Dmitriy Pavlov

As a product manager, you have to figure out, okay, which one of those three do we actually focus on?

What's the most important thing to our customers?

And previously, what we can do is we can look at word counts.

We can look at TFIDF, term frequency to inverse document frequency, which is useful up to a point directionally.

But really, you're not understanding what are the underlying causes?

How do they...

So what we're able to do is actually look at those three issues and say, actually, you know what?

People are way more anxious about installation not working out than anything else.

And the thing that actually is predictive of a positive experience is a positive installation.

We can actually understand with our value, with the Pearson product efficient, how predictive of something is an experience.

And this is kind of new in the space.

And what we're doing is we're looking at things like internal customer support requests.

And we have actually on the telecom side, they have hundreds of thousands of calls coming in.

And an agent has to look at this call and say, hey, this is an issue or this is not an issue.

This is important or it's not.

What we can do is because we actually collect data externally, anything that's open source, we look at things like reviews and competitor reviews.

And what we're able to actually find is that if an agent says, this may be not an issue, we can say, actually, this one issue represents about 20% of the total issues your competitors are experiencing.

So this is going to be a really severe issue.

And furthermore, we can actually catch that before issues come into these companies.

So we can start predicting away really severe issues that they will experience.