By Maurice N'Diaye, Catalix Core Team Marketing Expert and Partner at Synomia
Catalix is an AI school for business, based in France
Synomia is a consulting agency specializing in data marketing and the creation of innovative strategies to help companies reach their projected growth, also based in France.
Among the many trends driving companies towards customer centricity, customer feedback management is a particularly complex, though strategically important, challenge to deal with.
Customer Experience and Relationship Management (CERM) Software Market reached an estimated $42.14 bn (1) last year (that's 15.5 percent growth), and needless to say, the big issue in understanding customers to better create differentiation is the increased ability to collect, aggregate, process, and activate customer feedback.
For that matter, AI should be expected to bring a consistent answer for companies that aim at creating value out of this gigantic pool of data, either for process automation, process optimization, or experience innovation.
Let’s just quickly run an overview of three typical use cases leveraging AI for CFM.
Case 1: Clustering
Whenever the customer service department manages to detect and record multi-channel customer interactions (call center, chat, social networks, mobile apps, emails, NPS surveys, e-commerce platform reviews, etc.) there comes a need for organizing this data: qualification, quantification, and hierarchization. First stage analysis will focus on quantified metrics, such as NPS, but most of the valuable information lies in unstructured streams: text, image, and voice. Only there will the marketer find actionable insights about the how and why of behaviors. What are the sources of satisfaction/dissatisfaction, which parts of my funnel create the most friction and why, what is the image of my brand through each channel?
The tasks of processing and sorting those types of data are highly complex, and require advanced AI capabilities to be delivered in sustainable, business-relevant ways. We're talking here about Natural Language Processing, Speech Analytics, and Image Recognition. Those three fields of AI are in fact very different sorts of AI, but a comprehensive CFM will require a combination of these capabilities.
Case 2: Predictive analysis
Another range of applications you want to be able to engage when dealing with customer feedback is prediction. Are there actual patterns emerging from customer behaviors and conversations, that can help anticipate upcoming crises, business opportunities, consumption trends, or churn? For that matter, just as for any kind of prediction, the ability to create massive data sets and to reduce bias will improve the performance and sharpness of the prediction. The goal here is to enter a wide range of parameters, the combination of which should lead to the identification of underlying patterns that the human brain alone could not apprehend.
Case 3: Customer interfaces
I remember hearing Google Senior Fellow and SVP Jeff Dean when inaugurating Google’s center for AI in Paris. He said that what changed radically with the rise of AI applications, was the shift in the way we use machines from execution to interaction.
Indeed, now that we’ve passed the Turing test in various ways, we can create a whole new range of customer interactions and diversify customer touch points. Connected speakers, chatbots, AR ... these are new ways of interacting and thus of collecting feedback.
Then why do so many firms struggle with the implementation of efficient AI components within their CFM stack? Probably because of a combination of 5 key success factors that are tricky to master all at once:
Technically speaking, the simple fact of gathering streams of data coming from multiple touch points, at different time frames, in different formats, in different languages, from different segments creates an incredibly complex galaxy of pipes, APIs, and databases. Not to mention data quality when sources come from outside such as social media. And let alone methodological headaches from the interpretation and weighing of signals coming from different platforms and collected in different ways, like customer service spontaneous conversation records vs NPS oriented open-ended questions.
Choice of technologies
In a market where a new player comes in every other day with a disruptive-mind-blowing-never-seen-before-already-market-leader-must-have-technology, it is hard for executives to make the right decision quickly. First, technologies are difficult to differentiate for non-experts. Marketing pitches from tech companies all tend to deliver the same promises, with often very different realities and performances.
Second, a same business problem can sometime be tackled in very different ways that will leverage very different technologies. Understanding your customer base and segmentation can be done based on transactions (what they buy), or on conversation (what they say), or else on profiles (who they are). Of course, you want to do it all, but if you have to choose for budget reasons, let’s say, then you need to arbitrate between structured data clustering and NLP. And that is amplified by the constant evolution of techs in terms of maturity, cost, performance, applications. Look at how speech analytics or image recognition could totally change the way we analyze customer feedback if they were mature enough.
It may go without saying, but clearly determining the needs you want to fill when implementing an AI-based feedback management system is not always that easy. What specific KPI do you want to act on, are you looking for real-time activation and reaction, or in-depth insight detection, or marketing innovation in customer interactions? Too many companies are trying to address all these use cases at the same time.
But the one-stop-shop approach is not compatible with the radically different specificities between operational performance and strategic analysis, that will require different features. For instance, focusing on real-time operations means fast rather than sharp classification capabilities, when focusing on strategy requires advanced processing and modeling technologies. As of today, there are still no players able to reconcile both, and this is just one example.
Sponsorship & Organization
The best AI will never bring any impact if it is not well supported by adapted sponsorship and leadership. Taking a turn in the way you handle feedback will require both investments, and change. And the higher the vision is upheld within the organization, the easier it will be to fund a new project, and to make sure employees will show the right level of dedication and get enough room to make the change happen. That's true for any innovation project actually.
People & Culture
Probably the main key to success here, is how you onboard your organization towards AI-based CFM. You need to setup a client-centric culture, explain that AI doesn’t have to replace humans but can help them do their job, and find the right combination of skills, between in-house and outsourced (and the balance can change overtime, depending on the fact that you want to favor speed of delivery or build a strategic asset).
In a series of articles called “The customer insight series,” and based on surveys conducted over the past decade, the BCG suggests a framework about the place of Customer Insight in a company’s strategy, and the way it can drive competitiveness and growth if it is well implemented (2), from “traditional market research provider” to “business contributor”, then “strategic insight partner”, and finally “source of competitive advantage”. Executive support being the game changer in the speed of the process.
This framework could work as well here, as CFM can move up from a “traditional feedback aggregator” to become a “source of competitive advantage”. Harnessing the power of AI and understanding its potential impact will be critical in doing so, so the sooner you get started, the better!
1 N. Gupta , Y. Dharmasthira , J. Poulter, Customer Experience and Relationship Management Software, Worldwide, 2017, Gartner Research