LTV Is All You Need

Lifetime value, commonly abbreviated as LTV, is one of the more aptly named statistics out there. Simply put, you tally up the revenue each individual customer has ever brought your company, line them all up, and take the median (and sometimes the mean).

And voilà. You now have your lifetime value.

In layman’s terms, lifetime value tells you how valuable a typical customer is to your company. In that simple form, it already answers the question we started with: how much can I afford to spend to acquire a new customer?

That single number, a point metric in data science terms, is where most companies stop. Which is a shame, because a single number is a castrated version of its greater self.

Lifetime value isn’t a point metric. It’s a graph. Or perhaps more accurately, a time series.

The mountain

Let me explain with an example.

Think about a hike. When you reach the top of the mountain it’s true to say, “I am 14,000 feet up in the air.” That’s a very helpful thing to know. But expert hikers and climbers know that, while true, it doesn’t tell even 1% of the story.

You want to know how long it took to get up there. What the peaks and valleys were along the way. You’d be far more interested in the topographic map of the mountain, one that shows not just how tall it is, but the shape of what it took to get up there. Otherwise, every mountain in the world of around the same height gets treated exactly the same.

Lifetime value is the exact same thing. You might say a customer is worth $10,000 to you, but you’d be stripping out all of the interesting story.

When we calculated lifetime value up above, that number was correct. But it would likely have been different if you had calculated it last month. Or last quarter. Or two, four, ten years ago.

So instead of throwing all that history away, let’s graph it.

Fantastic. Now we can see how fast it’s growing.

But it’d be a crime to stop here. That single line is doing a lot of work. It’s standing in for every customer you’ve ever had, when we all know full well that not every customer is made the same. Some are far bigger than the rest. Some are far smaller.

So let’s split that line up by how big a customer is, which, conveniently, is exactly what percentiles do.

You probably remember percentiles from the doctor’s office growing up: you weren’t just 4’2”, you were in the 70th percentile for your age. Same idea here. Only one in ten ACME customers is worth more than the 90th percentile line. A quarter are worth less than the 25th. And the median is just the 50th: half above, half below.

Now we’re talking.

We have something a lot more interesting. The number of questions I can ask of a simple graph like this, without going any deeper, just went up tenfold. For example:

Are the lifetime values of my customers increasing over time?

They should be, but it’s entirely possible they’re not.

If they’re flat, the typical customer isn’t worth any more today than they were years ago, and that raises some uncomfortable questions. Do we stop giving customers a reason to come back after their first project? Is all of our growth just coming from winning new customers faster than we lose old ones?

And if they’re going down? Houston, we have a problem. Unless, maybe, you made your services cheaper and far more people are coming in, so the typical customer is simply worth less. That can be fine. But it also means what you can afford to spend on a new customer keeps shrinking, so you need ways to acquire them that keep getting cheaper.

Are my best customers pulling away from everyone else?

Look at the range between the median, the typical customer, and the 90th percentile. We probably already knew our biggest customers were important to us. But man, look at that concentration risk increase year after year.

ACME’s 90th percentile customer is worth 7.7 times the median. And the gap isn’t just big, it’s only going to get bigger: draw a straight trend line through each, and the 90th percentile’s is about 11 times steeper than the median’s. Those customers are clearly different from the typical customer ACME goes after.

It makes you think this company might be better off shifting its ICP to go after more of those customers, instead of the 50% that apparently may not be worth the time. Worth further investigation, of course.

So, how much can I afford to spend to acquire a new customer?

Now we have a more interesting question to ask: which customer are we trying to acquire? Just the typical customer? Then our original number was correct. Or do we go a step further and ask how much we could, or should, be spending to acquire a customer in the 75th or 90th percentile? That’s clearly where we want to be spending our time.

Here’s the quick math for ACME, with a 40% gross profit margin, a 25% close rate, and 30% of marketing-qualified leads turning into sales-qualified ones.

The difference is already huge. So are we spending enough? Are we spending even a tenth of what they’re worth? (Hint: most companies aren’t.)

And for extra credit: percentiles split customers by how big they are, but what if you split them by what they bought?

You might have many different product lines. Wouldn’t it be interesting to see the lifetime value of customers who bought a certain product first?

Now we’re really cooking.

You might find that certain types of products are far better for you to be selling. Obvious, yes. But you may also find a few surprises: products that, when they’re associated with an account or happen to be its very first deal, come with a far higher lifetime value.

I call these your seeker missiles. Products with a heavy correlation (not quite causation) to accounts that go on to buy more. At ACME, that’s Launch Pad Repair. Few customers start with it, but the ones who do end up worth the most.

And products are just one way to slice it. What about lead source, or the industry a company is in? The region they’re from? Maybe even your sales reps, or their teams.

Every new slice adds more questions you can ask. Why does my Google advertising find customers who grow into bigger clients better than Facebook does? Why do trade shows tend to be the best? Why are Jim’s accounts worth so much more than Tony’s?

Ah, I could list off fifteen more questions. But I’ll stop belaboring the point, because we have an entire second graph to look at.

If you’re not familiar with the term, a cohort is a fancy data science word for a group. A cohort analysis just means looking at a metric one group at a time. In our case, we group customers by the year they first became a customer, and then ask: what’s the lifetime value of the customers who first bought from us in 2020? What about 2018? And so on.

When you do that, you get a graph like this one.

Each box runs from that year’s 25th to 75th percentile customer. The black line through it is the median, and the red one is the mean. In essence, it’s the line graph from above with every year pulled apart.

And while this is still just lifetime value, it starts asking and answering even more fascinating questions.

Are our customers coming back at all?

If every year looks about the same, even though the older years have had far more time to buy again, your customers probably aren’t coming back. And if the line graph is still climbing anyway, look at the 90th percentile. It’s likely a handful of big customers pulling everything up.

That usually means two things: it’s mostly a one-and-done business, and it spends its time on a very small number of customers who bring in almost all of the revenue. Mmm. Smells like concentration risk.

Did something change in a particular year?

Like ACME’s 2015 and 2016. Maybe your ICP changed. Maybe you stopped going to the trade show where the bigger customers came from. Maybe a couple of your reps left, and their customers have just sat there, never mined at all.

This is where I’d go back to my sales team and ask: is there really a big difference between the customers we won in 2015 and 2016 and the ones we won around them? If the answer is that they’re at least somewhat similar, send your reps back out to those customers and see if there isn’t more business to be had.

Now for the story I promised: how not knowing lifetime value almost killed a campaign with a 20-to-1 return on ad spend.

I was working with a company in the industrial services space, and one of the services we advertised was line striping. You know, the lines on a warehouse floor that keep people safe and meet OSHA standards.

We were among the first to figure out how to advertise it across multiple mediums. Eventually, people picked up on what we were doing and copied it. The competition heated up, and ad prices went up with it, until the return got close to one-to-one. And on a low-cost service like line striping, that’s losing money.

So, inevitably, the campaign ended up on the chopping block.

But I thought, “Well, hold on. Let’s do some deeper analysis.”

We pulled the purchase history out of their CRM and started building out lifetime value. Not just for their customers overall, but for the customers whose very first purchase was line striping. What we found was fascinating.

More often than not, a line striping customer came back for a second project within two years, and that second project was usually way bigger than the first. This company also did a lot of flooring and painting work, and some of these accounts went on to be two, three, four-timers.

Then we went back to the whiteboard to double-check our work (always do this), and the pieces started coming together. Line striping was getting us in the door. And for anyone in the services industry, you know that once you’re on a company’s approved vendors list, everything else gets a whole lot easier. The bigger projects were close to inevitable.

Line striping was our seeker missile. And we were about to kill it.

That changed the whole conversation. A line striping job might be in the low thousands, but we’d been valuing every lead as if that first job was all it would ever bring in. We needed to value them far closer to what those customers are worth over their lifetime. Counted that way, the campaign wasn’t returning one-to-one. It was returning closer to ten-to-one. The first sale wasn’t the end of the story. It was our price of entry onto the approved vendors list, and we could afford to keep paying it.

A year later, we reanalyzed the data to confirm or deny what we’d said. We were completely right, and it actually came out closer to 20-to-1.

Enough story time.

You don’t have to build this yourself. We did it all for you. Drop in a small CSV file and every graph is drawn right in your browser. The file never leaves your computer. Go on airplane mode if you want to make sure.

But if your team would rather build it in-house, there are a couple of things they should take care to get right.

1. Put a gate in front of new customers

   Most people include every single customer the moment they buy. Here’s why that can throw off your numbers.

   Say your company is ten years old, and you just hired us at Moneyball to help grow your business. Because of what you learned here today, you’re able to bid more, and all of a sudden five times as many new clients are coming in.

   But we know that in your business, it generally takes a customer about two years to come back and buy again. If you add all of those brand-new customers to your lifetime value calculation right away, you skew the number lower than it really is, because they haven’t had enough time to buy a second time yet.

   Here’s what that looks like at ACME, exaggerated a little so it’s easy to see:

   Nothing about ACME’s customers got worse. The chart just started grading brand-new customers against ones who’ve had a decade. So a customer only counts once they’ve had enough time to make that second purchase.

2. Beware how big or small a slice is

   A lot of the value of lifetime value comes from all the ways you can segment it: the percentiles, the cohorts, the split by product. Hopefully I was able to illustrate that for you. But there’s one danger you need to look out for.

   Always know how many customers are inside the segment you’re looking at. You may have noticed the counts sitting under the cohort years and beside each product line up above. They matter. You might come across a segment with only five or ten customers in it, next to another with a hundred or two hundred. The smaller one may very well have useful information, but I wouldn’t call it trustworthy next to one that has really proven itself out.

   There’s a whole slew of extra detail on how to know when a segment is big enough to trust, but that’s an article for another time. For now, if you see a population, as we call them, far smaller than the others, take what it says with an extra grain of salt.

Most companies stop at a single number. You now know lifetime value is a graph, and what to ask of it. Here’s where to go from here.

1. Calculate yours

   Drop in a small CSV from your CRM and you’ll get every graph in this article for your own business, without the file ever leaving your computer.

2. Get help reading it

   If you’d like someone to help you interpret the graphs and figure out your next best step, just give us a ring. We’d love to talk with you.

3. Hop on the newsletter

   Enjoy articles like this one? Join our newsletter to get notified when new ones drop, along with the free tools we build and other insights. All learned from the School of Hard Knocks and passed along to you.

   Email address

   Subscribe

So the next time someone asks, “Okay, so what’s your lifetime value?”, I hope you answer with a graph.