Why sentiment scores need context

In an idea world you could just pull up a screen and it would show you a perfect single sentiment your audience was feeling- thrilled, happy, neutral, angry- and exactly what you needed to do in order to encourage or fix it.

Unfortunately, technology is not quite there yet— but we do have a pretty good start. Let’s get into it.

Got questions or something we missed? Find us on Twitter @UnionMetrics.

What sentiment looks like

If you’re using Union Metrics analytics, sentiment looks like this:

Screen Shot 2018-05-04 at 3.56.43 PM

That’s a recent example from our Twitter account. Looks great, right? We can just move on! Not if you want to get the most possible out of sentiment reporting.

It’s important to recognize what the negative sentiment is about, even if it’s small; that’s how you catch issues before they snowball into something bigger. And it’s equally important to make sure things have been properly classified- positively or negatively- since robots still aren’t great at detecting sarcasm (we’ll get into this in the next section).

Digging in

So what were the negative tweets? One was a garbled translation that didn’t make much sense, so it could be classified as spam. The other was someone who didn’t like seeing a promoted tweet in their timeline, which is fair, if not completely under our control.

When it comes to negative sentiment,  you want to concentrate on constructive criticism from customers, fans and followers you can use to build a better relationship with them as well as improving your business processes and products.

If someone is just trolling, that’s only constructive if you notice and overall increase in volume directly related to your content or timing. What have you changed recently that seems to be attracting trolls? Does it make sense to change it back?

Some tweets that are classified as negative aren’t though, and if you’re using our analytics you can click on the face to change it to reflect this.

You also want to dive into the positive tweets: Are they really positive? What kind of feedback are you getting there? Some tweets might get classified as positive when they’re really more neutral, or someone might be a big fan of your brand so they make their feedback really upbeat. You still want to make sure you take that feedback into account to keep that person a fan of your brand.

Keeping an eye on the topics and sentiment related to them over time is also important, as is noting the percent change in the intensity breakdown. If you focus on different time periods, how do these things shift? Can you point to a specific thing in your strategy- the launch of a new campaign, a specific piece of content or format you started experimenting with- that influenced a shift either good or bad?

What else can sentiment do?

We’ve written a lot about what sentiment can do for your brand, including:

We wrote up a comprehensive, updated piece about all of these recently which you can find here.

As always, let us know if you have any questions!

Photo by Nathan Dumlao on Unsplash