I’m trying to get into better shape
To track my efforts, I use a scale that will measure body fat percentage. It’s a bit of a pain to configure it every single time to measure my body fat. I have to click buttons a total of 58 times (really, 58 times) to get it to my 5’9” male adult frame. It starts out at adult, then I click “configure” then click to change from male to female, then configure again, and then a click for each half-inch between 3’6” and 5’9”.
But then I finally realized
I really don’t care about the absolute numbers. I don’t particularly trust that the scale is accurate from an absolute perspective anyway (caliper test FTW). All I’m using the scale for is tracking trends. What I really care about is that my overall body fat is dropping over time.
So now I just leave it set to being a 3’6” female and get my measuring done lots faster. Yeah, it thinks that I’m some kind of freak of nature, but the trend shows exactly what I want it to with a lot less work.
Why am I telling you all this?
Because absolute measurements for media monitoring are equally as useless.
Tracking trends is what really matters
Basically what I mean is that the specific numbers you pull out don’t matter; what’s meaningful is the trends and patterns that the data represent.
Let’s say that I’m looking into how often my brand is mentioned in social media, and my monitoring system tells me that I’ve got 20 positive mentions in the past month.
On its own, almost useless.
Here’s what I really need to know
1.) Is that better or worse than last month?
2.) Is that better or worse than my competitors?
Here’s what I need to answer those questions
1.) Multiple data points (either spread in time or across companies – or better, both).
2.) Consistent measurement methodology.
Comparison is key
By comparing, and in particular comparing over time, you can actually see if things are better or worse – and know if what you’re doing is making any difference at all.
Comparison also helps to wash out measurement error – as long as the methodology is consistent, and you’re not missing some large block of important data, you can see whether you’re being effective or not – even with the imperfect science of text analysis.
(As an aside, if you’re interested in delving into the whole realm of “accuracy” – please see my Symposium talk.)
I can’t overstate how important this is
The entire purpose of text analytics is to save time and resources in providing valuable information and insights; don’t throw that away by focusing on snapshots.
You’ve got to start by tracking trends and patterns overall before you go looking deeper. By starting with a general view and gradually zooming in, you avoid the trap of asking and pursuing questions that turn out to be irrelevant.
So, next time someone asks you “how many mentions did we get this month?” – make sure your answer includes a point of comparison like, “as compared to … , we’re doing…” just like I say “oh, good, compared to last week, I’m down 1% body fat”.
But, tell me, honestly….
Does this blog make my butt look big?
Seth Redmore has over 15 years of experience in product management and 10 years of experience in text analytics – from the perspective of a user as well as a vendor.