Our POV: The 2018 Forrester Wave on Enterprise Social Listening, and a New Path Forward for the Industry

The newest Forrester Research Wave Report on Enterprise Social Listening (Q3) was published last week and the results are, well, mostly uninspiring for those looking to embrace innovation in the industry. Over reliance on the same data, commoditized features and a general lack of breakthrough innovation dominate most of the industry, report the authors in a rather hard-nosed but clear-eyed 40-criteria assessment on ten of the largest solution providers. On the surface the analysis paints a rather bleak assessment for those organizations hoping for new approaches that can help them maximize the impact of social listening.

But rather than portend just more-of-the-same over the next 18 months, we believe the report instead describes a general calm before a coming storm of innovation and transformation that will be more disruptive than at any time over the last decade. Indeed, there are some exciting developments happening just below the surface of the industry that have the ability to forge a critical new path forward and help unleash a wave of growth and excitement that is long overdue.

As a participant in the Wave Report several times and a recognized leader in social intelligence for more than fifteen years (before turning our focus more exclusively towards our machine learning text analytics technologies), the report’s general conclusions rang largely true. Brands are indeed demanding greater innovation, quickly and vendors are struggling to adapt. The good news is that we see change closing in fast from the horizon through refreshed “listening” approaches that directly address some of industry’s malaise. This includes expanding the critical mission of “social listening,” applying new game-changing AI-based technologies for advanced analysis, helping it expand into new areas of the organization and providing substantially more value and impact to organizations that go far beyond many of today’s narrow approaches.

Indeed, probe a bit deeper, and you’ll find this kind of innovation is already well underway. It is not just challenging the status quo already but also forging a new road forward; one with much higher revenue potential and with the ability to truly help drive customer intelligence in organizations who embrace it.

We believe the next 18 months will see several trends converge that will drive the industry forward. One is the maturity of new and highly effective AI and machine learning language technologies that are game-changers in enhancing our ability to analyze language effectively and accurately, and separate meaningful signals from this messy unstructured social noise. The second is the rapid growth of, customer-intelligence-hungry, quant-driven organizations with strong data science capabilities, sitting on top of massive unstructured data sets brimming with untapped insights and and need to integrate social data with other data streams for more real time insight and efficiencies. The third area is an emerging industry regulatory environment, driven through efforts like GDPR, The World Economic Forum and The Partnership on AI requiring high data accuracy, accountability and fairness to consumers. It is a powerful brew that is going to force unprecedented change upon the industry and provide the foundation for a new definition of social listening within organizations that will unlock greater value than ever before.

We expect this change and transformation to expand its foothold rapidly not just because there is a clear need for key industry players to differentiate (there is), but more importantly because brands are demanding it.  We hear constantly from brands that their needs in this area are accelerating in ways that have made it difficult for providers to keep up through the typical roadmap process. Existential challenges are often the best motivator for innovation and, as the report describes, staying status quo is likely a very treacherous road when the market has shifted. It’s become quite clear the leaders versus laggards over the next 18 months will largely sort themselves out based on those who can move quickly to embrace and capitalize on these new needs versus those who stagnate and surrender to industry homogeneity. So what exactly is coming soon?

AI Gets Real: AI may be considered hype across many industries but advances in machine learning are demonstrably having a tremendous impact on our ability to process and analyze language. The best algorithms driven through deep learning and “human-in-the-loop” active learning approaches are powerful and now able to approximate human performance for accuracy and comprehensiveness but at the scale and consistency that only software can provide. This even applies even to challenging areas like sarcasm, slang, mixed sentiment and implicit meaning. The problem for consumers of social data services currently is, as the report clearly states, that most providers are touting some form of “AI” making it difficult to evaluate solutions effectively. “Brands should be wary of over-exuberant AI promises and understand that social listening platforms still require humans to train the data in a semi-supervised environment before becoming operational,” the authors write. Smart companies will not accept AI claims at face value. They will and should require solutions to go beyond marketing speak and be able to verifiably demonstrate the substance of their technologies and processes. This may require bringing in experts with deep data science expertise to help in the evaluation. But healthy skepticism should not give way to cynicism. Indeed, by exploring outside just the confines of this report, they will find mature, powerful technologies that meet and exceed their growing demands. To quote author William Gibson, “the future is here – it’s just not evenly distributed.”

A Demand For Higher Standards: But with this great power comes even greater responsibility. Anyone with enough technical chops with some training data and access to open source technologies can build a crude classifier to analyze language that might function with some level of competency in a highly-controlled setting. But with growing recognition of potential unintended bias in machine learning models and the clear need for models to “generalize” effectively (that is work effectively with datasets it has not seen before) conversations about whether they should use AI (it’s critical they do) will shift to how they adopt and apply it correctly. This requires embracing new technical features, aligning with data science teams, including clear performance measurement on each classifier before and during deployment, ensuring unbiased training input, as well as implementation of protocols and processes to keep humans in control of the models to intervene strategically when needed. Social listening will need to align with emerging organizational “AI strategies” that emphasize trust, accountability, fairness and abide by process and service level agreements to mitigate potential risks from the technologies due poor implementation or misuse. Indeed, a growing number of smart brands are already requiring solutions to get specific about what techniques are being used, demonstrate how they handle mixed sentiment and provide clear evidence on how they are avoiding bias in their models. Perhaps most importantly, they are requiring vendors to provide clear verifiable classifier performance results with a commitment to provide that performance on a contractual basis. Brands need access to this performance information and vendors need to be able to provide it on demand and with transparency.

Brands Take Control:  As a result of the above, brands will take even more direct control of their data quality and models. New machine learning-as-a-service solutions for text (and images), such as our Conversus.AI platform, are already integrating via plug-ins into some leading current platforms to allow even general analysts at brands to build, modify, test and deploy their own models for immediate use. This robust customized approach will supersede the basic, one-size-fits-all, general analysis offerings that currently dominate the market. Benefits to brands include adjusting the models to fit their own unique taxonomies (not all brands define “trust” for example the same way) and better aligning the models to their data and insights requirements while protecting their IP. It also helps ensure all models meet the rigorous precision and quality requirements (via automated F1, AUC scoring, etc) their organizations demand while positioning their resident domain experts to be “in-the-loop” to manage, oversee and improve the models for maximum impact (and mitigate risk) by putting the power of machine learning directly into their hands. This will require social listening platforms to become even more of an open ecosystem allowing greater API integrations, third party model inclusion and integration with other related platforms as designated by the preferences and demands of their clients.

Models Become the New “Product’: Machine learning models pre-built for specific functional use are going far beyond the “basic” sentiment and emotion common analytics needed in most platforms. New emerging pre-built models accelerate insights and will save companies substantial time and money by not trying to build these models from scratch. These models can be chosen to align with organizational priorities and plugged right into a wide range of areas where they can provide the greatest value. These can include brand attributes (trust/distrust, innovation, value, etc.), customer experience (journey analysis), product innovation, social customer care and much more. Aligning models to specific business functions will make it much more efficient and effective for diverse parts of the organization to make use of the data, an area that the report says is an important need. Further, given that language differs by industries, classifiers will also become more industry-based and specific as opposed to cross industry general approaches that dominate the market today. Expect within the next 18 months that brands will be choose the models through a more “App Store” experience, simply subscribe and deploy immediately to their preferred platforms and data. While there will also be great demand for unique custom models, the move toward prebuilt models will save companies substantial time and money, unlock insight more effectively, and accelerate their use of AI in the enterprise.

Social and VoC Merge: As brands finally begin to effectively separate the critical and meaningful signals from the vast social noise for accelerated insight, they are now beginning to turn the technology into another vast untapped resource. According to Forrester, companies on average are processing less than 25% of their unstructured customer data, which includes not just social data but also call center transcripts, long for survey verbatims and more. And for good reason: processing this data through standard text analytics with any level of confidence has been quite challenging. In the race to customer intelligence, these new sophisticated AI-powered social language models will increasingly be applied to other VoC data initiatives. Social listening and VoC analysis will naturally merge providing new opportunities for growth effectively chomping into that vast unstructured data for insights.

Cost Discussions Turn to Value: The growth of the current incarnation of the social listening market has generally plateaued globally, by our own estimate, in the $400-500 million range. With perceived commodization and saturation, most platforms are forced into shootouts and tug-a-wars with each other over existing budgets which necessitates requiring deep cost cuts to win. It can become a race to the bottom. But these new approaches change that dynamic by unleashing the application of this data into new parts of the clients’ organizations (and new budgets) through a more value-based pricing model. For example, a global client of ours, one of the world’s largest software companies, is using our AI-powered social intelligence models to replace its traditional buyers journey survey-based analysis, providing real time insight while saving approximately $10 million a year annually. Another automotive company, through just one highly effective machine learning model, was able to more than double the accuracy of its customer care classifiers for its social customer care initiative, leading to vast cost savings, reduced latency, less human data cleansing and improved customer experience resulting in millions of dollars of value. The value of these models in many cases now supersede the entire cost of a most traditional annual listening platform contracts and we expect that to only grow as this classified, accurate data is integrated into more critical areas of organizations and demonstrate tangible value. As a result, the growth ceiling for the entire industry will level-up significantly.

Ecosystems Dominate: Effective analysis of social and VoC data will mean models need to become portable and be applied to analyze the data wherever the data resides, be it a traditional social listening or management platform, VoC platform, customer data database, attribution system or business intelligence platform. Right now, most large organizations have multiple platforms with little to no consistency in how the data is processed and analyzed. All listening platforms, for example, classify data differently resulting in a mishmash of data that cannot integrate. The pursuit of “one truth” that unifies the analysis of this data consistently across the organization will demand models fulfill “ecosystem” approach rather than walling them off in a specific, siloed platform.

Data Gets Predictive : AI-powered social data will ensure it becomes critical components of CX, brand health, product innovation and predictive analytics. This highly-refined and accurate data has proven time and again to have quantitative value. Data scientists will increase their demand and access to the data for modeling. In several cases, our team has effectively used social data to build out machine learning-powered attribute classification that was able to accurately predict the outcome of more traditional brand health surveys 12 weeks in advance and provide greater understanding of key drivers. In other recent example, a partner of ours as able to use custom classifiers for “motives” to predict product sales six weeks in advance for a global packaged good firm. Demand for this data with this kind of quantitative power is indeed going to become de rigueur for attribution modeling, brand health, customer experience, predicting customer satisfaction and proving that better, faster and cheaper can indeed exist.

The future of the industry is bright. As we head into 2019; into a time when the collective voice of customers and citizens, empowered through social channels, have become a primary agent-of-change and transforming governments, societies, industries, brands and products, there’s arguably no greater obligation than for the industry to accelerate change to “get it right.” And there has been never been as much demand for brands needing to understand their customers in real time as the present. As the Forrester authors say clearly in the report, the technology has great potential, but its heyday (is) still to come. “We’re still holding our breath (for this potential to be fulfilled) in 2018,” they write. We expect that given the rapid innovations now available and spreading from important corners of the industry, together with a strong appetite for brands to look at their strategies with fresh eyes, the report’s authors, customer-centric brands and the industry itself will finally be able to begin to breath more easily in the year ahead.

Interested in learning more and seeing Conversus.AI in action?  Simply schedule a demo here!

Also, if you have any additional questions or comments please feel free to email us here: sales@converseon.com.

©Converseon All Rights Reserved