Insight Wert
The Leaky Bucket: Why Obsessing Over CAC Is Costing You More Than You Think
Every growth meeting starts the same way. CAC is down. New users are up. The acquisition machine is working. Everyone's happy. Nobody looks at the back door.
This is the leaky bucket problem — and it's quietly destroying the unit economics of companies that, on paper, look like they're growing.
What CAC Actually Measures (And What It Doesn't)
Customer Acquisition Cost is a useful metric. It tells you how much you spent to get a customer in the door.
What it doesn't tell you is whether that customer was worth getting.
A CAC of €40 looks great. Until that customer stays for six weeks, generates €35 in revenue, and leaves. You didn't acquire a customer. You paid €40 for a six-week trial that cost you money.
Now multiply that across thousands of customers, across multiple channels, over twelve months.
The growth chart still goes up. The revenue numbers still look reasonable. But somewhere underneath, the bucket is leaking — and you've been pouring faster to compensate.
The Metric Everyone Tracks vs The Metric That Actually Matters
Here's the uncomfortable truth about CAC: it's the easiest growth metric to improve in ways that make your business worse.
Lower your prices — CAC drops. Run a promotion — CAC drops. Target a broader audience — CAC drops. Acquire customers who were never going to stay — CAC drops.
Every one of those moves can make your acquisition numbers look better while quietly degrading the quality of your customer base.
The metric that actually matters is LTV:CAC ratio — not CAC in isolation. And even LTV is only useful when it's broken down correctly.
Your average LTV is lying to you. Not because someone calculated it wrong. Because averages hide exactly the information you need: which customers are actually worth acquiring, and which ones look fine until month three.
| What average LTV tells you | What you actually need to know |
|---|---|
| A customer is worth €X on average | Which channel produces the highest LTV customers |
| Retention is Y% on average | Which cohort churns fastest — and when |
| Revenue per user is Z | Which product or plan destroys LTV silently |
The difference between these two columns is the difference between pouring into a bucket and finding the hole.
Where the Holes Actually Are
Most leaky bucket problems aren't visible at the top level. They're hiding in the breakdown.
Channel-level LTV variance
This is the most common one. The customer acquired through paid social in Q2 churns at twice the rate of the customer who came through organic search. Same product, same price, completely different lifetime behaviour. Your blended LTV average smooths this out and makes both look acceptable.
The result: you keep investing in paid social because the CAC looks competitive, without realising the customers it produces are fundamentally different — and fundamentally less valuable.
Cohort decay
Month-one retention looks fine. Month-two looks fine. Month-three is where things quietly fall apart. But if you're only looking at overall churn rates, you'll never see the month-three cliff until you've been feeding the top of the funnel for a year.
Plan or product mismatch
In subscription businesses especially, entry-level plans often attract customers who were never going to upgrade — and who churn the moment a cheaper competitor appears. These customers look identical to high-value customers at acquisition. They're not.
None of these are visible if you're only watching CAC.
The Compounding Effect Nobody Talks About
Here's what makes the leaky bucket particularly dangerous: it compounds.
You acquire 1,000 customers. 300 churn in three months. You acquire 1,000 more to replace them — plus additional customers for growth. Another 300 churn. And so on.
The acquisition machine keeps running. The spend keeps going out. The team keeps hitting their "new customer" targets.
Meanwhile, the actual customer base barely grows — because you're running to stand still.
Retention is a marketing problem. Not just a product problem. If your paid channels are consistently producing customers who churn faster than your organic channels, that's an acquisition targeting problem — not a product problem. You're reaching the wrong people. Efficiently.
This is why companies can show strong acquisition metrics quarter after quarter while revenue growth stagnates. The top of the funnel is working. The bottom is leaking. And nobody connected those two facts because they were being measured by different teams with different dashboards.
What Fixing This Actually Looks Like
The fix isn't a new dashboard. It isn't a new attribution model. It isn't even a new retention strategy — at least not yet.
The fix starts with connecting data that's currently sitting in separate systems.
Your ad platform knows where customers came from. Your CRM knows which ones stayed. Your subscription backend knows what they paid and when they left. BigQuery is where those three conversations finally happen in the same room.
Once they're connected, the questions become answerable:
- Which acquisition channel produces the highest LTV at 6 months? At 12?
- At what point in the customer journey does churn spike — and what were those customers doing beforehand?
- Which customer segments are we acquiring at a loss without knowing it?
These aren't exotic questions. They're the questions every marketing team should be able to answer on a Tuesday morning. Most can't — not because the data doesn't exist, but because it's never been connected.
The practical starting point
Before you can do cohort analysis, LTV modelling, or churn prediction, you need one thing: reliable data flowing from every touchpoint into a single source of truth.
Which means clean tracking. Properly configured consent. Ad platform data reconciled with your CRM. No 40% discrepancy between what Google Ads reports and what actually landed in your backend.
You can't model LTV on broken data. You can only model what broken data looks like — and make decisions based on that. Which is exactly how the bucket stays leaky.
The Question Worth Asking This Week
Pull up your acquisition data from twelve months ago.
Find the cohort of customers you acquired in that month. How many are still paying? What's their average revenue to date? Break it down by channel.
If you can answer those questions in under an hour, your data infrastructure is in good shape.
If you can't — if the data lives in three different places, if the channel attribution doesn't match across systems, if "cohort" isn't a filter you can apply — that's where to start.
Not with a new strategy. Not with a new tool. With the foundation.
Not Sure If Your Tracking Gives You the Data You Need?
We built a free tool that checks your website tracking in 30 seconds — consent setup, conversion tracking, server-side configuration, and common GA4 errors that silently corrupt your data.
Run Free AuditFrequently Asked Questions
The leaky bucket problem describes a situation where a company keeps acquiring new customers but loses them almost as fast through churn. The acquisition numbers look healthy, but the actual customer base barely grows because the bottom of the funnel is leaking.
CAC only tells you how much you spent to get a customer in the door. It says nothing about whether that customer was worth getting. A low CAC is meaningless if those customers churn within weeks and generate less revenue than they cost to acquire.
The LTV:CAC ratio compares customer lifetime value to acquisition cost. It tells you whether the customers you acquire are actually profitable over time. A healthy ratio is typically 3:1 or higher. Looking at CAC without LTV is like celebrating a sale without checking the margin.
Different acquisition channels often produce customers with very different lifetime values. Paid social might deliver low-CAC customers who churn at twice the rate of organic search customers. Blended averages hide this — you need to break LTV down by channel to see which sources are actually profitable.
The first step is connecting your acquisition data with your retention and revenue data in a single source of truth — typically BigQuery. You need clean tracking, properly configured consent, and ad platform data reconciled with your CRM before you can do meaningful cohort analysis or LTV modelling.