How to stop expensive work running on unverified signups
Every unverified signup can start work you pay for. How to delay background jobs until verified, cap what's left and see where basic gates break.
Verification email $0.001
Workspace setup and database writes $0.039
Sample data and search indexing $0.07
AI onboarding summary $0.24
One signup $0.35
22 signups a minute x $0.35 = $7.70 a minute
$7.70 x 60 minutes = $462 an hour
$462 x 7 hours = $3,234 by morningIt's only an example, but the sums work the same with your numbers. The sum is running right now. It's 4:07 a.m., you went to bed at 11:40 with a quiet dashboard, and at 12:20 a script found your signup form. That was 227 minutes ago. There are 4,994 new accounts (227 x 22), and they've cost you $1,747.90 so far (227 x $7.70). Your phone is face down on the nightstand and nothing has paged you, because nothing has failed. Every function returned a success and every job finished.
If you hit snooze once at 7:11, those nine minutes cost $69.30 (9 x $7.70). When you pick up your phone at 7:20 there are 9,240 accounts (420 x 22), $3,234 is gone, and the script's still running while you look for your laptop.
Hardly any of those accounts has confirmed an email address. They never needed to, because your app did all the work the moment the form was submitted. Closing that gap is what this post is about: delay background jobs until verified, so an account that never opens an inbox costs you a tenth of a cent instead of 35 cents.
Your AI step shares a key with your product. Say your provider allows 200,000 tokens a minute. 22 summaries x 8,000 tokens is 176,000, which leaves 24,000 for everyone else. By 8 a.m. your paying customers usually need about 60,000, so 40% of their requests get through (24,000 ÷ 60,000) and the other 60% come back as errors. The first message lands at 8:04: "Is it down? Our whole team's stuck."
Say you've got $41,000 in the bank and spend $4,100 a month. That's ten months of runway, and last night took about 24 days of it ($3,234 ÷ $4,100 x 30). Six customers ask for their $39 back, another $234. The signup graph you check every morning now has 9,240 people on it who don't exist. By 9 a.m. you have to choose: shut off signups and lose the free plan that brought you nearly every customer you have, or leave them open and keep paying $7.70 a minute to see if it stops by itself.
That night was a scenario. This one happened:
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 ($12,000 ÷ $300 and $12,000 ÷ $200), and the registrations alone set it off.
Why signups trigger expensive work
Most signup code has one obvious place to hang things: the moment the account row is created. The workspace gets built there, the sample data gets seeded and the AI step runs, so a new person sees a finished product in their first minute. A script gets the same first minute, and submitting your form 22 times costs it almost nothing.
The obvious fix is to turn on email verification. Do it, but then look at what it changes in your code. In a lot of apps it only decides whether someone can sign in. The hook that queued four jobs has already fired by then, so you can have verification switched on and still pay 35 cents for every account that never confirms.
How to delay background jobs until verified
Price everything that runs at signup. Search for every place that reacts to a new account: auth hooks, webhooks, database triggers, onboarding code. Write a cost per run next to each one.
Move the trigger to the confirmation. Before someone confirms, the only things that should happen are an account row and one email. Queue the rest in the same statement that marks the account verified, so a link opened twice queues one batch. This version is for PostgreSQL:
with verified as (
update accounts
set verified_at = now()
where id = :account_id
and verified_at is null
returning id
)
insert into jobs (account_id, kind)
select id, 'set_up_workspace'
from verified;Accounts that come in through social sign-in may arrive already verified. Send them through the same function, so there's only one way into the queue.
Give unconfirmed people something to look at. One static sample workspace, shared by everyone, costs next to nothing and keeps some of that first minute. Limit resends too.
Move the costliest step later still. A confirmed email doesn't mean someone's sitting there. Run the AI summary when a person opens onboarding and asks for it. In the scenario, that's $0.24 of the $0.35.
Give free work its own lane and budget. Use a separate queue, a concurrency limit and a separate API key with a hard spending cap, so a bad night can't use up what paying customers need.
Expire what never confirms. Delete unconfirmed accounts after seven days.
Find out in minutes. Alert on signups per minute against the same hour last week. An alert at 12:35 would have turned that night into a 15-minute problem: 15 x $7.70 = $115.50.
Run the same night through the new flow. Say 1 in 10 of those accounts confirms, and a quarter of those ask for the AI step. Both rates are illustrative.
9,240 emails x $0.001 = $9.24
924 confirmed x $0.109 = $100.72
231 AI summaries x $0.24 = $55.44
Total = $165.40That's $3,068.60 less than the $3,234 night, from moving triggers around. The middle bar below is every job moved to confirmation, with the AI step still running there: $331.72 across 9,240 accounts.
Where the basic gates break
Say the next wave uses mailboxes that can receive mail, and 7 in 10 accounts confirm instead of 1 in 10. That's 6,468 confirmed x $0.109, plus 1,617 AI summaries x $0.24, plus $9.24 in email, which comes to $1,102.33. Email verification proves someone can open an inbox, and a catch-all domain can give one person as many working addresses as they like.
So you add another gate. Each of these does stop something.
| Gate | Stops the jobs for | Still pays for | Turns away | Upkeep |
|---|---|---|---|---|
| Email verification | Typos and scripts with no inbox | Any mailbox that can receive mail | People whose email lands in spam | Watching bounces |
| Unique emails | The same address twice | Every different address | Almost nobody | Very little |
| Disposable-domain blocks | Known throwaway services | Domains that aren't on your list yet | People using privacy forwarding | A list someone updates forever |
| Gmail dot and plus rules | One inbox written several ways | A second real mailbox | People who sort their mail with plus addresses | Upkeep as mail services change |
| CAPTCHA | Simple scripts | Automation that gets past it | People using screen readers or privacy tools | Fees and lost signups |
| Phone requirement | Casual second accounts | Anyone with another number | People without a reliable mobile | A charge for every text |
| Card requirement | Scripts with no card, casual second accounts | Anyone with another card | Students and people who want to try first | Fewer signups |
| One signup per IP address | Repeats on one connection | Any other network | Offices, campuses and households | Exceptions and tickets |
| VPN blocks | Known VPN ranges | Everything off your list | Staff on company VPNs | Another list to maintain |
| Rate limits | Bursts from a few sources | Traffic spread thinly | Real users on a shared network | Tuning per route |
| Usage caps | One account overrunning | Each new account's fresh cap | Anyone at the cap on a good day | Deciding when to stop |
| Shorter trials | Some value per account | More accounts make up the difference | Every honest user | Fewer people reach paying |
| Manual review | What a person can spot | Anything that arrives overnight | Whoever looks odd to a tired reviewer | Hours you don't have |
Gmail ignores dots in personal addresses, and a plus sign and a word added to an address still reach the same inbox, so a unique-email rule sees different people. Your app just emailed 9,240 addresses, and many of them don't exist. Bounces at that volume can hurt your sender reputation, and then a real person's confirmation lands in spam and their jobs never start. Your gate is now keeping out the customer it was meant to protect. By hand, 30 seconds per account comes to 77 hours for one night.
Stack everything and an honest person gets a code, a puzzle, a text and a card form, while you look after two lists, email rules and limits you picked at 2 a.m. Each of those gates asks about an inbox, a phone or a connection. None of them asks whether this account should get the free thing your jobs are about to pay for.
Deciding who gets the free work
AbuseZero exists to answer that, so you don't have to build your own abuse prevention system. Where your gate saw one confirmed inbox, multi-layer AI risk analysis examines 140,000+ signals, detects linked claims and flags abuse that single checks miss. Your domain list goes stale, and AbuseZero checks mailbox aliases, disposable domains and mail routing for you. Your per-IP rule catches the office next door, while AbuseZero sees beyond an IP address with VPN, hosting, network reputation, Tor and proxy intelligence. A CAPTCHA puts a puzzle in front of everyone. AbuseZero blocks spammy bots, browser automation and malicious scripts while allowing genuine AI agents and assistants. AbuseZero uses a proprietary algorithm, with AI that checks the claim again. Creating a fresh account isn't enough to erase previous activity.
It's one API call and a clear allow or deny, under rules you control. Protect a benefit end to end shows where the call goes, and Observe mode shows what would be denied before anyone is blocked. Your signup flow, queues and spending caps stay yours, and you'll still want every step above. AbuseZero's evaluation covers the moment a free benefit is claimed.
Frequently asked questions
Will waiting for confirmation hurt activation?
It can, because some people won't come back. A static sample workspace and a clear "check your email" screen help.
What if the jobs are running right now?
Pause the free work first and go straight to the first-hour runbook.
Is a confirmed email enough to grant free credits?
It tells you an inbox works and says little else. An AbuseZero evaluation at the moment of the claim returns allow or deny, and the quickstart walks through your first one.
Protect the free work you pay for
Start with Builder
Pay as you go. Protect your first benefit today.
See pricing
Builder, Pro, Scale and Enterprise.
Talk to us about Enterprise
High volumes and custom terms.