Instead of paying to advertise your product, you pay to advertise the payout: "earn $100 a week sharing our app". The hope behind an ad for the payout is that cheap banner ads recruit an army of people who sell your product for you on commission, which is why so many companies have tried it.
The short version:
- Buying ads to recruit referrers loses badly for a consumer product. On Google Display's $0.63 average cost per click, and with 2% of clickers signing up to refer, one recruited referrer costs about $31 in ads. Roughly 90% of people who join a referral or affiliate program never refer anybody, so one referrer who actually refers costs about $315.
- Paying the reward on signup can lose money at any volume. A $25 reward paid on every referred signup, when 30% of those signups go on to pay, has already cost $83 in rewards per paying customer before any ad spend. A stranger reached through Meta ads costs $42 to $48 per paying customer. Pay the reward on the first purchase instead.
- Whether recruiting referrers with ads breaks even depends on the size of each referrer's audience, not on your ad. Even on generous assumptions, each recruited referrer who actually refers has to produce about five or six referred signups to break even, and on ordinary numbers closer to 60. An ordinary consumer produces one or two referred signups, and somebody with an audience produces dozens.
- An ad that promises the sender money works against what makes referrals succeed. Field experiments in the Journal of Marketing Research found referral rewards aimed at the invited friend beat rewards aimed at the sender. An ad saying "Earn $1,000" is the most self-interested framing there is.
- A referral ad that promises a dollar figure breaks the ad policies of both Google and Meta. Google and Meta both reject specific income promises. In the United States, "earn up to $10,000" is a regulated earnings claim that has to be what a typical person earns, not merely possible. The FTC, the US consumer protection regulator, sued Arise Virtual Solutions, a company that recruits work-from-home call centre agents, over an advertised "up to $18 an hour" that fewer than 1% of workers earned.
- Uber ran the largest referral-incentive programme ever, then shut it down. Driver referral incentives cost $199M in 2017 and $103M in 2019. Uber shut both its rider and driver referral programs in 2020.
- A shareable referral link is the best way for customers to refer, but referral links leak onto coupon sites. On coupon sites, referral links collect rewards on customers who were already coming to you.
- Advertise the referral programme to the customers you already have, not to strangers. Your own customers cost nothing to reach and already like your product. People with audiences are best recruited by contacting them by hand, not by bidding for ads.
An ad-funded referral program advertises at a salesperson
An ordinary ad campaign sells to the person who will pay you. An ad that recruits referrers puts a stranger in the middle:
- An ad that recruits referrers buys a distributor, not a customer. That ad becomes the first of four steps before you earn any revenue, not the last.
- A cash referral offer gets people to sign up in every audience, which is why recruiting referrers with ads looks promising.
- A referral reward is a real cost of every sale, not a marketing expense. Every customer who arrives through a referral carries a permanent payout, unlike an ad you can switch off.
- You are legally responsible for how every referrer you recruit behaves. What your referrers text, whether they disclose being paid and where they post your code all become your legal risk.
An ordinary ad campaign fails by wasting money. An ad that recruits referrers fails by acquiring people whose incentives are not aligned with yours, and then paying them.
Referrals are the cheapest customer you cannot buy more of
The idea of buying referrers with ads starts from a correct premise:
- A referral program produces the cheapest paying customers of any marketing channel measured in this series of articles. The customer acquisition cost breakdown puts a referral program at $25 to $65 per paying customer with the reward included, against $42 to $48 on Meta.
- Customers who arrive through a referral are also worth more to the business than other customers. Marketing professors Philipp Schmitt, Bernd Skiera and Christophe Van den Bulte tracked about 10,000 customers of a German bank and found referred customers were worth at least 16% more than comparable non-referred ones.
- A referral program is capped by the size of your customer base, and buying referrers with ads is an attempt to break that cap. Only 5% to 15% of a company's customers ever share a referral link at all.
- The logic of buying referrers with ads is sound as far as it goes. If referral program members can be bought through ads, buying them turns a referral program capped by your customer count into one with no cap.
Use Repeat Recorder to hear how you describe your product
Only 5% to 15% of your customers ever share a referral link. What those customers share is usually a one-sentence description of your product. Repeat Recorder, one of my own apps, records you and plays each recording straight back, so you hear your own description the way somebody else hears it.
- Repeat Recorder plays each take back automatically and then starts recording again, so you can try the same sentence twenty times without picking the phone up.
- A referral program produces paying customers at $25 to $65 each, against $42 to $48 for a stranger on Meta, and all of that saving comes from a customer describing your product in their own words.
- Takes worth keeping go into favourites and into folders you name yourself, dragged into whatever order you want. The product description you settled on is still in Repeat Recorder next week.
Five multiplications turn a $31 referrer into a $315 one
This is what $1,000 of Google Display ad clicks produces after the five steps between the ad money and a paying customer:
$1,000 at Google Display's average $0.63 cost per click1,587 clicks
× 2% fill in the signup form32 referrers
× 10% of program members ever refer anybody3.2 active
× 2 referred signups each, a normal consumer figure6.4 signups
− 15% of those are self-referrals and fake accounts5.4 real
× 30% of real referred signups become paying customers1.6 customers
Ad cost per paying customer, before you pay a single reward$617
- One referrer recruited through Google Display ads costs about $31 in ad spend. $31 is already two thirds of the $42 to $48 that a finished, paying customer costs on Meta.
- One ad-recruited referrer who actually refers anybody costs about $315. Around 90% of the members of a typical affiliate program send it zero visitors.
- Assuming 2% of strangers who click a display banner sign up to refer is generous, not harsh. Google Display ads turn roughly 0.59% of clicks into a signup or sale for an ordinary offer.
- Doubling the 2% signup rate still leaves each ad-recruited paying customer costing about six times the $42 to $48 a stranger costs on Meta.
| Stage in the chain | Realistic value | Best case you could argue for | Who actually controls it | Can you fix it with money? |
|---|---|---|---|---|
| Cost per ad click | $0.63 Display average | $0.30 on Meta | The auction | Somewhat. Better ad design moves it 2x at most |
| Click to referrer signup | 2% from cold banner traffic | 5% with a strong landing page | You | Yes. Landing page work doubles it |
| Signup to ever referring anyone | 10%, since about 90% of affiliates never send anyone | 25% with onboarding and nudges | The recruit | Barely. Reminder emails do not give a recruit an audience |
| Referred signups per active referrer | 2 for a consumer | 30 or more for somebody with an audience | The recruit's reach | No. You can only select for it |
| Referred signup to paying customer | 30% inside your funnel | 40% for a well-matched referral | Your product and your price | Somewhat. Ordinary funnel work |
| Fraud loss on referred signups | 15% when the reward is cash | Under 3% if you pay only on a first purchase | Your payout trigger | Yes. One rule change |
A $25 bounty commits $83 per customer before you buy an ad
The $83 cost comes from paying the referral reward on signup rather than on purchase:
- Pay the referral reward on signup and you pay for every referred signup, including people who never buy. If 30% of referred signups become paying customers, a $25 bounty costs $83 in rewards for each paying customer.
- Paying $83 in rewards per paying customer is a cost on every single customer, so more referrals cannot rescue it. If the reward per paying customer already exceeds what a stranger costs on Meta, no number of referrals makes the program profitable.
- Pay the referral reward on the first purchase and the same $25 reward costs $25 per paying customer. Paying on first purchase also removes most referral fraud, because a fake account has to spend real money to collect the reward.
- Paying the reward only on a first purchase means fewer referrers bother, and the referrers you lose are mostly people who were never going to produce a customer.
$83 What a $25 reward paid on signup costs per paying customer when 30% of signups pay, against $42 to $48 for a stranger on Meta.
Work out your own break-even
An earn $1,000 ad recruits people who want money, not your app
Ad platforms keep showing an ad to the kind of people who respond to it, and the people who respond to an "earn money" ad are the people interested in being paid:
- The people most likely to click an "earn money" ad are the people least likely to be believed when they recommend your product. A recommendation beats an ad only because the person recommending appears to have nothing to gain.
- Someone recruited by a banner about money has never used your product. A banner-recruited referrer's pitch collapses into "use my code", the weakest possible message.
- Advertising for people who want a payout selects for people who will game the payout.
- Ad targeting cannot separate genuine referrers from people gaming the payout, because the "earn money" ad itself decides who responds.
- People who were paid to install an app also stop using it sooner. In mobile advertising, ad networks that pay people to install apps keep around 1.1% of installers using the app after 30 days, against 2.4% to 4.4% for gaming apps generally.
| Who answers an "earn money" ad | Share of the signups | Referrals worth having that they produce | How persuasive their recommendation is | Fraud and compliance risk | Worth recruiting? |
|---|---|---|---|---|---|
| Curious clicker, signs up and forgets | The clear majority | Zero | None. They never say anything at all | None. They are inert | No. Pure cost |
| Side-income seeker working several programs | A large minority | A handful, mostly low quality | Low. Cannot vouch for the product | Medium. Spams codes to aggregator sites | No. Costs you margin |
| Organised fraud ring with fake identities | Small in number, large in payout share | None. Volume, and all of it worthless | None. There is no real person to persuade | Severe. This is their whole business | No. Paying on signup funds them |
| Existing happy customer who saw the ad | Tiny | One or two, but genuine | High. They actually use the product | Low | Yes, but not this way. Reach them in-app for free |
| Micro-influencer or niche community owner | Rare | Dozens, sometimes hundreds | High. Their audience already trusts them | Medium. Disclosure obligations are on you | Yes. The only group worth paying to recruit |
| Coupon and deal aggregator site | Few accounts, enormous traffic | High volume, mostly traffic you already had | None. They intercept at checkout | Medium. Legal, and it eats your margin | No. They bill you for customers you had |
The loudest money framing is the worst-performing one
Research on referral programs finds the opposite of what most people expect:
- A referral reward paid to the invited friend beats one paid to the person who sends the invite. In field experiments, marketing researchers Rachel Gershon, Cynthia Cryder and Leslie John found prosocial incentives that pay the recipient recruited more new customers than selfish ones that pay the sender, published in the Journal of Marketing Research in 2020.
- People send more referral invites when the reward goes to the friend, because the invite looks like a favour. A referral invite that pays the sender feels like a sales pitch, so senders send fewer of them.
- Most referral programmes reward the sender, the design those researchers found works worse.
- An "earn $1,000" recruiting ad puts the whole reward on the sender and then advertises that reward. As a result, every referral message a recruit sends reads as "I am being paid for this".
An ad that recruits paid referrers buys recommendations at a premium price, then spends the money in the one way that guarantees the recommendations sound paid.
Use Camera to Clipboard to make an ad that shows the product
Google and Meta both reject a referral ad that names a dollar figure, so the ad has to show the product doing something a viewer recognises, in a photo or a captured screen. Camera to Clipboard, one of my own apps, gets a photo from your phone into your design tool in one step instead of three.
- A photo you take inside Camera to Clipboard is on your computer's clipboard the moment you take it.
- A Camera to Clipboard photo pasted into Photoshop or GIMP arrives as a new layer or a new document, which is where an ad actually gets built.
- Rewards aimed at the friend beat rewards aimed at the sender, so the picture in the ad has to show what the friend gets rather than what the sender earns. A photo of the product in the middle of a task shows the friend what they get.
Paying on signup puts a bounty on fake accounts
An ad that advertises a referral payout finds the people most motivated by the reward, then tells them the reward exists:
- Self-referral, one person referring a second account of their own, is the most common referral fraud and is cheap to run.
- In 2014 Uber rider Blake Jareds collected $50,000 in Uber ride credit from his own referral code. Jareds named his code "UBER$20FREERIDE" to rank in search, then posted it on Reddit.
- Referral rings, groups of real people referring each other in a circle, slip past fraud checks that look at one account at a time. A referral ring looks exactly like a group of friends.
- When you pay the referral reward matters more than how well you detect fraud. Requiring a genuine first purchase before paying the referral reward makes fraud cost more than it pays.
The detail that decides everything else
Reward on the referred person's first purchase, never on their signup, and never on an install. Paying on the first purchase stops referral fraud paying. When 30% of referred signups pay, that one rule change cuts the reward cost per paying customer from $83 to $25.
Uber paid $199 million for referrals, then shut it down
- Uber's referral bounties were enormous and heavily advertised.
- Uber's spending on driver referral rewards fell every year from 2017 to 2019. Uber's driver referral incentives ran at $199 million in 2017, $136 million in 2018 and $103 million in 2019. Uber discontinued both the rider and driver referral programs in 2020.
- Uber's referral programs worked for as long as they did because Uber is a two-sided marketplace of drivers and riders, not because referrals work by themselves. Every new Uber driver reduced wait times for every rider.
- So many Uber drivers quit that paying to recruit drivers had to be repeated rather than paid once. A study by Stanford researchers working with Uber found 68% of Uber drivers quit within six months, so the recruiting spend repeats every year.
- Uber's referral spending is best read as a way to win market share, not a way to make a profit.
The other big spender, and what happened next
Temu, the cheap-goods shopping app, ran a consumer version of paid referrals, pairing referral rewards with an ad budget JPMorgan estimated at a $3 billion annual run rate. In January 2024, Morgan Stanley analysts reported about 20% fewer households shopping on Temu than in September 2023. Cash rewards recruit shoppers, but cash is much weaker at keeping them.
Shared links have the least friction and the worst leakage
A referral link that is easy for a customer to share is just as easy for anyone else to post:
- A referral link credits the referrer automatically when the friend clicks it, and needs no customer support.
- A referral link is also the easiest thing in the world to post publicly. Referral links pile up on coupon sites where people look for a code before checking out.
- A customer already about to buy from you who searches for a code and finds a referral link costs you a reward you never needed to pay. Paying a reward on a sale you already had is the affiliate cannibalisation problem.
- Referral codes are harder to share than referral links, but easier for you to control. A referral code can be capped, expired, made single-use or tied to a named partner.
- The strongest referral setup is a link, with the reward paid on the friend's first purchase and public posting of the link banned in the program terms.
| Mechanism | Friction for the referrer | Friction for the friend | How well it tracks | Exposure to fraud and leakage | App store rules |
|---|---|---|---|---|---|
| Personal shareable link | Lowest. Copy and paste, one tap | Lowest. Tap and land, nothing to enter | Automatic, until a cookie is blocked | Highest. Ends up on aggregators | Fine on both stores |
| Personal coupon or promo code | Low. Short string, easy to send | Medium. Has to be typed at checkout | Exact, and it survives lost cookies | Medium. Cappable, expirable, revocable | Fine on both stores |
| In-app invite from the contact list | Lowest. Pick names, tap send | Medium. Arrives as an unsolicited message | Exact | Highest legal exposure. You wrote the text | Allowed, and the message is attributed to you |
| QR code for in-person sharing | High. Both people need to be present | Lowest. Point the camera | Exact | Lowest. Hard to farm at volume | Fine on both stores |
| Reward paid to the friend, not the sender | Lowest. It reads as a favour, not a pitch | Lowest. They get something | Same as a link, so the same cookie problem | Medium. Still worth a fake account | Restricted on iOS. Only the sender may be rewarded |
| Paid affiliate account with a dashboard | Highest. Application, tax form, approval | Lowest. An ordinary link | Exact, with reporting per partner | Lowest. Everyone is identified | Fine, it sits outside the app |
A dollar figure in the ad breaks Google and Meta policy
The referral ad that would make recruiting cheap, one naming a dollar figure, is the version Google and Meta reject:
- Google treats a promised income as an unreliable claim. Google's misrepresentation policy prohibits get-rich-quick offers and exaggerated return claims, including in on-screen text.
- Meta rejects a named dollar amount with near certainty. Money-making offers are among the most heavily policed categories under Meta's unacceptable business practices standard.
- The referral ad that passes Google and Meta review is the weak version with no dollar figure. Removing the dollar figure removes the reason recruiting referrers with ads looked cheap.
- Google and Meta reject an income claim in an ad even when the claim is true. Ad reviewers judge how an income claim is worded, not whether it is accurate.
- Appealing a rejected ad is slow, and every rejection stays on your ad account's record. Repeated ad rejections bring more scrutiny to every other ad you run.
Nine US legal boundaries this idea has to clear
Recruiting referrers with ads makes three of nine US legal rules much harder to follow:
- The FTC, the US consumer protection regulator, requires an advertised earnings figure to be typical, not merely attainable. In 2024 the FTC brought a case against Arise Virtual Solutions, a company that recruits work-from-home call centre agents, over an advertised "up to $18 an hour" when fewer than 1% of workers earned it.
- The FTC put 1,100 companies selling money-making opportunities on notice for penalties in 2021. The FTC's Notice of Penalty Offenses makes civil penalties of up to $53,088 per violation available.
- Every paid referrer is an endorser under US rules, and every endorsement needs a disclosure that the referrer is paid. The FTC's Endorsement Guides require that disclosure every time, and you cannot supervise thousands of strangers.
- A company can be legally treated as the sender of the referral texts its users send. Robinhood, the share-trading app, settled a claim over its refer-a-friend texts for a reported $9 million under Washington state's Commercial Electronic Mail Act (CEMA).
- Pay referral rewards only for people a referrer brings in directly, keep the referral program free to join, and pay only on sales. Under the Koscot test, the FTC's test for an illegal pyramid scheme, referral rewards have to come from real sales to end users, not from recruiting.
Apple, Google Play, US tax law and sector bans limit how referral rewards can be paid
- Apple restricts who may receive the referral reward in an iOS app. Under the App Review Guidelines, the reward in referral flows is expected to go to the sender, so on iOS you cannot run the friend-benefiting design the research says performs best.
- Google Play restricts rewards given for installing or reviewing an app. Google Play allows referral programs, but a reward paid for the install itself puts the app at risk.
- US referral payouts are taxable income, and the company paying them must report them. From 2026 the IRS threshold for reporting payments on forms 1099-NEC and 1099-MISC rises from $600 to $2,000 under the One Big Beautiful Bill Act.
- Finance, insurance, healthcare and legal services businesses face laws that restrict or ban paying referral fees. Anti-kickback and finder's-fee rules limit, and in some cases prohibit, paying for referrals in those sectors.
| Rule | Who enforces it | What getting it wrong costs | Does recruiting by ad make it harder? | The fix |
|---|---|---|---|---|
| Earnings claims must be typical | FTC, and state attorneys general | Up to $53,088 per violation | Much harder. The ad is the claim | Drop the figure, or publish typical results |
| Endorsers must disclose being paid | FTC Endorsement Guides | Penalties plus the brand damage | Much harder. Thousands of strangers | Force a disclosure into the shared text itself |
| Refer-a-friend texts | TCPA, and Washington's CEMA | Class actions, one settled at $9m | Much harder. More senders, same author | Never pre-compose the message or read contacts |
| No rewards for recruiting recruiters | FTC, via the Koscot pyramid test | Existential. The program is shut down | No change. It is a design choice | One tier, free to join, paid on real sales |
| Apple: reward the sender, not the invitee | App Review | Rejection, and release delay | No change. Same rule either way | Sender-only rewards on iOS builds |
| Google Play: no incentivised installs | Play Console policy | Listing suspension in the worst case | Somewhat. Volume attracts attention | Reward the purchase, never the install |
| Tax reporting on payouts | IRS, from $2,000 a year in 2026 | Penalties and unpaid-payee disputes | Somewhat. Many tiny payees | Collect tax details before the first payout |
| Sector bans on paying for referrals | Finance, insurance, health and legal regulators | Licence risk, well beyond a fine | No change. It is banned regardless | Check before designing anything |
| Platform ad policy on income claims | Google Ads and Meta review | Disapproval, and account scrutiny | Much harder. It is the whole creative | Advertise the program, never the payout |
This is a summary of published US rules and cases, not legal advice. Get a lawyer to look at a referral program before you launch one.
The three versions that work need many customers per recruit
Recruiting referrers with ads is wrong for consumers referring consumers, which is the version most people picture. Three variants of recruiting referrers do work:
- Recruiting distributors, people with an audience, rather than ordinary consumers works. Somebody with an audience produces dozens of referred signups, and you find people with audiences by searching for them and writing to them, not by buying banner ads.
- A marketplace short of drivers or sellers can justify paying to recruit referrers. A referred driver, seller, tutor or host makes the marketplace better for everybody already on it.
- Advertising the referral program to people who already use your product works. Your own users cost nothing to reach, they already like the product, and their recommendations are credible.
- All three working versions depend on one number, how many referred signups each active referrer produces. If each active referrer produces 30 referred signups instead of 2, the ad cost per paying customer falls to one fifteenth.
- The recruiting channel decides the outcome of a referral program, not the size of the reward.
| Your situation | Referrals one recruit produces | How much extra a referred person is worth | Verdict on recruiting referrers with ads | What to do instead |
|---|---|---|---|---|
| Consumer app, one-off or subscription | One or two | 16% more, the one measured figure | No. The worst case | In-app prompts to existing users, free |
| B2B SaaS with a sales-assisted funnel | A few, but high value each | A lot. A warm intro shortens the sales cycle | No. Your buyers are not reached by banner ads | A partner program, recruited by hand |
| Two-sided marketplace short of supply | A few, and each one is inventory | A lot. Supply improves the whole market | Defensible, if you can keep paying as recruits quit | Cap the bounty, pay it in stages on activity |
| Ecommerce with repeat purchase | One or two | Moderately. Repeat purchase carries the reward | No. Codes leak to deal sites and cost margin | Post-purchase referral offer, on-site |
| Creator, publisher or community product | Dozens, if the recruit has an audience | A lot. Audiences arrive already interested | Only for the audience owners, and by outreach | Find and contact them directly |
| Fintech, insurance or healthcare | Varies | Varies, where it is permitted at all | No. Often prohibited before economics apply | Check the sector rules first |
Use Repeat Recorder to rehearse a pitch to a community owner
The only referrers worth recruiting already have an audience and have to be contacted by hand. Repeat Recorder, another of my own apps, records you and plays each recording straight back without you touching the phone.
- Repeat Recorder replays each take automatically and then starts recording again, so you can run the same thirty-second pitch ten times over without picking the phone up.
- An ad-recruited referrer costs about $31 in ads, and about $315 once you account for the 90% who never refer anybody. One community owner who agrees to promote your product is worth more than a month of ad-recruited referrers.
- Takes worth keeping go into favourites and into folders you name yourself, dragged into whatever order you want. The version of the pitch that worked on your last call is still in Repeat Recorder before the next one.
Conclusion: how to recruit referrers without buying ads
- Buying ads to recruit referrers pays for a salesperson rather than a customer, so one recruited referrer costs about $31 and about $315 once you count the 90% who join a referral program and never refer anybody.
- Every ad-recruited referrer who actually refers has to produce about five or six referred signups to break even, and closer to 60 on ordinary numbers. An ordinary consumer produces one or two referred signups, and somebody with an audience produces dozens.
- Advertise the referral program to the customers you already have, and find the people with audiences by hand instead of bidding for them. Pay the reward on the referred person's first purchase, because a $25 bounty paid on signup costs $83 for each paying customer when 30% of signups pay.