A referral mechanic on a waitlist is the cheapest growth lever available and the fastest way to ruin the data you were collecting. Both are true at once, and which one you get depends on decisions you make before turning it on.
Why position-jumping works when a discount does not
The thing being offered is not a reward. It is a queue position, and queue positions have three properties that make them unusually effective for a product that does not exist yet.
They are visible. You are #312. That is a number about you, and a number about you invites improvement in a way "get 20% off" does not.
They are relative. Moving up means moving past people. Discounts are absolute and forgettable; rank is competitive and sticky.
They cost nothing. You are not discounting a product with no price. You are reordering a list. That means the mechanic works before you have decided anything about pricing, which is exactly when a waitlist exists.
The corollary is the constraint: position only motivates if position means something. If everyone gets in on the same day, the queue was theatre and the people who shared for it will notice. Either stage your access genuinely, or do not lean on rank.
What it does to your data, and why that matters
This is the part usually left out.
The moment sharing is rewarded, your signup list stops being a sample of people with the problem and becomes a mix of those and people recruited by someone chasing a number. The second group did not arrive because they have the problem. They arrived because their friend asked.
That is not worthless — it is real reach — but it contaminates precisely the measurements a waitlist is for. Your conversion rate drops and you will not know whether that is the mechanic or the market.
Two things make this survivable:
Segment referred signups from direct ones and read every metric separately. If direct signups answer your free-text question at 40% and referred ones at 8%, that gap is the finding, and averaging it produces a number describing nobody.
Keep the free-text question mandatory-ish. People recruited by a friend, asked what they currently do about the problem, will often simply not have an answer. That is diagnostic rather than annoying.
The abuse arrives within a day
Not "may arrive". Turning a queue position into a reward creates an incentive to manufacture signups, and somebody will, quickly. The categories are predictable:
- Disposable addresses, generated in bulk.
- Plus-addressing, where
you+1@,you+2@all deliver to one inbox. - Small rings, where a handful of people refer each other in a loop.
- Ordinary bots filling any form they find.
Osokoro ships four controls for this and they are worth turning on before the mechanic rather than after: double opt-in, CAPTCHA, disposable-domain blocking, and referral-abuse detection. Double opt-in is free on every plan including the free one, because a referral count built on unverified addresses is not a count.
Only verified signups should ever move somebody up the queue. If an unverified address advances someone, you have built a machine for generating fake positions, and the people gaming it will find that out before you do.
Rewarding effort rather than volume
The default design — one place per referral — optimises for volume, which is where the abuse pressure comes from. Some alternatives that produce a better list:
Thresholds rather than a linear count. Something real at 3 and something better at 10, rather than a place per head. Fewer people bother, and the ones who do are the ones actually invested.
Reward on verification, never on signup. Obvious once said, routinely not implemented.
Cap it. Nobody needs to be able to earn two hundred places. A ceiling removes the leaderboard- farming behaviour without removing the mechanic.
Offer something other than rank. Osokoro has perks and a leaderboard on the paid plans, and the useful thing about a perk is that it can be genuinely scarce — early access to something specific, input on a decision — where rank is infinitely divisible.
What is on which plan
Being concrete, since it affects what you can actually do:
- Referrals and email verification: every plan, including Free. These are the two that stop the mechanic being a fraud generator, so they are not a paid feature.
- CAPTCHA: Starter ($10/mo) and up.
- Leaderboard, perks and broadcasts: Everything ($20/mo).
Free gives you 1 waitlist, 50 contacts and 100 monthly sends, which is enough to run the experiment and not enough to run the campaign.
The measurement to actually watch
Not total referrals. Referrals per verified referrer, and the shape of its distribution.
If a handful of people account for nearly all of it, you do not have a referral loop — you have four enthusiasts, which is lovely and is not a growth mechanism. If it is broad and shallow, with many people bringing one or two, the loop is real.
The second pattern is rarer and much more valuable, and the two are indistinguishable in the total.