Any business can pay its users to stay.

Say we run InvoiceViz, a SaaS-based invoicing solution that costs $99/mo to use. If we send our users a $100 check each month that they choose to stay, then they are very likely to stay indefinitely. We will go broke, but, nevertheless, our users will stay.

Now, going broke isn’t a desirable outcome, so let’s look at some alternatives.

Any given user, at any given time, has a certain amount of remaining value to us if we just leave them alone.

If we estimate that a given user will stay subscribed to our $99/mo service for three more months and then leave, we’ve got this much remaining value in that user:

$99 * 3 = $297.

If we do nothing, then that’s what that user is worth to us.

But what if, 60 days from now, we send that user a check for $50?

$50 might make our user stay, right? Let’s say sending such a check has a 30% chance of making them stay two months more than they would have otherwise. Then, we can expect the remaining value of that user to be:

$99 * (3 + (30% * 2)) = $356.50

That’s better! By spending $50, we made $59.40 more than we would have otherwise, for a net gain of $9.40.

As long as we pay users a little less to stay than we profit by having them stay, we come out ahead.

That’s fantastic! We’ve just figured out a money machine that will pay money indefinitely, right? Well, there’s a problem with this machine. For one, the effectiveness of this bribe will diminish over time. The first incentive might buy 2 more months. A second might buy one more month. With a third incentive, the user knows what’s coming so they stay just long enough to take our money and run. Oops, now we’re bleeding money.

Because retention incentives only work so many times, we need to find ways to stretch them farther.

How can we keep more users, longer, for the same amount of money? The main thing we need to do is increase the perceived value of our incentives without increasing the real value of the incentives. If we give users something that feels like a million bucks, but only actually costs us ten bucks, then everybody wins.

Behavioral economics can help us structure incentives to be more effective.

Turns out, humans are notoriously bad at gauging the value of something in a vacuum. When we estimate the value of anything, we can only do it relative to something else that’s similar. This is why, for example, we now routinely pay $3 for a cup of coffee just because we bought it in a fancy-looking coffee shop.

We can use this tendency to take our $50 retention incentive and make it feel like $51, $75, or even $200 to our end users.

First, we can chop our incentive up into pieces.

If we take the same $50, dice it up, and hand it out in chunks, then it can feel like, say, $55 to our end user. But we can’t just dice it up into arbitrary chunks and fling it around haphazardly. We need to be deliberate about how we hand it out in order to maximize its effect. In fact, if we’re too sloppy about handing it out, we can actually subtract perceived value instead of increasing it.

Handing out incentives in increasing increments makes them feel larger.

Instead of giving out one incentive for $50, we’re going to give out two: one for $15 and one for $35. We’ll end up spending the same total amount of money, but will generate more goodwill in the user. But beware. It turns out if we give incentives out in descending value our users will actually hate us. They’ll hate us so much that we’re actually better off giving out a one-time $35 incentive than giving out a $35 incentive followed by a $15 incentive. So lets not mix up the order, ever.

Accelerating incentive schedules feel like a money monsoon.

Just like accelerating a car onto a freeway is more thrilling than cruising at speed on a freeway, receiving accelerating incentives is also more exciting than receiving steady incentives. So, instead of paying $5 once a month for five months, we might pay on a schedule like this:

Week 8: $5
Week 14: $5
Week 17: $5
Week 19: $5
Week 20: $5

This way, we end up paying the same total amount of money over the same total amount of time, but it feels to our user like it’s just raining cash towards the end of the schedule.

Surprise incentives feel like a social contract, planned incentives feel like a business deal.

When we hand an incentive to a user, we want to trigger something called a reciprocity effect. In short, that means the user feeling some sort of social pressure to “pay us back,” even though we didn’t ask them to. Research shows that surprise gifts feel like a social contract and trigger a reciprocity effect. Planned gifts (like the kind we see in frequent flier plans) feel more like a business transaction. Business transactions do not trigger a reciprocity effect. So, one easy way to stretch our $50 is simply to not announce it ahead of time.

Only give $50 to people for whom $50 is significant.

To put it bluntly, if we give $50 to a millionaire she won’t give it a second thought. But if we give $50 to an unemployed shoe shiner, he’ll definitely remember us. The easiest way to make our $50 stretch is to only give it to users who will care.

If the incentive feels like a key for special access to something exclusive, that’s even better.

In addition to the $50 incentive, let’s also sign the user up for the special “power user newsletter”. The power user newsletter is only available for those in the $50 Special VIP Gift Group. Now our user not only thinks they’re getting more cash than the next guy, they also think they’ve entered an exclusive club. The only cost to us is actually producing a useful, no-bullshit newsletter. And that’s probably a good idea anyway so it’s a win-win.

We should also hand incentives to our users in a different form that’s harder to compare to cash.

Handing out incentives as cold cash is logistically easy and certainly appreciated by users. But cash is too easy to compare to the value of other things. Maybe that $50 is as much as the user spent on dinner last night. Or it’s two hours of their salary. Or it’s one hour of parking in Manhattan. Whatever the case, our cash gift is anchored to all the other financial interactions that the user has in daily life. And we’re not at all in control of what that anchor is.

But, we can instead spend $50 on something that’s harder to compare to cash. We could instead send a gift certificate to a high-end restaurant. Or we could send an R/C Helicopter (like New Relic keeps wanting to send me). Or we could offer tickets to a sold-out event. All of these things are better than $50 cash because they’re anchored to something that feels “premium” or special. And they’re all things that are hard to compare to our daily Starbucks coffee and so are difficult to gauge the value of. When choosing the thing we want to give, we just need to make sure anything a user will naturally compare it to seems expensive. So, for example, if we’re giving a gift certificate to a restaurant, we should make sure it’s an expensive restaurant and not something low-tier like fast food.

We must have believable motives.

People hate to think they’re being manipulated. And many of these recommendations ride a very fine line between stretching our incentive money and downright manipulating people. Do not cross over that line. Not only is it unethical, it is also counterproductive. If our user feels like they’re being manipulated, then all the boosting effects of these tactics will evaporate. In fact, we might even subtract value.

So, what’s the best way to stay on the right side of that line?

Be genuinely interested in making the user feel good.

If we approach this exercise with a user-focused mindset, then the odds of doing something wrong plummet. Let’s think of this exercise as “how can we stretch the $50 we have to make the user feel as appreciated as possible” rather than “how can we squeeze as much money as possible out of this user?”