Augmented BI tools speed insights with less manual labor
Through AI, machine learning and NLP, augmented analytics reduces the manual labor of traditional BI work to more quickly and accurately locate and analyze relevant data.
Augmented analytics embeds machine learning, natural language processing and other AI functionality into business intelligence software to help citizen data scientists and business users more quickly and accurately locate pertinent data, distill patterns of information, model the data for analysis and interpret BI insights.
Advancements in AI and machine learning technologies beyond traditional BI tools allow businesses to discover nontrivial insights, such as predicting the lifetime value of customers and their preferences to help improve marketing efforts, said Sewook Wee, director of data science and analytics at real estate site Trulia LLC. But there are limitations -- in particular, finding the talent, time and resources required to build high-quality analytical models for augmented BI analytics tools to run. "The speed of technical advancement and access to data are there," Wee explained. "The limits to augmented analytics are a matter of prioritization and resource allocation."
Therefore, companies should familiarize themselves with the ways augmented analytics tools can complement BI before embarking on a major initiative. "It's important to avoid the temptation to try to build the perfect solution from day one," Wee cautioned. Starting with a simple model, it's possible to prove business value, learn and iterate. Those early models become key inputs for future models.
When Trulia was building its first price-range prediction model, it only assigned one price range per user, and that worked for many customers, according to Wee. But the company quickly learned that many users have different target price ranges for different geographical locations. For example, a potential buyer interested in purchasing a home in the San Francisco Bay Area might have different price targets depending on square footage, amenities, commuting distance and neighborhood.
Identify and prioritize insights
Sewook WeeDirector of data science and analytics, Trulia
This may seem simplistic, but it's important to know what's going to be analyzed or discovered by a BI system using augmented analytics tools since the natural language processing (NLP) service needs guidance about the insights it's analyzing and deriving.
"You can't just approach it with no guidance and expect the NLP to just know what you're looking for," said Stephen Blum, founder and CTO at real-time network-as-a-service provider PubNub Inc. "Augmented analytics are only as smart as your implementation."
It's good practice to identify the areas that require deeper insights, then test and validate findings, instead of expecting augmented analytics to replace traditional BI tools. That's a good way to get to know the augmented BI analytics tools, write the correct algorithms and design better models. It may take some time to develop the expertise to design and implement augmented analytics models and algorithms.
"Though the NLP system does the heavy lifting when it comes to processing and analysis, your model plays a massive role in how relevant the augmented analytics are," Blum advised. Therefore, companies will need a certain level of expertise in NLP, machine learning services and data science. In the long run, Blum expects that BI vendors with augmented analytics features will speed the learning process.
While AI and machine learning are powerful tools, they still need to be focused on a specific problem to produce results. "Start with something that's hard for people," suggested Jake Freivald, vice president of product marketing at BI and data management tools provider Information Builders. Detection of patterns, for example, often can be difficult to visualize because there's a lot of "noise" in the data, but natural language generation can convert the most relevant information into easy-to-understand text.
Stephen BlumFounder and CTO, PubNub
Business users may struggle to understand the limitations of their models, so Freivald said it's probably best to let data scientists determine the models to be used and to grant business users access to the models through dashboards and visualizations instead of the more complicated modeling tools.
Another alternative is outsourcing expertise. Consulting services can help complement in-house expertise, particularly in the early stage of an augmented analytics project. "These third-party vendors not only have high levels of data science expertise, but are also able to learn across clients to leverage combinations of data and insights accessible only to them," said Eli Finkelshteyn, CEO and co-founder of search discovery platform-maker Constructor.io.
Look for partners with a combination of technical innovation, expertise and reference customers. It's also important to have stakeholders regularly use the new platform and provide feedback even when the platform is not giving brilliant insights, Finkelshteyn added.
Simplify access to insights
AI applied to NLP for querying data and generating insights in a conversational manner helps business users expedite data set queries, said Thierry Audas, senior director of product marketing at SAP. By contrast, traditional BI workflows require more familiarity with how the data is modeled and more time to construct the right queries and interpret outcomes.
According to Audas, augmented analytics is most successful in helping business users with no data science skills to formulate BI strategies. Sales managers, for example, can ask targeted questions without waiting for the data science team to generate a report. And with the click of a button, they can trigger automated machine learning algorithms for clustering, classification and regression to uncover key factors influencing the success or failure of a sale.
Mark PalmerSenior vice president and general manager of analytics, Tibco Software
To get the full benefit of augmented analytics tools, it's important to create and maintain a closer, more collaborative working relationship between data scientists and business analysts. "Automated machine learning and augmented analytics are great technical features," noted Mark Palmer, senior vice president and general manager of analytics at Tibco Software Inc. "But business analysts need to better understand the statistical theory behind observations or models, and, similarly, [it's imperative] that the data science team understands the business context and needs better."
A team of business analysts that includes a data scientist can better integrate and refine models within visualizations to tackle business problems. These teams can also find new ways to combine text analytics with data science to create readable guides and help business users easily understand metrics, statistics and other findings. Augmented BI analytics tools make it possible to infuse BI into data science or automatically generate machine learning recommendations. "However," Palmer noted, "technology cannot bridge the understanding gap to map these statistical and technical observations to business problems."
Read more about augmented learning