Algorithms predate AI. In fact, they are the digital representations of the businesses’ competitive differentiator. A retailer’s focus on low price is an algorithm. As is a manufacturer’s practice of sourcing top-quality raw materials. An AI algorithm is the digitized version of these differentiators.
The advantage of these algorithms is that they can identify strengths and opportunities that companies aren’t even aware of. What’s more, AI algorithms can deepen expertize that’s already there, and generate new ways to take it further. And yet, they are nothing without human insights.
These insights are derived from ethnographic research — or what we call thick data — to blend human science and data science to make core business processes smarter and improve customers’ lives. Here is a look at three such companies we worked with for using human-centered algorithms to improve customer engagement.
After years of successfully marketing its credit cards to small and medium-sized companies, one of our financial services clients saw card offer acceptance begin to decline. Committed to reversing the drop-off and boosting conversion, the company partnered with us to learn more about the reasons for the decline.
Our team recommended learning more about the small-business owners who are prospective customers by asking some fundamental questions, such as: What are their buying needs? When are they most receptive to offers? We devised a field study in which we interviewed 20 representative clients in the New York City area in a variety of industries.
We found that most of these small business owners rarely delegate decisions, and like to manage their own finances. Perhaps most surprising, particularly in our tap-and-swipe era, was that because the business owners typically partner with local bankers for loans and lines of credit, the telephone is their preferred mode of communication for discussing financial services.
We figured that, for this client, it’s important to factor the telephone preference into its core algorithms, along with the discovery that the key decision makers for corporate credit cards are the business owners.
By carefully building hypotheses and then translating them into data proxies to improve the existing targeting model, the company expects to sharpen its target list and the timing of its card offers. The new engagement model will identify the interaction points likely to be most successful. What are the conditions under which they’ll take an offer? All of that information will now be fed into its algorithms.
Utility customer experiences are typically uneventful: Customers pay a monthly bill and get a service — no upgrades, no loyalty programs, no perks. However, European energy company is using AI to upend that model. With an eye toward new services and channels, we worked with them to change the nature of the way it connects with utility customers.
Its call center was sagging under the weight of incoming customer calls, most of which were billing queries. Customer satisfaction was low. Net promoter scores were dismal. Given the rise of new options in the energy sector, the company sought a fresh, more positive way to engage with customers and grow its business.
We started by reviewing their field notes and interviewing some customers. We then hosted a workshop with call center staff and conducted one-on-one interviews to further explore the customer experience.
From that research emerged a new way for the company to view its business: Rather than connecting with customers around billing issues, the energy company could shift its focus to their residences. In the new engagement model, the customer is the home.
By integrating multiple data sets — energy consumption, customer profile, contact center logs — our team created an AI algorithm that indexes customers based on details such as usage. Now, instead of engaging with customers around problems, the algorithm engages with them around goals such as reducing carbon footprint and lowering energy costs.
When a regional theme park sought to boost ticket and merchandise sales, it needed to know more about what motivates park-goers to visit its site. How do they experience the visit? How do they take it with them when they leave?
Our team began by interviewing dozens of guests including families, couples, and friends. We then combined the data collected at the park with other data sources such as third-party demographic data. We also looked at the wristband data collected from the park’s customers.
Based on this analysis, our teams identified three visitor needs the theme park could deliver on. For each need, we created an index, suggested pilots and developed a set of key performance indicators (KPIs). The needs are:
Here’s how your organization can begin to strike the right balance of big data and thick data to ensure the success of your AI initiatives.
To learn more, read “Through Thick and Thin: Making AI Work in the Real World”, visit the Digital Business section of our website, or contact us.