AIBusiness recently did an interview with Catherine Havasi, CEO and Co-founder of Luminoso, a Cambridge, MA-based text analysis and artificial intelligence company, spun out of the MIT Media Lab and its crowd-sourced Open Mind Common Sense project. Luminoso empowers companies to capture, measure, and act on customer feedback in real time to improve experience, product offerings, loyalty, and strategy.
Below you will find the interview with Catherine Havasi, sharing her thoughts around artificial intelligence in the enterprise, how to succeed in this evolving world of technology, and looking into the future of the coming year.
In your presentation at the AI Summit you raise the question: “how has the fire hose of available data changed the business world?”, could you elaborate on this?
“It’s a given at this point that it’s easier for customers to reach out and talk to a brand through chat, social media, reviews, or email. Almost everyone’s done it at this point – and it’s a huge wealth of data for brands but also a challenge”, Havasi says.
“Consumers who lead feedback expect to be heard by brands and expect a dialogue with brands and that’s a huge responsibility”.
Havasi also mentions that, in addition to this, brand loyalty is down and it’s easier for consumers to research products. “Speech recognition has advanced to the point that listening to your contact center is possible and more channels are opening up as ways for consumers to call for help. All of this adds up to an enormous amount of information”, Havasi explains.
You say you believe that AI has made real-time analysis possible, and that it is its business practices that are lagging behind. What can the businesses do to change this?
“It really fits into two big categories”, Havasi begins. “Coming up with a plan for gathering insights and then planning for dealing with those insights”.
“The first is simply designing a process for gathering the data and working with it – in as automated a way a possible. But this is harder than it sounds – often the key to finding some deep insights is to look at omnichannel data and getting the various data owners within a company to work together can be bushwhacking a difficult path”, she explains.
“My advice would be to start working on analytics programs within different channels (contact center, mobile app, HR, for example) while putting in place a plan for connecting those programs and deriving insights across channels”, Havasi says.
She explains that showing the stakeholders that they personally get value out of connecting analytics programs can sometimes be helpful. Sharing an example of this, Havasi said:”We were working with a big box store that was having customer satisfaction problems around pre-orders. Only by combining employee feedback with customer feedback were they able to fully understand the scope of the problem in their business process that was causing the difficulty”.
“Once you have the insights, come up with a plan for getting those insights to relevant parts of the organisation and giving them the power to act on customer feedback”, Havasi says.
It doesn’t help to know that a tab on the back of a razor is breaking off for consumers with certain hair types if you can’t figure out how to fix the product or design future products better.
She mentions that too often they see that insights stay within the analytics or contact center departments, when in reality it needs to make it’s way through the organisation.
Why would you say that companies who evaluated products and services quarterly, are now struggling to work in the real-time world?
“Companies are used to making decisions and reacting to feedback very slowly – they are used to running surveys quarterly or monthly and not looking at their live contact center data on a deep level”, Havasi begins. “When they find a problem, the avenues of communication are often not in place to react to that problem quickly”.
Explaining this more thoroughly, Havasi mentions a prime example of this, looking at the annual HR survey many companies do. “By the time you’ve run the survey, analysed the results, thought about the changes, and implemented them, most of the employees who were churn risks have already left!”, Havasi says
“Per Gartner, while 95% of companies surveyed collect customer feedback, only 10% implement improvements based on insights from that data, and a mere 5% let customers know what was changed as a result of their feedback”.
Havasi also refers to an earlier Gartner study that said that by 2017, 70% of the most profitable companies will use real-time analytics but the other statistics don’t point to that and neither does her experience talking with large companies about digital transformation.
“Things are definitely changing and many leading companies are looking to undertake a transformational initiative in 2017”.
How do you locate the “why”, in order to make sense of your sentiment analysis?
“If you only measure sentiment, you lack the “why” in the outcomes you discover which is often the key to improving outcomes”, Havasi says. “Additionally, often sentiment isn’t tied to the critical metrics you’re looking for”. I’ll give two examples. First, we were working with an agency that was doing
“I’ll give two examples. First, we were working with an agency that was doing a TV spot for a women’s hygiene product. The spot had gone viral and the agency was looking to more deeply understand the impact this had on the brand”, Havasi begins.
“If you looked at the sentiment of consumer reactions you found a very positive response to the spot. However, if you looked to see what was driving that positivity you’d find that consumers were very pleased that the *agency* had chosen such a forward way to address the topic – people felt it was good advertising and a step forward but no one was talking about the feminine hygiene brand and people’s thoughts on it hadn’t changed much”, she explains.
“On the flip side of that, in the contact center the squeaky wheel isn’t always the one that gets that impacts the KPIs. If you were to look at the early reviews of the Kindall Fire you’d find the number one complaint was that a plug wasn’t included with the product”, Havasi says.
“However, the rating and recommending behavior of those consumers wasn’t driven by that complaint. There were many other problems that more seriously impacted KPIs including the experience of consumers who bought a Kindall with an international credit card and were unable to use the app store”.
Why would you say that measuring only sentiment, measures nothing?
“Often, finding the answers to questions that really determine ROI are not simply positive or negative but are deeper insights in consumer behavior”, Havasi explains. “For example, if you were looking to understand – say – changes in consumer buying behavior due to the local produce movement you wouldn’t be able to gain any traction on the question through sentiment”.
Additionally, if you’re trying to determine how to improve your NPS score or to optimise your contact center the emotional state of the people who call is almost a given – and it won’t help you get to the results you need”.