Surveys are great for getting customer insights at scale, especially those with freeform text boxes which don’t restrict the answers of your participants. On the flip side, it can be quite a pain to extract meaningful info from this, especially if you have thousands upon thousands of answers to sort through.

We recently did some work categorizing survey results and it was brutal, to say the least. But we did learn a lot on how to handle data work like this. Here’s some of those lessons:

1. Use Excel

No doubt it’s still the king of data apps, Excel is the go-to tool for anyone trying to work with data. Everyone knows how to use it, it’s easily available and can do tons of powerful functions without much overhead in learning.

If you fancy using another tool, don’t. Or at the very least, export it into Excel.

Why? Because you need to split the work. In order to efficiently be able to split huge tasks like these, people need to use the same platform. Which is why we use Excel. Also, it’s not online so in the event of loss of connectivity, there’s no issue. And it can handle big data sets fine.

2. Farm out the work

First, look through the data and try to do the work yourself for a bit. I categorized some and I was able to both know how long it takes to do one unit of work, as well as finding out how it’s done. The latter is so I can give clear instructions on how to do the work.

Give each guy one question to categorize. This develops some sort of economy of scale since they will be focused on just that topic.

In this case, I actually let the guys create their own categories from scratch so this helps. One thing I forgot to do was give access to the entire data set so they have a clearer picture of the whole survey and how their question fits in it.

3. Restrict the output

I made a mistake in that I didn’t give restrictions on how many categories there should be and how long they need to be.

The easier it is to cull insights from data, the better you did your job. I should have restricted the number of words per category and the number of categories.

4. Do the work in parallel

My common trick is to get lots of guys to do the work and give the one who does the work fastest more work. Of course, I also check their work to see if they did okay with their categories before assigning the next set.

What I do is give them a bit of the work first, see if that’s done well, then give the rest.

5. Check and clean the work as it’s being done

My mistake here was not doing this immediately as soon as the first guy finished, leading to a ton of work later.

Granted I was busy with other stuff but still, this should have made it easier to do the cleaning up of duplicate categories, fixing misspellings and doing a quick tabulation of the final tally of categories.

If you’re going to do some survey work, make sure that you get guys who know what they are doing and test them first. Then you can farm the tasks to them. Or just let us do it, we already know how after all.

———
Photo credits: henryfaber and musescore