonight's traffic, and plenty of paying customers work behind one
Add phone verificationIt slows scripts down, but it's new code on a bad night and every text you send is another bill
Require a card from everyoneIt would cut fake signups, but it turns away real ones too, and payment code shouldn't be written tired
Raise the cap so the jobs finishYou'd be paying for the fake accounts to finish

What to set up before the next one

Do these tomorrow, once you've slept.

  • Kill switches. One flag each for free grants, AI onboarding and new signups, all reachable from your phone.
  • Spending caps. Real caps at every provider that bills by usage, with an alert well before each one.
  • Alerts on signup spikes. Alert on signups per minute against the same hour last week. By the time an invoice tells you, the money's gone.
  • Delay expensive work until an account is verified. Email verification won't tell you someone is new or honest. It only tells you they can open an inbox. But every account that never confirms now costs you one email instead of $0.40. There's more on this in why new founders wake up to surprise bills.

Why tonight's blocks won't be there next time

Give it a week. The domain list needs new entries. The IP blocks cover last night's addresses and nobody else's, and one of them may well include a paying customer's office. The CAPTCHA is costing you signups, so someone takes it off. The rate limit is still set to a number you picked at 2:30 a.m. You could review new accounts by hand instead, but at 30 a minute that's one account every two seconds.

Each of these is small. Together they're a second job, redone after every wave, usually at night. And none of them answers the question that costs you the most once the form is submitted: should this account get the free benefit at all?

Before the next free benefit goes out

That question is the job AbuseZero was built for. Before a trial, free plan or batch of credits goes out, your server makes one API call and gets a clear allow or deny. With multi-layer AI risk analysis, AbuseZero examines 140,000+ signals, detects linked claims and flags abuse that single checks miss, which is exactly where a domain list or an IP block runs out. Signals are read together, never one at a time. There's also a proprietary algorithm, with AI that checks the claim again. AbuseZero blocks spammy bots, browser automation and malicious scripts while allowing genuine AI agents and assistants. A CAPTCHA switched on at 2:30 a.m. can't tell those apart. The rules are yours, and Observe mode shows you what would be denied before anyone is blocked. You don't have to build your own abuse prevention system, or rebuild it after every bad night.

Your application still runs signup, and you'll still want the caps and kill switches above. What changes is the moment a free benefit is claimed. The quickstart protects your first one.

Frequently asked questions

How do I know if it's a signup attack or a real traffic spike?

Compare signups with your normal for that hour, then look at what the new accounts do. A launch brings people who use the product. An attack brings accounts that do nothing once the work you pay for has run. Going viral can look similar for the first few minutes.

Should I shut off signups completely?

Only if you can't contain the cost by pausing free grants and costly jobs first. If you do, say so on the page, and turn it back on as soon as your caps and queues are in place.

How do I keep free credits away from fake accounts?

Hold the credits until an account is verified, then evaluate each claim before you grant it. AbuseZero's evaluation returns allow or deny. See how to stop repeat credit abuse.

Get ready before the next bad night

Keep reading

Copying your prompt…

Copied the to your clipboard. Give it to your coding agent & it will handle the rest.