Data Driven OS

In 2018 I watched Joel Marion do something to a room of 400 entrepreneurs that I've never forgotten.

He was giving a talk on scaling — ten times bigger, ten times faster. Partway through, he stopped and asked for hands.

Who knows their average customer lifetime value? About fifty hands.

Okay — who knows their LTV at day 0, day 30, day 60, day 90, day 180? Around twenty hands.

Now: who knows that, broken out by product, by funnel, by traffic source, and by affiliate?

Four hands.

The room went from 400 to four. He asked each of them how big their business was. Fifteen million. Twenty-five. Thirty-five. Fifty.

And he just said: I'm not surprised.

Someone asked him how he does it himself. His answer: "I pay a developer and a data scientist 200 grand a year to make it happen for me."

That was the moment I understood what I was going to spend the next several years on. Not because the four had some secret. Because the thing separating them from the other 396 was visibility, and visibility is buildable — but at the time, only if you could afford $200k a year to build it.

That's the part I wanted to change.

Your average customer is worth $150.

Nobody is average.

Underneath that number, customers from one funnel go on to reach $220. Customers from another stall at $95. Same business, same period, same $150 average — and the two groups are not remotely the same customer.

Now apply one CAC ceiling to both. You spend too much acquiring the $95 buyers, because the average says they're worth more than they are. And you refuse to spend enough on the $220 buyers, because the average says they're worth less than they are.

The average didn't lie. It did exactly what an average does. It just isn't the number either decision needed.

This is why "what's our LTV?" is close to unanswerable as asked. The useful version has a scope on it: what are customers from this funnel, this channel, this affiliate worth over time — and what can I therefore afford to pay for more of them?

Do you know what your best-performing segment is actually worth, separately from the blend?

The first purchase gives you a useful piece of information. It doesn't tell you the entire story.

A customer who buys once and disappears behaves very differently from a customer who returns, purchases additional products and continues generating value over time. Looking only at the first transaction can leave a significant part of the customer journey out of the analysis.

That is why customer lifetime value matters. It connects what happens at acquisition with what happens afterward, giving operators a broader view of customer economics.

For businesses focused on sustainable growth, the question isn't just how many customers you acquired. It's what those customers actually do after they arrive.

Save this for your next conversation about acquisition economics.

Someone asked me what founders should stop bragging about, and the answer came out faster than I expected.

Customers per day, and revenue.

Both are on every stage, every screenshot, every post. And both can go up while the business gets worse, which is the whole problem with them as a scoreboard. I've seen operators celebrate a record volume month that made them less money than the quarter before, and genuinely not realize it for another two quarters.

Volume is not an achievement. It's an input. You can buy it. Anybody with a credit card and no discipline can post a record month.

The actual flex — the one almost nobody posts — is how low your volume can be while your profit is at its highest. That's a much harder number to hit. It means your unit economics are so good you don't need the volume to make the math work. It means you could survive a 20% rise in acquisition costs without changing anything, while your competitor has to shut ads off.

That's the number I'd want on a slide.

Genuine question, because I don't think it's obvious: what would you put on that slide instead of revenue?

A few things I'd check, in rough order of how often they turn out to be true.

You run one maximum CAC across products whose customers repeat at very different rates.

A campaign got cut before the customers it brought in had time to place a second or third order.

Subscription buyers and one-time buyers sit inside the same blended average.

Two channels looked identical in week one, and nobody went back to check whether they still looked identical by day 60.

The LTV figure behind your CAC target was set months ago — before your offer mix, your refund rate and your customer behaviour all moved.

If more than one of those lands, the constraint probably isn't your media buying. Media buying is usually the most competent function in the building. The constraint is what's known at the moment the decision gets made — and at real spend, holding or pushing a budget is a bet with actual money on it.

Which of those is closest to home?

In engineering, precision isn't optional. Small calculation errors can lead to very different results.

That same mindset applies to business. Decisions around acquisition, profitability and growth deserve more than estimates or incomplete reporting. They require reliable information that leaders can actually trust.

Tyler Ryan's background as a NASA software engineer helped shape the philosophy behind Data Driven OS. Build systems that reduce uncertainty, surface what matters and help businesses make confident decisions backed by real numbers.

Learn more at www.datadrivenos.com

Here's a week of YouTube ads from a client's account:

Day 0: buyers worth under $1.
Day 7: those same buyers worth about $17.
Cost to acquire: $5.

Day 0 says you're losing $4 a customer and should shut it off today.
Day 7 says pour everything you have into it.

Same campaign. Same customers. Seven days apart.

The thing is, the day-0 number isn't wrong. It's completely accurate. It's just measuring the wrong moment. And it happens to be the number sitting in front of you when your finger is over the pause button, which is how it ends up making the decision.

He told me something about the channel he'd been running before this, and it's stuck with me. He said if he'd had this working back when he was scaling Facebook, he'd have spent a lot more, because his 30-day value was much higher than he realized at the time.

He didn't lose money on a bad campaign. He lost the upside on a good one, and never knew.

That's the version of this that costs the most and shows up nowhere.

How long do you let something run before you call it?

One of the best results I've been part of would have read as a disaster on a dashboard.

Revenue fell 60%.
Gross profit didn't move.

What changed: we went through their customer journey and found where value was leaking out the back end. Refunds went from $100,000 a month to $50,000. That put $40,000–$50,000 a month straight into profit, without buying a single additional customer.

His words: "our gross profits stayed steady even as revenue dropped 60%."

I keep coming back to this one because every instinct in direct response pulls the other direction. Revenue is the number you report. Refunds show up 30 to 60 days later, attached to no particular decision anyone remembers making. So they get tolerated as a cost of doing business.

They're not a cost of doing business. They're a change in what your customer is worth, which changes what you can afford to pay for the next one.

If your refunds jumped three points this quarter, would you be able to point at which offer did it?

Most people can't. That's the whole problem.