In this article I'm going to share my fastest four steps to get you from a mountain of data to something much more useful: great ideas to test.
We live in a world of insane amounts of data. It can be valuable but also often leads to analysis paralysis and never-ending exploration that goes no where.
Well, I've been there. A lot. I’d rather you follow this approach to smash out great ideas 10x faster.
Unfortunately, most people think going from mountains of data to game-changing ideas is easy and get stuck in exploring their data… forever.
There are no answers in exploratory analysis. Only inspiration.
When people first get into exploring their data, they wrongly think:
interpreting data is easy
brilliant ideas will reveal themselves
there must be amazing insights in the data
they can observe and let the answers reveal themselves
If you’ve been there, you know it ends with a slap in the face.
Let me save you the pain. Follow this 4-step approach and I promise, you’ll save yourself days or even weeks of wasted analysis.
Step 1: Find 3-5 undesirable patterns or outliers
Most people face two challenges when they start exploring data: not knowing what to look for and not knowing when to stop. Here’s my advice.
Look for:
Interesting patterns or outliers – trajectories, distributions, relationships, or noticeable deviations from them that you didn't expect or grab your attention.
That are undesirable – if they were DIFFERENT (higher trajectory, not an outlier, etc) that would mean more value for customers and your business.
Stop when:
You've got 3-5. Be sure to document your findings with screen shots, reminders how to reproduce the results, commentary about what you find interesting, etc.
If you're after some more detailed guidance for how to explore data, checkout this Convert article, "How to Turn Mounds of Data into Usable, Meaningful Insights (2024 Guide)" by Uwemedimo Usa. I don't take the exact same approach but step 3-7 are particularly useful for advice on how to explore data.
Step 2: Write 2-3 possible explanations (hypotheses) for each
This is THE crucial step most people miss.
We tend to (a) keep exploring and think an answer is right around the corner or (b) assume we know what’s going on and skip ahead.
Like our cat friend below, both are lazy. Great tests come from deeper thinking!! Not from data or snap assumptions.
So put on your thinking cap. Come up with 2-3 different explanations (hypotheses) for why the data is the way it is instead of the way you want it to be.
Here’s an example:
---
Say you’re in an e-commerce business and found an undesirable outlier: conversion rates drop dramatically going from “Add to cart” to “Checkout page.” Why so much lower than you’d hope?
Two hypotheses:
Customers can’t find the check out page because the UI is confusing.
Customers are using the cart to save products to look at later (rather than purchasing).
---
Do this for each finding from Step 1.
By now, you’ll have 3-5 undesirable patterns/outliers with 2-3 explanations each.
That's up to 15 hypotheses you can test!
Step 3: Trim your hypothesis list
It’s now worth quickly dipping back into data to debunk any obviously wrong hypotheses.
To do this, for each explanation in your list, think if there’s any easily-accessible data that could debunk it.
For example, I might debunk explanation [2] above by checking if users who drop off are also using a “liked products” feature.
Remove any that are easily debunked and move on.
You should now have done some analysis, a whole lot of thinking, and landed with about 5 hypotheses for what’s causing maybe 2-3 undesirable patterns/outliers.
Step 4: Design hypothesis tests
All that’s left now is to design your tests!
My pro tip here is to design hypothesis tests, not solutions.
By this I mean don’t try to think of a way to solve the situation. Instead, think of a way to test if your explanation is correct or not.
For example, explanation (1) above was that customers couldn’t find the check out page. A solution-based approach might be to get a designer to go and redesign the page. But this takes a lot of work that mightn't even pay off!
Instead, a hypothesis-test approach might be to do whatever to make the check-out page button more visible (bigger, brighter, etc).
I hope this example demonstrates how hypothesis tests are faster to design, faster to execute on, and faster to inform whether you’re moving in a valuable direction.
If you're looking for more advice, checkout this Harvard Business Review article, "A Step-by-Step Guide to Smart Business Experiments," by Eric T. Anderson and Duncan Simester.
Wrap up!
Well, that’s a wrap! Here’s the summary.
Next time you’re feeling stuck exploring a mountain of data, think through these steps to get to great test ideas 10x faster:
Find 3-5 undesirable patterns or outliers
Write 2-3 possible explanations (hypotheses) for each
Trim your hypothesis list
Design hypothesis tests
I hope this helps speed up your innovation journey!
Until next time, thanks for reading! 👋
P.S. Would you rather have an expert mine your data and turn out game-changing test ideas at record pace? Because we'd love to help! Click the link below to schedule an obligation free discovery chat.