If your company has an online presence that your customers interact with, they expect to be able to converse with your site or app.
This week, I’ll explain what “conversational” data science is and why it’s so important to your customers. Moreover, I’ll give a few examples of websites you might visit every day where it’s done really well. I bet you never thought about how well these sites converse with you!
What is conversational data science?
Conversation data science is definitely not Hal or some prescient A.I. that responds to your every whim. At their simplest, conversational algorithms ask you for input and do something (hopefully magical/predictive) with it
For example, when you first create a new Netflix account, they “onboard” you by asking you to rate a bunch of popular movies/shows. That’s the first half of the “conversation”. Yours. What would you do if you clicked around liking movies and shows for five minutes and Netflix didn’t personalize movies to your preferences?
Netflix takes your information and uses it to rank the shows it thinks you want to watch based on other users just like you – this is called Collaborative Filtering. If they didn’t, you’d click that little [x] for wasting your time, and you’d probably never come back.
But they do pay attention to your clicks, and as you continue the “conversation” by watching movies and otherwise interacting with their site, they respond in kind by constantly updating your predicted ratings for movies and shows. This makes finding the next thing to watch a piece of cake.
You may have thought about the convenience of these interactions, but have you ever thought about what would happen if Netflix didn’t respond to your input and actions? Immediately the conversation would switch to a one-way discourse, and no one likes that, on- or off-line
Yet this is exactly what a lot of companies are doing.
Here are some examples:
- Ever get a marketing email from a company that knows a lot about your purchase behavior but the products in the email don’t get your gender right?
- Sometimes personalization stops at product teams and doesn’t make it to the marketing team.
- Ever seen product recommendations showing “most-popular” items and wondered how in the world those items could be popular and with whom?
- Before 1:1 personalization is possible, many companies use popular products for recommendations site-wide. The problem with this approach is that customer segments often have conflicting ideas about what’s “popular”.
- Popularity works much better as a recommendation technique when we break it up, for example into segments, such as what’s popular for each of these three age ranges: 18-30, 31-50 or 51 and older.
- Do you regularly visit a news site and look for the same types of articles, yet each time you come back, the site pretends it’s never seen you before? Same uninteresting content to sort through with the same irrelevant ads?
- Data scientists can extract topics and concepts from text, allowing us to recommend items similar to what people are already reading. This technology has been around for long enough that readers expect it. But that’s not the whole story …
- News recommendations are tough. Imagine a person is only interested in sports and never reads current events; if we were to personalize their news feed, what threshold would we use for recommending content regarding something like a nearby natural disaster?
Getting personalization right is a multi-tiered process, going from no personalization whatsoever to simple, to robust and companywide.
When done right, and without big product changes, we generally see a 30% lift in customer engagement. Depending on the business model, this sort of lift can drive big revenue across multiple channels, from increased fees with brand partners to higher customer lifetime value
Hit reply if you’re interested in talking about how we can help you increase customer engagement and revenue by implementing conversational data science in your business.
You can read more about conversational recommendation here, replete with example screenshots.
What is Emotional A.I. and how is it Embracing Computing?
The next step on the evolutionary roadmap of the computer is emotion recognition. A.I. provides the digital equivalent of a computer dipping a toe into an ocean; add in emotion, and computers can jump on a boat and swim away from shore
How is A.I. influencing the Alcohol Industry?
A.I. can power ‘digital tasting systems’ that are becoming ever more sophisticated so that one day, we will see the opinions of human experts being combined with the entirely objective feedback from an A.I. tester so that beers and spirits can be generated with ever-increasing quality.
Gartner 2020 Magic Quadrant for Data Science & Machine Learning Platforms
Last week, Gartner released its highly-anticipated report and magic quadrant (MQ) for Data Science and Machine Learning Platforms (DSML). The names remain the same as in the 2019 MQ and 2018 MQ reports, reflecting a more mature understanding of the DSML field, but the contents -especially the leader quadrant- have changed dramatically, reflecting accelerating progress and competition in the field.