Amazon Music—Listening to the Wrong Customers
Companies love to say they’re “customer-obsessed.” They pour millions into surveys, feedback loops, and user research—all in an effort to make better decisions. But what happens when they listen to the wrong customers?
Amazon Music found out the hard way.
A while back, Amazon made a seemingly strategic decision to expand its music library from 2 million to 100 million songs. Sounds like a win, right? More music, more choices, happier customers.
But there was a catch.
For customers on the standard tier, Amazon removed the ability to choose exactly what they wanted to listen to. You could request a specific artist or album, but after the first song, Amazon would shuffle in “similar” music.
So if you asked for Hans Zimmer, you’d get one track from him… followed by John Williams, James Horner, or whoever the algorithm thought you should like. Not exactly the experience people expected.
Customers weren’t happy. And they let Amazon know—by canceling subscriptions, switching to competitors, and venting online.
What went wrong?
The Danger of Misreading Customer Data
If you’ve ever worked in a company that makes product decisions based on customer feedback, you might be thinking: Wait a minute. Amazon wouldn’t make a decision like this without research.
And you’d be right.
Amazon didn’t roll this out blindly. They likely sent surveys, ran focus groups, and analyzed data before making changes. In fact, they even stated that customers asked for this new experience.
But here’s the problem: not all survey data is reliable.
Let’s talk about the “politeness factor.”
If you’ve ever taken a free sample at Costco, you’ve probably been asked, “Would you buy this product?” And chances are, even if you had zero intention of purchasing it, you gave a polite answer:
"Oh yeah, this is great!"
Surveys suffer from this same issue. When asked, “Would you be willing to pay more for access to a bigger music library?” many customers probably said yes—not because they actually would, but because, in theory, it sounds good.
But when it came time to pay up? A significant portion of those customers didn’t follow through.
Understanding Survey Bias: The Five-Point Scale Trap
To really understand what went wrong with Amazon Music’s decision, we need to dig into how customers respond to surveys—specifically, the five-point scale used to measure likelihood to buy.
You’ve probably seen this kind of survey before:
How likely are you to purchase this product?
1️⃣ Not at all likely
2️⃣ Unlikely
3️⃣ Neutral
4️⃣ Somewhat likely
5️⃣ Extremely likely
Companies love to use this scale to predict customer behavior, but here’s the catch: not everyone who says they’re “likely” will actually buy.
Here’s what the data tells us:
Only about 64% of those who say they are extremely likely (5) to buy actually follow through.
Only about 42% of those who say they are somewhat likely (4) will actually make a purchase.
That’s a massive drop-off.
And when companies—especially those making big product or pricing changes—fail to account for this reality, they end up overestimating demand and making decisions based on numbers that look better than they actually are.
These percentages fluctuate depending on industry and price point, but they serve as a safe baseline.
How This Played Into Amazon’s Mistake
Let’s say Amazon Music surveyed its customers and asked:
"Would you be willing to pay more for access to 100 million songs?"
Plenty of people probably responded with a 4 or 5—because, in theory, more music is great. But when it came time to actually upgrade their subscription, many didn’t.
If Amazon had adjusted their projections based on real-world buying behavior instead of raw survey responses, they might have realized that:
✅ Fewer people would upgrade than expected.
✅ More people would cancel than they anticipated.
✅ Negative cannibalization was a real risk.
Instead, they rolled out a change that ultimately backfired—and had to reverse course.
Positive vs. Negative Cannibalization: A Business Lesson
What Amazon was hoping for was something called positive cannibalization.
Here’s how it works:
A company introduces a new product or pricing tier.
Some customers move from the old product to the new one.
Because the new product is more profitable, the company makes more money overall.
This happens all the time. Apple releases new iPhones knowing people will upgrade. Netflix introduces ad-supported tiers expecting some users to shift but also expecting new subscribers.
But there’s a downside—negative cannibalization.
That’s when a new product or pricing change doesn’t bring in new revenue—it just drives people away.
Amazon’s hope was that many free-tier users would upgrade to the new Amazon Music Unlimited plan. But instead, many just left, taking their money to competitors like Spotify and Apple Music.
When a business decision leads to:
Lower revenue per user
Fewer total customers
Higher churn than expected
That’s negative cannibalization—and that’s exactly what Amazon Music experienced.
A note on negative cannibalization: It probably goes without saying that you want to avoid negative cannibalization. However, it may surprise to you hear there are rare instances where it’s the right move, and that’s to keep the competition at bay. If you’re in a highly competitive market, or new entrants are showing up regularly, it’s not a bad idea to launch a product or service that ensures your customer stay with you instead of going over to your competitors.
Why This Matters for You
At this point, you might be thinking: Okay, but what does this have to do with me? A lot, actually.
The core lesson here isn’t just about Amazon—it’s about how we evaluate information and make decisions.
Too often, whether in business or in our personal careers, we look for data that supports what we already believe instead of trying to disprove our assumptions.
Albert Einstein was famous for doing the opposite. When he developed a new hypothesis, he didn’t search for evidence that proved him right—he actively tried to disprove himself. Only when he failed to do so did he trust that he was on the right track.
Imagine if Amazon had approached its survey data the same way:
Instead of just asking, “Would you pay more for a bigger music library?”, they could have tested:
“If we made this change, would you stay subscribed?”
“If you had to choose between this and a competitor, which would you pick?”
Anytime you see promising survey results, ask yourself:
Are these numbers reflective of real behavior or just intent?
Have we accounted for drop-off rates?
Are we testing for what people will actually do, not just what they say they’ll do?
Misreading customer data isn’t just an Amazon Music problem—it’s a common pitfall across industries. And the better you get at identifying these biases, the smarter your decisions will be.
Final Takeaways
Not all feedback is equal. People say one thing but do another—especially in surveys.
Understand the impact of change. Are you moving people toward a better experience or driving them away?
Challenge your own ideas. Don’t just look for data that proves you right—actively try to find what might prove you wrong.
Amazon eventually walked back its changes, proving that even the biggest companies can misread their audience. More importantly, it also shows the willingness to be humble and admit when we got something wrong. The lesson here applies to all of us—whether you’re launching a product, planning a career move, or making an important life decision.
Be thoughtful. Be curious. Data matters—but only if it’s interpreted correctly. And don’t be afraid to test your assumptions.
When you do, you’ve done the calculations necessary to reduce potential risks, giving you even greater confidence when making a change.