Referral Growth

For a long time the only way to get new users was to buy them. Money went into search ads, banners went up, booths went out at conferences. The arithmetic was simple. Double the spend and roughly double the signups arrive. Stop the spend and the signups stop with it.

The trouble is that the price per user climbs over time. As more companies bid on the same search term, a single click costs more, and people who arrive because of an ad tend to leave about as easily as they came. The money it takes to land one customer is called customer acquisition cost (CAC), and when that figure rises from 30 dollars to 70 while the revenue that customer leaves behind stays flat, there is a point where every new signup makes things worse.

The monthly review kept landing in the same place. Eight hundred new teams this month, 48,000 dollars in ad spend. Someone asks what it would take to get 1,600 next month, the answer is 96,000 dollars, and the meeting ends with the conclusion that without the budget there is no growth. As long as signups are tied to the budget line, that meeting comes back unchanged every month.

Bought traffic has one more property worth naming: it doesn't accumulate. Spending 48,000 dollars to land 800 teams this month does nothing for next month. Set next month's budget to zero and new signups drop close to zero, because last month's money bought last month's users and nothing else.

Meanwhile, some products were adding users with almost no advertising at all. Looked at closely, they had one thing in common. The people already using the product were pulling other people in. Sharing a document meant inviting someone, the invited person had to create an account, and that person then pulled in their own team. Signups were tied to the number of existing users rather than to a budget.

What separated those products from the rest was rarely a clever campaign. It was that using the product properly required pulling someone else in, so every additional user created a small amount of pressure on the people around them.

Referral growth is the name for treating that pattern as something you design rather than something you hope for. Where the other branches of growth widen one step of the funnel, this one folds the end of the funnel back onto the front. When people who left the funnel produce the people entering it, the intake grows with the user base instead of with this quarter's ad budget.

Five terms sit under referral growth. They split into what you build by hand (design) and what sits underneath and has to be understood before you build anything (structure). Structure comes first in practice. Attach a beautiful invite screen to a product nobody has a reason to share and nothing happens.

Design has two terms, separated by what they touch. One creates the reason to invite; the other creates the path for doing it.

1. Double-sided incentive — a reward that goes to both the person who refers and the person who accepts. Rewarding only one side makes the asking awkward, because sending a link now looks like collecting five dollars at a friend's expense. Rewarding both sides turns the same action into handing someone a gift. Pagedeck gives one month of the paid plan to the inviting team and one month to the invited team. There's a second decision here: when the reward lands. Pay out at signup and you get a wave of accounts that never come back. Pay out after the invited team creates its first document and the volume drops while the share of people who stay goes up. The type of reward matters too. Cash-like discounts move fast but attract people who take the discount and leave, while rewards that only exist inside the product — extra storage, an unlocked feature — move slower and bring people who stay.
2. Invite flow — the sequence of screens a user passes through between deciding to invite someone and actually entering their address. Where the button sits, how many taps it takes, whether you pull in a contact list or hand over a link to copy: all of it lives here. It gets its own term because identical rewards produce results that differ by multiples depending on this sequence. An invite button buried in settings produces 40 invitations a month. The same button offered at the top of the screen to someone who just opened a document alone produces 400. The intent is the same; the distance to reach it is not.

Structure has three terms, and together they decide whether a product can grow this way in the first place.

1. Network effect — the property where each user gets more value as the number of users grows. A messaging app is the clearest case. Installed alone it is useless; once ten people you know are on it, there is finally a reason to open it. In a product with this property, users invite others without needing a reward, because inviting is in their own interest. A product where one person alone gets the same value either way — a personal expense tracker, say — has almost none of it.
2. Cold start problem — the other face of the network effect. Value requires people, there are no people at the beginning, so there is no value, and without value nobody shows up. Open a secondhand marketplace and the first week has twenty items listed, so buyers leave empty-handed, and with no buyers, sellers stop listing. The usual fix is to shrink the board. Instead of launching nationwide, run it inside one neighborhood, one campus, or one company until that narrow slice reaches density.
3. Word of mouth — spread that happens with no reward and no invite button, because someone liked the product enough to mention it. There is less to design here, and in exchange these users tend to stay the longest. It isn't entirely out of your hands, though. Giving the product a name that is easy to say out loud, and making sure a shared screen renders properly for whoever receives it, are both things you can work on.

Here is how the five move together. Structure supplies the reason to invite, design turns that reason into an actual action, and the people who arrive that way go on to invite the next group. How hard that wheel turns is counted with the viral coefficient, the average number of new users one user brings in. If 100 users bring 30 people, it's 0.3. Above 1 the user base grows on its own without advertising, but very few products get there. Even 0.3 to 0.5 cuts the number of people you have to buy by that much, which is why the number is worth tracking long before it approaches 1.

"Thirty-four percent of this quarter's signups came in through invite links, and our ad spend matched last quarter."

Intake went up without more money, so the referral side is actually turning.

"Signups doubled after we launched the reward, but one-month retention dropped by half."

The reward attracted people who wanted the reward rather than people who wanted the product.

"Moving the invite button out of settings and onto the document screen is small enough to fit in this sprint."

A proposal to change the position of the invite flow rather than touch anything structural.

At Pagedeck this comes up every other week in the growth review. Ravi puts the number of teams that arrived through invite links next to the amount paid out in rewards. He shows two lines rather than one because signups going up isn't by itself good news. Last quarter 320 teams came in through invitations, 90 of them were still active a month later, and dividing the cost of the free paid-plan months by those 90 made referrals more expensive than ads.

His response was not to raise the reward but to move when it pays out. After the change, both sides got their month only once the invited team created three documents. Volume fell from 320 teams to 210 while the number still active a month later rose from 90 to 110.

Mia watches something else in the same meeting: who is sending the invitations. Teams that have built more than ten documents sending invitations means the product is delivering value, but accounts that did nothing on day one firing off ten invitations each means the flow is being worked for rewards and needs a gate. Theo made that distinction possible by logging which screen each invite link was pressed from. Grace closes with the same question every time: if every reward were switched off tomorrow, how many invitations would still go out? If the answer is close to zero, what the team built is a discount campaign rather than a growth structure.

The most common mistake is treating referral growth as the job of bolting on a reward program. What's visible is the invite banner and the coupon, so that reading comes naturally. But a reward nudges someone who already wants to share; it cannot manufacture the wish. Put five dollars behind a product nobody would recommend and the only people passing the link around are the ones who need five dollars, while the people receiving it redeem the coupon and disappear. Six months of that leaves a budget going out every month and a per-user referral rate that hasn't moved. The order has to be reversed: first check whether anyone at all is sending invitations with no reward attached, and if some are, aim the reward at widening that behavior. Switching every reward off for a month is the fastest check available. If invitations hold at roughly half, the product gives people a reason. If they fall to nearly nothing, the reward had been standing in for that reason all along.

The second mistake is counting referred users as free. No ad was bought, so the cost looks like zero, and the reward usually books as a product discount rather than marketing spend, which keeps it out of sight. The real total includes the plan months given to both sides, the engineering time spent building the invite flow, and the work of filtering out accounts created to farm rewards. Leave those out and referrals always look cheaper than ads, which pushes the team to raise the payout regardless of how many of those users stay. Divide by the number still around a month later rather than by the number who signed up, and the ranking often flips. Which channel is cheaper is something you calculate, not something decided in advance.