Referral Program Examples: What the Famous Six Still Do, and What They Stopped Doing

Somewhere in your bookmarks there is a list of referral program examples, and the reason it is still there is that it looks finished. Every entry has a reward, a logo and one line about why it worked. What none of them carry is a date somebody last checked the program still exists in that shape. That gap costs you: you will spend weeks arguing about the size of a reward and no time asking whether the mechanic you are copying still runs. So which parts of that list are still true, and which ones did you already build on?

Six companies fill the same slots in almost every version: Dropbox, Airbnb, Uber, Tesla, Apple and Robinhood. I read five of them on their own pages on 4 November 2026, and the sixth through two secondary sources that agree with each other; the audit below labels which is which. Four are wrong in circulation: Airbnb, Robinhood, Tesla and Apple. The other two still run, but not the way the list describes. Checking a reward against a brand's own page has become routine. Checking whether the program survived its last terms revision has not. The most careful page in this field runs the first check and still prints an offer that was retired in September.

One boundary before the audit starts. The pillar page for this cluster owns the arithmetic of a reward against the acquisition it displaces, plus the rule for when to shut a program down. This page owns the case evidence, and it labels which checks came back verified. If you want the program shape those numbers sit inside first, start with our guide to the customer referral program.

Dropbox is the example everyone opens with, and half the story is missing

Almost every round-up opens with Dropbox: 500 MB per referral, 16 GB cap, both sides get storage. Accurate once, half accurate now. Dropbox's own page on 4 November 2026 splits its customers in two. Basic users earn 500 MB per referral, up to 16 GB. Plus users earn 1 GB per referral, up to 32 GB. Both sides get the storage. A referral only counts when the invited person creates and validates a new account, and a shared folder does not count.

The paid tier now earns twice the reward and twice the ceiling. The program was re-tuned to favor the customer who already pays, the opposite of the free-tier growth story that gets retold. The 16 GB figure everybody quotes is the Basic ceiling, not the program ceiling.

Why did the original work? Three conditions held between 2008 and 2012. Free storage was scarce, so a reward paid in storage felt like more than a discount. The reward cost Dropbox almost nothing because it was more of the product. And the cap meant a heavy referrer eventually ran out and had to buy more. None of those three is a given for you. Storage is not scarce, your marginal cost may not be near zero, and a cap only works if hitting it feels like a step forward.

One number to retire: the 3,900 percent growth figure. Vendor blogs repeat it and none traces it to a Dropbox announcement, filing or named executive.

The six famous referral program examples, graded on what they do now

Each program is graded on three things: what the reward was, what is true on 4 November 2026, and whether the reward cost stayed under the acquisition cost it replaced.

Tier A means I read the company's own page or help center on 4 November 2026. Tier B means secondary reporting confirms the change but the company's own page does not. Tier C means nobody verified it, and anything Tier C either carries that label in the text or stays out.

Six rows, and four of them contradict the version in circulation.

Most of the fourth column reads "no published amount," which is honest rather than lazy. None of the six publishes the acquisition cost its program was meant to replace, and three do not publish the current reward.

Airbnb, Tesla and Apple: the three that get described wrong most often

Airbnb's guest referral is restricted to users resident in Brazil, mainland China, India, Japan, Malaysia, the Philippines, Singapore and the Republic of Korea. The United States is not on that list. What is open in the US is the host program: refer a host, earn a cash reward, eligible locations and listing types only, with amounts that expire on 28 December 2026.

Tesla's vehicle referral is retired in the US, and Tesla runs an FSD referral instead. That is Tier B: I did not read it on Tesla's own US page. Two independent secondary sources, rivo.io and notateslaapp, describe the same change, which is also why I will not print a current credit amount for a car purchase. The car discount every "Tesla grew without advertising" paragraph rests on is not there to quote.

Apple is a category error rather than an outdated fact. The referral product is the Apple Card: new cardholders who apply through a link and buy within 30 days earn 75 USD in Daily Cash, one offer per account, credit approval required. There is no consumer Apple hardware refer-a-friend program. What listicles reach for instead is either this card offer relabeled, or the iTunes affiliate scheme Apple killed for apps in 2018, which was an affiliate program, never a referral.

Robinhood shows why checking the offer is not the same as checking the program

This is the cleanest illustration of the problem, because the strongest page on this subject ran the check and still got the version wrong. joy.so's round-up of 23 referral programs says it checked every offer against the brand's own page rather than a coupon site.

Its Robinhood entry then reads: the referrer gets a free share worth 5 to 200 USD, capped at 1,500 USD a year. It is described as a mystery reward, and the lesson drawn is about the excitement of not knowing what you get.

Robinhood's own support page on 4 November 2026 describes a different program. Referrals must begin signing up on or after 1 September 2026 at 9 AM Pacific, and that effective date resets every referrer's cumulative count to zero. The milestone ladder, as published:

The cap is 15,000 USD per person per year. The friend must open and fund an investing account with a deposit above zero. Rewards must be claimed within 60 days.

Robinhood moved from a variable one-off lottery to a milestone ladder with a hard ceiling, and reset the counter to reward volume from a standing start. That is a switch from acquisition excitement toward retention, the opposite of the mystery-reward lesson every listicle, including the careful one, still draws from it.

Why did a page that explicitly checks offers against brand pages get this wrong? It confirmed the offer existed on a Robinhood page, not that it exists in the current version. An updated support page keeps its old URL, search results keep their old snippets, and the archived version looks identical until you read the effective date.

Put the two claims next to each other and the argument makes itself. Checking an offer against a brand page is table stakes. Checking whether the program survived its last revision is not. One takes a page load. The other takes reading for a date.

Three examples with numbers you can actually check

The six are known for growth stories, not for legible design. If you want to copy a referral program instead of admiring one, look at designs whose terms fit on one screen.

Google Workspace pays per seat, and the promo code rule kills the usual advice

Google's own referral page publishes rates by tier. In the United States, Business Starter pays 8 USD per user, Business Standard 15 USD per user, and Business Plus 23 USD per user. The reward is one-time per new user, capped at 100 users per referred account and 200 users per year. The friend gets a first-year discount through a promo code, one per referral.

The per-seat structure does something most referral designs do not: the reward scales with the size of what you referred. Refer a ten-seat Business Plus account and the reward is 230 USD on one referral, a number you can put next to a sales-assisted acquisition cost.

The trap is in the promo code terms. Google's codes cannot be posted publicly and have to be distributed individually, which quietly forbids the tactic most referral advice opens with: putting your code in a banner, an email footer or a LinkedIn post. A 90 day payout hold is claimed by at least one competitor page, but it appears nowhere in Google's own text.

Hostinger is the honest borderline case: a commission in referral clothing

Hostinger's referral page pays 20 percent of the purchase price of each eligible sale, headlines "earn up to 450 USD for your first referral," gives the friend 20 percent off, sets no limit on referrals and pays out by PayPal, wire or Hostinger balance. A percentage of every sale with no cap is how affiliate programs pay publishers.

ReferralCandy publishes customer referral rates, and no other vendor I found does

If you want referral rates rather than reward structures, one vendor publishes them. ReferralCandy's 2026 examples page lists Toki Kids at a 12.04 percent average referral rate with 517,960 USD in referral revenue and 41 times ROI, Riff Raff and Co at 24 percent, Farm Hounds above 600,000 USD at 35.3 times ROI, and Branch Basics above 1.5 million USD.

They are customer-reported through a vendor platform with an interest in publishing favorable cases, and the page does not distinguish a self-reported figure from an audited one. Useful as illustration. Not benchmarks.

The transfer test: five questions before you copy anything

You cannot copy a program. You can copy a design decision, and only after you check whether the conditions that made it work exist on your side. Five questions, in order.

1. Does it still exist in this form? Does the current version of the terms describe the thing you are about to copy? Check the effective date, not the page's existence. Robinhood's support page still loads. What it describes is a different program.

2. What action triggers the reward? Dropbox pays on a validated new account, not a folder share. Airbnb pays after a qualifying stay. Robinhood requires the friend to fund an account with a deposit above zero. That distinction is where your fraud exposure lives. A reward for an email address is not a reward for intent.

3. What gate blocks self-referral and low-quality volume? Vera Bradley requires a minimum purchase on the friend's order before the reward lands, cheap and effective, per a June 2025 vendor write-up rather than its own page. Google caps users per account and per year. Robinhood caps the annual total. A gate is what makes the growth worth having.

4. What does the reward cost against the acquisition cost it replaces? The arithmetic is a separate job, so I will only say what the audit showed. None of the six publishes the figure the reward should be compared against. If your own program cannot answer it with your own numbers, you are running a discount with no control group. Report the answer next to the other numbers you already track as marketing KPIs.

5. Can your customer reach ten other people with this problem? This decides whether the program works at all, and it has nothing to do with the reward. ReferralCandy's July 2026 analysis of 500 established Shopify stores found that 1.1 percent of advocates made 11 or more referrals and produced 30.2 percent of referred customers. The denominator there is successful advocates, not all customers. The program runs on a small group of people who know a lot of the right people, so judge the design on whether it makes the ask easy for them.

What to copy when you do not have a brand

You cannot copy Dropbox, because you do not have Dropbox's conditions. Storage was scarce, the reward cost near zero, and the cap converted free users into paying ones. Without those three, it is a story about 2010 with a referral page attached.

Start with the gate, not the reward. Vera Bradley gives 20 percent to both sides with a minimum purchase on the friend's order. Small brand, legible in one sentence, and the minimum spend blocks the self-referral loop that eats most programs before they scale.

Then instrument the window. ReferralCandy's data puts the typical first referred sale at 14 days from activation. That is a working expectation for how fast a referral should convert, and it beats a vague sense that referrals are slow. Report it next to your acquisition costs and your referral rate.

Then treat distribution as the actual work. Impact.com's consumer referral research, in the report they distributed in 2024, found that 44 percent of consumers participate in referral programs, and that of the 56 percent who do not, 60 percent have never received a referral code or link from someone they know.

For most companies the binding constraint sits before the reward: the ask never reaches the customer. Every hour arguing about whether the reward should be 10 USD or 20 USD is an hour not spent on the mechanic that decides participation, which is whether a happy customer can find their link and send it without thinking.

That is where I would spend the budget. The reward is the easiest part to copy and the least likely part to be your problem.

The division of labor, and one thing to do this week

This page covered the case evidence. The arithmetic of what a reward costs against the acquisition it displaces, and the rule for when a program should be shut down rather than rescued, are deliberately not repeated here.

If you take one thing from these referral program examples, take the verification habit rather than any of the six designs. Open your own referral page and read the effective date at the top. Then ask whether the version your customers joined is the version the page describes.

The next example you copy will be current on the day you copy it. The question is what happens to it in fourteen months.