Back to Blog

How fake signups distort your trial conversion rate

Fake signups inflate your signup count and drag your conversion rate down. Work out your real number and see what the basic fixes do to it.

Sanket DattaFounder & CEO

The reply from the investor's associate is one line long, and you can't stop rereading it. Say you run a $49-a-month product with a 14-day free trial, you're in the middle of a fundraise, and the number on your growth slide is one you're proud of: 9,000 trial signups last quarter, up from 6,000 the quarter before. The reply to your deck is one line: "How many of the 9,000 are unique people?"

You don't know. So you pull 200 trials at random and read them by hand, 45 seconds each, which takes you to 7:15 p.m. A few are the same Gmail inbox written three different ways, and others sit on throwaway domains. By the end you've marked 70 of the 200 as someone who'd already had a trial, which is 35%. You do the quarter before as well, finish at 9:45 p.m., and that one comes back at 10%.

Those two quarters brought in 162 and 180 customers. Put it all in one table and it's the worst thing you've read all year.

Quarter beforeLast quarterChange
Signups, as reported6,0009,000Up 50%
Signups, real people5,4005,850Up 8.3%
Conversion, as reported2.7%2.0%Falling
Conversion, real people3.0%3.1%Flat

That's the fake signups conversion rate problem. The same accounts made your signups look better and your conversion look worse in the same quarter, and neither number was telling you anything about your product.

You've already acted on both. You read that slide from 2.7% to 2.0% as a pricing problem, so you cut the price from $59 to $49. On this one quarter's customers, that's $10 × 180 × 12 months, or $21,600 a year. You read the 50% as a wave of new users and hired a second support person at $5,500 a month, $66,000 a year, for tickets that never came, because fake accounts don't write in. Serving the 3,150 fake trials at $3.20 each cost $10,080 last quarter on top.

That's $97,680 in all. If you burn $40,000 a month, you've given up nearly two and a half months of runway to people who don't exist. You also spent six weeks rebuilding onboarding to fix a drop that never happened, and the export feature promised to your biggest customer slipped to make room. They've asked about it twice.

It's 9:45 p.m. and the email is still open, with a partner meeting on Monday. You can send 5,850 and watch a 50% story turn into an 8% story, or you can stay quiet and keep presenting a number you now know is wrong. Nobody lied to anyone. Your database counted accounts and you reported what it counted, which doesn't make the reply any easier to write.

That was a scenario. This wasn't.

From our early beta

After a launch went viral on X, one team gained 6,400 new users in a month. 2,700 of them were duplicate accounts. About 32 became paying customers.

That's about 42% duplicates (2,700 ÷ 6,400). Count every account and conversion is 0.5% (32 ÷ 6,400). Take out the duplicates and it's about 0.9% (32 ÷ 3,700). The customers didn't change between those two rates. Only the bottom of the fraction did.

Same 32 customers, two rates00.20.40.60.81All accountsNo duplicatesConversion (%)
Same 32 customers, two rates00.51All accountsNo duplicatesConversion (%)

Why fake signups pull your conversion rate down

Conversion is a fraction. Paying customers go on top and trial signups go on the bottom, and the bottom counts accounts because accounts are what your database has. A free trial hands out something worth having, so one person has a reason to open five accounts and a script can open hundreds. Those accounts rarely pay, so they grow the bottom and leave the top alone.

The share of fake signups also moves around for reasons unrelated to your product. A launch or a mention on a deal site can change it from one week to the next, so your rate falls in a week when nothing changed, or rises the week you happened to ship a new pricing page.

Fake and duplicate signupsSignup count goes upCustomer count stays putReported conversion fallsGrowth looks strongerYou fix a funnel that worksYou plan for userswho don't exist
Fake and duplicate signupsSignup count goes upCustomer count stays putReported conversion fallsGrowth looks strongerYou fix a funnel that worksYou plan for userswho don't exist

The obvious fix is to filter the report by dropping accounts that never came back after day one. Say 5,000 of your 9,000 never returned. Drop them and you get 180 ÷ 4,000 = 4.5%, which is flattering and wrong, because plenty of those 5,000 were real people who looked around and left. A duplicate that used up its trial in an afternoon looks active and stays in.

How to work out your real trial conversion rate

You can get a usable answer in an evening with this formula:

Real conversion=Paying customersTrial signups×(1−Fake share)\text{Real conversion} = \frac{\text{Paying customers}}{\text{Trial signups} \times (1 - \text{Fake share})}

Pick one finished cohort. Use trials that started in a single quarter and have all ended, and count customers from that group only.

Draw a random sample of 200. Don't pick the strangest-looking ones.

Read each one by hand. Mark an account as a repeat only when you'd be comfortable saying so to the person: the same inbox written differently, a throwaway domain, or someone telling support it's their second trial. If you can't tell, count it as a person. You'll undercount, which is the safer way to be wrong.

Report a range. A sample of 200 at 35% is good to about 7 points either way. In the scenario, a fake share of 28% to 42% puts real conversion between 2.8% and 3.4%.

Do the previous cohort too. Two corrected numbers tell you whether the trend you've been steering by was real.

To see how far the reported rate can drift, hold real conversion at 4% and change only the fake share. These numbers are illustrative.

Fake share of 10,000 signupsReal peopleCustomersReported conversion
0%10,0004004.0%
10%9,0003603.6%
25%7,5003003.0%
42%5,8002322.3%
50%5,0002002.0%

Every row is the same product. For a fuller audit, see how to measure your free trial abuse rate and how to clean duplicate accounts out of your SaaS metrics.

The basic fixes, and what they do to your numbers

Keeping fake signups out of next quarter is harder than measuring them, and each basic fix carries an extra cost: it changes who's in the fraction, so your before and after stop being comparable.

Email rules

Email verification clears out typos and dead addresses, but a verified inbox isn't a new person, and real people lose the code to their spam folder. A unique-email rule stops the same address twice and misses every different address. Gmail ignores dots, and mail sent to jane+trial@gmail.com lands in Jane's inbox, so one mailbox can look like many signups. Disposable-domain blocks catch the throwaway services on your list, turn away people using privacy forwarding, and need someone updating that list for as long as you have a signup form.

CAPTCHA, phone and card

CAPTCHA stops simple scripts. It does nothing about a person opening a fifth account by hand, and it's hard on people using screen readers. Phone and card requirements make a casual second account more work, which has real value, but anyone with another number or card gets through, people without either can't, and every text costs you money. These two also move the metric the most. Require a card and signups fall while conversion jumps, and you can't tell how much of that jump is fake signups leaving and how much is real people who wanted to try first and walked away.

Network limits

One signup per IP address catches repeats on a single connection and lands on offices, campuses and households. VPN blocks catch the ranges on your list and shut out privacy-minded people and staff on company VPNs, and now you've got a second list to maintain. Rate limits stop bursts, need tuning for each route, and never notice 40 accounts that each stay under the limit. Usage caps hold one account to its allowance, cut off real users mid-task, and give every new account a fresh allowance.

Smaller trials and human review

A shorter trial lowers what each fake signup costs you and what every honest user gets, and it does nothing to stop someone starting another one. Manual review is the most accurate option here, and the sample above shows its ceiling. All 9,000 accounts at 45 seconds each is 112.5 hours, every quarter, and it only tells you about trials you've already handed out.

Stack all of this and your form gets longer for everyone, and your rate now depends on which rules were on in which week. None of it answers what your fraction needs to know: has this person already had this trial?

What to do with the corrected number

Plan on what fake signups can't touch. Paying customers and revenue are the same in both versions of that table, so hiring and pricing should rest on them first and on signups last. Report signups with the sampled share beside them, and log every signup rule you change on the same chart. If an investor already has the old figure, send the correction with your sample size and range. It's a hard email, and it's easier to send than to be asked about later.

Keeping next quarter's number honest

A spreadsheet of 200 accounts gets you through one email. AbuseZero is built for every trial after it, and it starts where those rules ran out. Your server makes one API call before a trial starts and gets a clear allow or deny, under rules you control. AbuseZero flags abuse that single checks miss and detects linked claims, with multi-layer AI risk analysis examining 140,000+ signals. There's also a proprietary algorithm, with AI that checks the claim again. AbuseZero checks mailbox aliases, disposable domains and mail routing, and it recognizes repeat devices across new accounts, even when email addresses keep changing. Creating a fresh account isn't enough to erase previous activity, and you don't have to build your own abuse prevention system to get there.

Observe mode is the part that matters most for your numbers. Every trial still goes through, and you see what would be denied before anyone is blocked, so you can measure how much repeat claiming your trial attracts. The quickstart sets up your first benefit.

Frequently asked questions

Do fake signups make conversion look higher or lower?

Lower, as long as you're dividing by all signups. They add to the bottom of the fraction and rarely to the top.

How many accounts do I need to sample?

200 gets you within about 7 points at a 35% fake share, and 400 gets you within about 5. If you're making a hiring or pricing call, do the 400.

Should I remove fake accounts from my past reports?

Keep the raw numbers and add a corrected column beside them, with the date and sample size.

Get a number you can stand behind

Keep reading

Copying your prompt…

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