You Don’t Have to Be Afraid of Numbers: 4 Simple Steps for Data Analysis
A process-guide for math-phobes like me
The new trend is in “measuring”, “big data”, or “analytics”.
“You can’t improve what you can’t measure” is the battle-cry of this data-driven movement.
But, the danger in this data “boom” is data mis-use or “data waste”.
Numbers Anxiety
Has that ever happened to you? Have you been presented a lot of data, but you didn’t quite understand what it was supposed to constructively mean for business?
I know a lot of us look at reports with financial data, social media statistics or Google Analytics metrics periodically. I’ve seen many reports show table after table, but at the end of it, the Client doesn’t know what he was supposed to do as a course of action, or ends up latching onto irrelevant things.
The beauty of data is that it’s supposed to make you feel more in control, have a better grasp of what’s happening to your brand.
I like that philosophy. I want people to have more peace of mind, or be less anxious, when they’re looking at their numbers. I want to be able to write about data analysis with as little mumbo jumbo as possible*. Mostly because I see a lot of people who get intimidated by “analytics”. I just want people to understand that it doesn’t have to be scary, so I decided to share my simple way of dealing with datasets.
What makes me think I’m qualified to talk about this?
I’m not “good at math”. I’m not a statistician, or a wizard at Excel (even after using it at work for 9 years).
But, I enjoyed and grew up in the market research and digital research fields all those years. I realized that I love analyzing quantitative and qualitative data — surveys, focus group responses, analytics, social media stats.
I know a lot of people I meet day-to-day don’t like math, too. And I’ll assume that that might be the reason they don’t want to analyze data.
But, hey, if I can, y’all can. Or, at the very least, I want to show you how you can hack numbers, too.
Defining “Analysis”
One of my earlier bosses told me that I was “good with numbers”.
That perplexed me for years. Didn’t know how I could suck at math, yet be “good at” quantitative analysis. Now, I realize that data analysis isn’t exactly about school math. Analysis is more about finding patterns and digging up the value behind the rows of data.
To backtrack further, in college, I was a pretty lazy student. My mind naturally tried to find the simplest ways to deal with shitloads of information. I think my job just taught me to keep getting better at it.
At first, when people would ask me to “teach” analysis, it would stump me. A former boss actually believed you couldn’t “teach logic” — that you should have been born with that common sense.
Having to teach others how to analyze data made me break down my own thought process, and organize it into a simple 4-step mental framework that I apparently use when dealing with datasets.
Trust me on this. You just need to:
- List
- Sort
- Cluster
- Track
List — Sort — Cluster — Track.
Memorize it. It isn’t too hard.
Process
1. List
Listing is the drudgery part of analysis.
Whether it’s for creating infographics, or figuring out weak performance points — it starts with collecting all the data that you feel could be important to your story.
When I say “list”, I mean copy or type up all your information. Do it in Excel, or any similar spreadsheet app.
It’ll be easier to manipulate individual cells in spreadsheet apps.
Analysis is still a tedious job, like any other craft. You can’t find the story if you don’t have all possible juicy ingredients in front of you. You also can’t foresee what exact story you’re going to find, so it’s best to start with a hefty base of information. Let the data surprise you. You can really only hypothesize about the findings you’ll see.
Additional reading: Please check out Sarah Slobin’s 7 1/2 Steps to Successful Infographics. She talks about the value of clean data.
*Listing doesn’t just apply to numerical data, but qualitative data, too. In market research, there’s a “codebook” — a fancy way of saying “list a good sample of qualitative responses, and organize them so you could count them better”. It’s really just a more systematic way of figuring out which responses crop up more.
You can also do a word cloud (Check out wordle.net) if you’re feeling a bit lazy. It can give you suggestions about which data is worth listing.
If you’re feeling creative or crafty, you can also write them down. I sometimes do that with qualitative data that I need to wrap my head around (similar to mind-mapping).
2. Sort
This is how you sift and knead the data — the most basic way to elicit any semblance of meaning or story in your information.
It’s “getting acquainted” with your story. When you sort different attributes, you start to get a feel of “strength”, “value”, “weakness”. What attributes matters and what needs work.
Find as many attributes to sort — time, performance, proportion. Type up other figures and dimensions that define your data.
When dealing with qualitative information (i.e. Facebook comments, descriptions of articles), what you normally sort would be frequency — how often the response appears in the sample. This is why the codebook is handy. It’s a documentation of how similar answers are lumped together to see which ideas are mentioned most.
Excel’s Conditional Formatting function is handy, at this point.
I suggest taking full advantage of it.
Why? You can select a whole column of data then it automatically creates color gradients or bars that visualize how strong the numbers are, relative to each other.
3. Cluster
Our brains naturally try to find patterns to streamline information. And, clustering is how you get to look for more angles to your analysis story.
I bet when you looked at your ranked figures, you already tried to classify them — see reasons behind why the strong ones were the strong ones and why the niche ones were niche.
Through listing the attributes, you’ll start to see which ones (objectively) occur together, particularly when driving higher, lower or mid-range numbers. Normally, I list all the attributes of what I listed to the right of the numbers (male, female, time of day, day of week, type of industry).
Yes, more listing. Analysis is still a tedious job — you have to work at it to get to the magic.
After that, I color it in. The “vision” function of the brain processes color before shape. It’s then easier to look for patterns, at a glance (or from arms’ length), through color rather than having to “read” numbers. It’s how I always end up with rainbow grids, when I need to analyze data.
But, this might also be because of the part of me that loves rainbows.
4. Track
Now, you now get to dig up even more stories.
My data philosophy (which my teams are probably sick of hearing) is “data is all about relationships”. A number doesn’t make sense by itself.
Another fundamental way to add another ingredient to your data recipe is time. I like to think it’s how your data starts living. You see it breathe, or move and grow.
A simple way to do this is to set up a periodic session of doing steps 1–3. Take your dataset, and measure the exact same attributes, over time, or various situations. If you used Excel (or a spreadsheet app), just copy and sort title columns to your heart’s desire.
So, say you took your “temperature” a month ago. You get to take it again now. And, hopefully next month. You are now “benchmarking”.
Look at the changes and differences between the data.
It may not seem exciting — but trust me, look at 3 periods worth of data. Sometimes, the initial patterns you saw in the first dataset will change (or won’t), and that’s a story in itself.
Then, you get to ask yourself — what has been happening in that period that made the data shift (or not shift)?
You have an objective prompt to look at the other (possibly external) factors that affect your business or performance. You can even start finding your “average” performance over time, places or situations.
Dig up answers to more questions: Is this better than last week? Is it better than last year? Is this higher or lower than your closest competitor? Is it different for men and women who visit the site? Is the figure similar or different from the same dataset from a different country?
Hoping that this helped, even a bit
In the next days, I’ll be showing samples of how to apply this in different situations. But, for now — List Sort Cluster Track.
Analyzing data doesn’t have to be a scary thing.
You don’t have to be a math whiz. You just need to be logical and to do a little grunt work.
Try it. Tell me about it. Tell me whether I’m missing some steps, or if there are hiccups you encounter.
I just want to help people make numbers make sense. :P
So tell me if this helped do that for you, or how it could be better.
*Inspired by how one of my colleagues, developer Jon Danao would say “If you can’t explain it in layman’s terms, then you probably don’t understand it.”
