A topic that we often address to prospective new clients and users is that we believe in full transparency, so we empower them to maintain control over what happens between when they send over their data and when they receive their narratives. NLG is still an innovative and recent technology, it’s not surprising that there are questions about the work that goes on behind the scenes and upfront, as well as what the process looks like from start to finish.
Myth: All NLG providers operate using the “black box” solution method
Reality: Self-service NLG provides transparency and complete control every step of the way
Black box NLG is a term typically applied to a canned, static, or “behind the curtain” NLG solutions. These solutions are prevalent across the industry and known for disappearing with your data for weeks or months at a time, then returning with narrative that you’re not able to directly edit. What’s worse, is the fact that non-domain experts are handling your data and trying to produce a narrative. Since they are not subject matter experts for which they are building out the logic, you’re left with an output that showcases very rudimentary insights and requires weeks of back and forth editing. Wordsmith users are never privy to this type of inefficient and self-serving, versus customer focused, model when they work with Automated Insights.
Data changes constantly, so your ability to quickly edit and update your NLG architecture within a platform such as Wordsmith is essential to a successful and valuable NLG deployment. Black box solutions like to advertise that they save you time by removing your direct involvement. Some may even tout their cutting-edge artificial intelligence, but the fact is that you have a plethora of non-subject matter experts taking their best stab at the story they think you want to tell. In reality, you spend more time submitting change orders to make even the smallest of edits to an often generic output. Your NLG solution should be as advanced as your data, it should give you complete control, transparency, and customization over your content.
What does this mean for data-rich industries?
When it comes to data-rich industries like business intelligence, financial services, and healthcare, Black Box NLG solutions tend to lack awareness of the underlying dataset. They rely on a predefined set of rules and conditions that parse the surface of your data to extract basic talking points in a pre-set, formulaic manner. These outputs typically include very basic metrics like minimum, maximum, summation, and average. While these metrics give some value, the key to driving change through data is the ability to analyze the entirety of your visualizations and tell a story across your dashboard. By removing any control you may have over the contextual layer (the step between processing data and generating output), black box NLG providers are forced to interpret fields of data in the same manner every time, no matter the industry or use case.
Here’s a common example from the financial services industry that would cripple companies using a black box solution:
Black box NLG providers would look at these two stocks and, without any input or guidance, point out that Company A has a lot of variance while Company B does not. This approach is fundamentally flawed, yet that is the narrative output that would be generated.
First, you already know, due to its nature, Company A will have much higher variance, so the information is unnecessary. Second, notice the subtle shift from using “volatility” to “variance” over the past few sentences. This shift is deliberate as the NLG solutions incorrectly identify each dataset as being equivalent. Without any guidance or consideration for contextual terminology, these solutions are unable to use company or industry-specific expressions, descriptions, and jargon. In the context of NLG for individual companies and industries, that functionality should be fundamental.
Self-service business intelligence is the ability of business users in an organization to independently carry out tasks instead of passing them to data scientists or IT experts for fulfillment. This approach allows individuals to access unique, role-based analytic insights anytime and anywhere, in addition to empowering them to act quickly and accurately to stay ahead of competitors and on top of organizational health.
As the leading natural language provider for business intelligence and analytics, Automated Insights has received feedback over the years from clients like Allstate, the Associated Press, Yahoo!, Greatcall, Arterra Wines, and countless others that highly value Wordsmith’s ability to put you in complete control of your content through customization and personalization through self-service business intelligence solutions. You are the expert of your data. You know the story it needs to tell; you know how to tell that story better than anyone.
Despite what a team of marketers want you to believe, anything other than self-service analytics (from any provider) is solely a blanketed approach that generates content in a vacuum. Without any of the context to makes your data and story unique, the narrative you generate will always fall short.
Missed the first part of our series? Check it out here.