When you see Artificial Intelligence in the news, the topic is typically robots taking over our jobs or the dangers of super-intelligent computers. You don’t see Elon Musk or Bill Gates opining on how AI is improving the way we do data analytics, for example.
But kiwi startup Thematic is doing exactly that. It’s using machine learning technology to make sense of customer feedback forms. Unfortunately that isn’t as sexy a headline as ‘Elon Musk’s Billion-Dollar Crusade to Stop the A.I. Apocalypse’ (an actual headline last year from Vanity Fair).
Despite flying under most peoples’ radar, Thematic hasn’t escaped the notice of Silicon Valley’s elite. The company, founded by husband and wife duo Alyona Medelyan and Nathan Holmberg, recently completed the exclusive US accelerator program, YCombinator, and finalized a US$1.2 million funding round.
What Thematic does sounds relatively simple, but it has real utility for businesses. Most of you will at some point have filled in a customer survey form. Thematic analyzes those results – and in particular the previously hard to analyze free text input – and provides insights.
How does this work? Alyona Medelyan told the Business is Boring podcast that “we automatically find themes” from the raw data and “then we visualise it using easy to understand charts, and let people interact with the data to understand their customers better.”
Specifically, Thematic uses Natural Language Processing (NLP) and a technique called sentiment analysis, which means determining how customers feel based on their language. The AI will do this by classifying text into sentiment categories, such as positive and negative.
As Stuff commenters well know, strong feelings lead to emotionally laden words. Thematic’s AI can interpret such language. Although interestingly, it cannot handle sarcastic comments.
I guess we will know the AI apocalypse is coming when we get sarcastic robots making fun of us. But until then, it’s important to remember that AI isn’t as all-powerful as it’s often made out to be.
AI can be very useful though. One of Thematic’s customers is Sky TV, which has been in the news lately due to its declining subscriber numbers. Sky TV has lost almost 34,000 satellite subscribers in the year to June.
You may think you already know the reasons for this – Sky TV is too expensive compared to streaming services like Netflix, and it has relatively poor streaming apps of its own. Easy. But naturally Sky TV wants to get more specific feedback about its programming, so it can try to at least stem the tide of unsubscribers. So every week Sky TV surveys thousands of customers about their experiences.
Using Thematic, Sky TV was able to input customer feedback data and get back responses “coded with one or several themes and sentiments.” Thematic also created interactive visualizations for Sky TV’s employees to view on their intranet.
As one example of how Sky TV used Thematic, it was able to quantify how customers felt about missing out on a popular sporting event. As I noted in a Stuff column last year, this may become a familiar feeling to all rugby fans in the coming years. Rugby rights are due to be renegotiated in 2020 and a global gorilla like Amazon could swoop in and buy them.
Perhaps to test how its customers might react to that, SkyTV used Thematic to analyze feedback about an unnamed “popular sporting event” it has already missed out on. The feedback from customers showed that this event was “very important to a subset of customers who were very vocal.” What’s more, this theme “spiked at announcement but produced the biggest impact […] when customers expected the event to start.”
Personally, I’d love to see Thematic analyze the comments made on Stuff’s article last week about Sky TV pricing changes. A manual scan indicates people think the pricing changes are like a band aid on a gaping wound.
Of course identifying and understanding the “themes” of customer feedback is one thing. It’s then up to SkyTV to come up with solutions.
Which brings us back to the current limitations of AI. For all its power in analzying and classifying data (such as customer feedback forms), most NLP systems aren’t capable of proactively solving the issues identified.
Machine learning is especially deficient in dealing with human emotions. Take all those unhappy soccer customers that Sky TV lost after it stopped showing Premier League matches. I don’t need Thematic to guess the themes derived from those customer feedback forms. Emotions like anger and frustration, no doubt.
But what can Sky TV do to make those customers happy again, other than paying a huge amount for Premier League rights? We’ll have to wait for the sarcastic AIs to figure that one out.
In the meantime, it’s heartening to see a New Zealand startup providing such a useful AI service to businesses. Understanding customers isn’t easy, since we humans tend to be emotional. Sometimes we need machines to help figure us out.