Data Scientists Don't Scale

Organizations have learned that the data about something can be more valuable than the thing itself. But what they are beginning to find out is that this bit of wisdom applies to data as well.

Despite technology's vast contributions to contemporary social and economic reality, skeptics continue to demand 'proof' of impact. It seems like it should be obvious, but the devil's deal between tech and finance has given the money guys the whip hand - and all they want is more. The result is that greenmailers like Carl Icahn can still shake down transcendent enterprises like Apple in the guise of improving shareholders' returns. And even Apple has to kowtow.

So when marketers complain about the imprecision of mobile advertising spends, let alone results, or economists debate technology's impact on productivity, it is only a matter of time before those buying Big Data demand to know where the path to promised profits lies. The reality is that this follows a well-worn template. First comes possibility, then capacity, then comes knowledge and then wisdom, which is what delivers all those juicy operational and financial improvements everyone thinks that investment in data is paying for. The demand for dominance waits for no one. JL

Stuart Frankel reports in Harvard Business Review:

We’re at the beginning of the next phase of big data, a phase that will have very little to do with data capture and storage and everything to do with making data more useful, more understandable and more impactful.
Big data is about to get a big reality check. Our ongoing obsession with data and analytics technology, and our reverence for the rare data scientist who reigns supreme over this world, has disillusioned many of us. Executives are taking a hard look at their depleted budgets — drained by a mess of disparate tools they’ve acquired and elusive “big insights” they’ve been promised — and are wondering: “Where is the return on this enormous investment?”
It’s not that we haven’t made significant strides in aggregating and organizing data, but the big data pipedream isn’t quite delivering on its promise. Despite massive investments in technology to store, analyze, report, and visualize data, employees are still spending untold hours interpreting analyses and manually reporting the results. To solve this problem and increase utilization of existing solutions, organizations are now contemplating even further investment, often in the form of $250,000 data scientists (if all of these tools we’ve purchased haven’t completely done the trick, surely this guy will!). However valuable these PhDs are, the organizations that have been lucky enough to secure these resources are realizing the limitations in human-powered data science: it’s simply not a scalable solution. The great irony is of course that we have more data and more ways to access that data than we’ve ever had; yet we know we’re only scratching the surface with these tools.
A few innovative executives understand this and have sought scalable, automated solutions that interpret data, unlock hidden insights, and then provide answers to ongoing business problems. Artificial intelligence (AI) is beginning to transform data and analysis into relevant plain English communication. AI is shortening employees’ data comprehension-to-action time through comprehensive, intuitive narratives.
Any organization where employees are spending valuable time on manual, repetitive tasks is a prime case for intelligent automated solutions. There may be no better example of this than the financial services industry.
Take for example the necessary but incredibly manual process of producing performance reporting for mutual funds. Typically, marketing teams toil every quarter to document portfolio performance and add commentary (see an example here). Today, some funds are using advanced natural language generation (Advanced NLG) platforms, powered by AI, to automatically write these reports in mere seconds. (My company, Narrative Science, works with multiple financial services clients to do this sort of work.)
These are not simple, static reports, but are data-driven, complex, and dynamic, reflecting the brand voice most appropriate for the firm. The reporting incorporates disparate data sources and performs real-time analysis on portfolio performance. These systems examine the facts, determine which of these facts are most notable, and output these facts as readable narrative text. For instance, the report could be a bulleted list of key findings for portfolio managers who just need the facts, or it could be a lengthy detailed summary to meet an investor’s needs.
To do this work, the system starts with the goals of the report (e.g., did this portfolio outperform the benchmark?). Once the business need is clear, the system pulls in the necessary information (portfolio results compared to the benchmark), performs the relevant data analysis (cohort comparison), and finally decides which data is required to complete the task (portfolio attribution data, returns data, expected values, actual values, and any confidence weights on the expected values).
Once the report is created, it updates automatically as the data does, eliminating the need to redo the analysis and reporting each time there is a change.
The use cases for these systems are countless, but they all start with the question: What do I want to communicate that currently requires a significant amount of time and energy to analyze, interpret, and share?
Take medical billing. AI scours thousands of billing records across hundreds of hospitals and generates narrative reports that immediately provide the desired analysis. These reports can highlight changes to a hospital group’s insurance-provider portfolio or changes in demographic mix that are affecting revenue and cash flow, while simultaneously identifying room for growth by suggesting changes to a doctor’s workload.
There are also a number of examples where AI solutions are improving customer experience. AI is the first technology to make personalized, “audience of one” communication a reality. Companies can speak to each customer as if they are the only customer, offering personalized reports that seamlessly integrate with consumer applications and websites.
Wealth management is starting to see this benefit. Many of us have now heard of “Robo-advisors,” the automated financial advisor that can offer a low-cost alternative to expensive, human advisors. AI is being embedded into existing advisory platforms, delivering personalized portfolio reviews and recommendations in natural language to customers who are unsure as to what their charts and graphs actually mean. Unlike the Robo-advisors that operate in a “black box,” these AI systems provide transparency by communicating what has happened and offering recommendations in plain English that people can immediately understand.The commonality across all of these new technologies is that they offer something additional humans cannot provide: the power of scale. Organizations that do not have a strategic initiative to regularly and organically engage with its customers will be at a serious disadvantage. Soon, AI-driven engagement models that interpret data and intuitively interact with clients will be the norm.
In the near-term, the adoption of AI within business intelligence platforms and customer-facing applications will accelerate. When a business user receives a dashboard, he will also receive an accompanying narrative to explain the insight within the visualization, easing consumption and quickening the pace of decision-making. Eventually, AI will offer even more complex analysis and advice. When a salesperson checks her pipeline status within her CRM application, she’ll be able to ask, “How am I performing compared to my colleagues this quarter?” and “Where should I focus my efforts to ensure I reach my quota?” The application will be able to respond immediately, with an explanation and sound advice she can understand and act on, in a back-and-forth, conversational manner.
The key to all of this is the intersection of AI and advanced natural language generation. We’re at the beginning of the next phase of big data, a phase that will have very little to do with data capture and storage and everything to do with making data more useful, more understandable and more impactful.