The Measurement Problem Every Analyst Faces (And How to Fix It)

You can't always measure what you want.

You want to know if your customers are happy.

But here’s the problem: you can’t just knock on someone’s door and ask, “Hey, on a scale of 1 to 10, how happy are you right now?”

Well, you could 😕, but good luck getting honest answers. People lie. People don’t know how they feel. And most importantly, you can’t do this for thousands of customers every single day.

So what do you do?

You use something called a Proxy Metric.

And if you learn this trick, you can solve a big data problem: how to measure things that feel impossible to measure.

Okay, But What Even Is a Proxy Metric?

A proxy metric is a stand-in. It’s something you can measure that stands for something you can’t measure directly.

Think of it like this:

You want to know if someone is rich. But you can’t see their bank account.

So instead, you look at their car. Their house. Their clothes. These things aren’t money, but they give you clues about money.

That’s a proxy.

In data, we do the same thing.

You want to measure: Customer Happiness

You can’t measure it directly because happiness lives in people’s heads.

So you find a proxy: Maybe it’s how often people come back to buy again (Repeat Purchase Rate). Maybe it’s how long they stay on your website (Time on Site). Maybe it’s how few complaints you get (Support Ticket Volume).

These aren’t happiness. But they leave clues about happiness.

Let Me Show You an Example

Let’s say you run an online store.

You want happy customers. But you can’t read their minds.

So you pick a proxy: Repeat Purchase Rate.

If people keep coming back to buy from you, that’s a good sign they’re happy. If they buy once and never return, that’s a red flag.

Is this perfect? No. But it’s better than nothing.

Another easy example

You sell furniture. You want to know if your furniture is good quality.

You can’t test every single chair and table. That would take forever.

So you pick a proxy: Return Rate.

If people keep sending your stuff back, your quality probably sucks. If almost no one returns anything, you’re doing something right.

Again, not perfect. But useful.

But Wait. There’s a Catch

Proxy metrics are like shadows. They show the rough shape of the real thing. But shadows can look different from the real thing.

Let me explain with an example.

The “Time on Site” Problem

You run a website. You want to know if users are engaged and happy.

So you track Time on Site. More time = more engagement, right?

Not so fast.

Maybe people are spending a long time on your site because they love your content. That’s great!

Or maybe they’re spending a long time on your site because your navigation sucks and they can’t find what they’re looking for. That’s terrible!

Same metric. Opposite meanings.

This is the danger of proxy metrics. They can trick you if you’re not careful.

So How Do You Fix This?

Great analysts find one proxy metric.

Elite analysts use many proxy metrics together to get closer to the truth.

Here’s what that means.

Instead of using only one number, you track several proxies and look for patterns.

Let’s go back to the “Customer Happiness” example.

Triangulating Customer Happiness

You want to know if customers are happy. So you track:

1. Repeat Purchase Rate (Are they coming back?)

2. Net Promoter Score (NPS) (Would they recommend you to a friend?)

3. Average Order Value (Are they buying more over time?)

4. Support Ticket Volume (Are they complaining?)

5. Time Between Purchases (How often are they coming back?)

Now you’re not just looking at one shadow. You’re looking at five different shadows from five different angles.

If all five metrics point the same way, you can be pretty sure about what’s happening.

If Repeat Purchase Rate is up, NPS is up, Average Order Value is up, Support Tickets are down, and Time Between Purchases is shrinking, then yeah, your customers are probably happy.

But if Repeat Purchase Rate is up but NPS is down and Support Tickets are spiking, something weird is going on. Maybe people are buying because they have no other option, but they’re not actually happy about it.

That’s triangulation.

You’re using multiple proxies to get closer to the truth.

Here’s What You Need to Remember

You can’t always measure what you want to measure.

But you can find something close.

Find the shadow. Follow the clues. Use more than one proxy to get closer to the truth.

And always remember: You’re not chasing the shadow. You’re chasing the thing that cast it.

That’s the difference between good analysis and great analysis.

That’s the proxy metric trick.