In today’s data-driven world, companies are scaling up investments in data storage solutions and optimizing data pipelines—all with the intention of helping people make better business decisions. However, they’re also realizing that data itself means little without the right tools to produce meaningful analysis.
After all, companies must come away from these costly investments with actionable insights. Where’s the value if you can’t generate insights to help drive growth, cut costs, and ultimately move the needle on performance?
Many companies are facing the fact that their traditional business intelligence (BI) investments in reporting and dashboards don’t necessarily get them closer to actionable insights, at least not at the scale needed to go on investing in those solutions.
Instead, the hope is that a new generation of big data and machine learning will mean that employees can ask a myriad of complex questions specific to their department, function, or role and receive answers in seconds or minutes, without intervention or support from a data analyst.
Insights should make teams smarter, more agile, and more efficient. They must be embedded in the decision-making process, accessible to business people without technical expertise.
Insights produced by traditional BI platforms tend to lack the depth, speed, and ease of understanding to be truly actionable at scale. Employees are accustomed to reporting that is either irrelevant or fails to point out the key problems.
In this post, I discuss why traditional business intelligence is insufficient to scale insights generation.
Traditional business intelligence is insufficient
Traditional analytics platforms rely on dashboards to illustrate stats and trends in the data. Dashboards are typically composed of distinctive visualizations that answer simple questions, such as “what were sales in Q3?” or “Report on Sales In Q3.”
While dashboards can show data, they can’t explain the results with drivers and analysis. An analyst must still interpret dashboards to find insights and piece the story together.
In reality, analysts download data from dashboards into Excel to determine root causes and uncover meaningful, actionable insights. While dashboards can provide helpful visualizations, analysts are still performing the bulk of the work, often outside the BI tools themselves.
That means analysis is not uniform across departments because best practices are difficult to convey and enforce. The occasional big success punctuates an expensive and mundane flow of largely irrelevant charts and grids.
These limitations can lead to low adoption, since business people need support from analysts to get actionable insights. In turn, data and analytics leaders who’ve carried out traditional BI are left with lackluster ROI.
As such, there’s an impetus for companies to upgrade traditional BI and analytics tools to drive the results they want to achieve.
These considerations illustrate the depth of the need for better insights at scale:
- Surface-level insights aren’t sufficient. A BI tool may show a trend or perform statistical analysis, but this information doesn’t help a business person get closer to understanding why and what steps they can take to impact a business outcome.
- The cost of analytics is increasing, along with demand for insights. Analysts are challenged to better serve the business, while at the same time, they’re asked to reduce costs. The insights generation process is too manual and time-consuming— the same way it’s been for 10–20 years.
- Existing BI and analytics tools are not broadly accessible. While dashboards can provide helpful visualizations, they often require technical expertise to interpret or build.
- The turnaround time for quality insights is too long. Insights can quickly become outdated when it takes days or weeks to produce reports or build dashboards. Even fast analytics tools don’t solve this problem, as quick insights can be disparate and lacking context. A competitive environment requires speed.
To solve for these challenges, companies need fast insights that replicate the knowledge of a data scientist— insights that can be infused with the business expertise of a department or category manager, a supply chain guru, or a strategic executive. In other words, companies can’t sacrifice deep analysis to get quick insights, and vice versa.
The solution cannot be traditional BI and dashboards. These tools have been around for 10–15 years with user adoption rates stuck around 30%. To help companies achieve the outcomes that measurably improve performance, they must look to future-facing technology: augmented analytics and GPUs.
Operationalize data science with augmented analytics and GPUs
Augmented analytics leverages the power of AI to put data science capabilities into the hands of business people.
This data democratization — essentially, allowing business people to access and understand data without intervention from a data scientist or analyst — is enabled by two key components of augmented analytics:
- Natural language generation (NLG) — NLG produces insights in plain language. Advanced NLG can produce entire data narratives that tell the story of business performance, trends, and opportunities.
- Machine learning (ML) — ML algorithms perform exhaustive data analysis, surfacing hidden insights by testing every data combination. To accelerate ML, companies can pair augmented analytics with GPUs (more on GPUs in a moment).
With NLG and ML, augmented analytics automates the analytics workflow. Augmented analytics not only performs the analysis far more quickly and exhaustively than a person could, but it also mitigates the need for interpretation.
As discussed, a report generated by traditional BI would likely include visualizations with simple answers that state surface-level trends. These answers might state a percentage increase or decrease for important metrics like sales or market share, but would require technical expertise to understand the “why” behind the numbers and generate actual insights.
Augmented analytics, in contrast, can answer “why” questions directly; ML algorithms determine core contributors and detractors, developing a full data narrative that helps business people understand where to focus their attention.
There’s no need to bring an analyst into the conversation. Augmented analytics enables business people to drill-down, to ask follow-up questions, and to quickly gain an unbiased, 360° view of performance.
Augmented analytics is a powerful tool. To generate insights above the gold standard, augmented analytics requires significant processing power, especially for the large data sets common in enterprise organizations.
Most analytics platforms currently run on central processing units, or CPUs. That’s the kind of tech that powers the computer you’re using to read this. While CPUs work well for lots of different types of analyses, such as trending and contribution, their use is limited for AI and ML. CPUs simply don’t have the speed that can provide a competitive advantage. As companies move to reap the benefits of AI and ML, they’ve also seen the need to move from CPU-based systems to GPU-based systems.
GPUs, or graphical processing units, can cost-effectively accelerate augmented analytics, scaling insights generation through massive processing power. Originally developed to create the kind of realistic 3D environments now seen in video games, GPUs can now be applied to advanced AI-driven technologies well beyond that humble start.
As such, GPUs enable complex analyses and insights generation that simply can’t be accomplished with CPUs. This GPU revolution runs ML algorithms faster, meaning insights can be delivered in seconds, even for complicated questions.
GPUs scale insights generation in four primary ways:
- GPUs help augmented analytics scale up from a handful of experts to enterprise rollout. For companies to achieve true intelligence with insights, they’ll need enough processing power to support every department and role that needs fast, on-demand answers.
- GPUs enable proactive analysis, finding insights before business users even ask the question. AI can leverage processing power to understand a user’s interests and behaviors, generating insights that are relevant to them in a newsfeed. As data changes, the platform can highlight notable changes or anomalies without being prompted.
- GPUs can help to understand the intent behind a user’s question. One question can contain a multitude of questions within it. Nested in a question like “why are sales increasing this year” is the question “explain the drivers of sales success this year, and identify any areas where we were not successful.” GPUs can answer the entire question, quickly, combining automated research with the work of many collaborators to build a coherent, comprehensive and actionable response.
- GPUs help advance the sophistication of questions that AI can answer. GPUs can support more complex ML algorithms without sacrificing speed or depth, unleashing enormous potential for the future of automated analytics workflows.
Together, augmented analytics and GPUs allow users to get answers to complex questions quickly. Pragmatically, this unique combination is currently implemented with two leading technologies: AnswerRocket’s RocketBots and RAPIDS— an open source initiative started by NVIDIA.
RocketBots are powerful ML algorithms for proactive insight generation.
Here’s how they work:
- RocketBots are invoked with a natural language question or through scheduled or event-triggered analysis.
- The RocketBot gathers the data and uses ML to analyze all possibilities, producing meaningful analysis and insights.
- The RocketBot composes a clear, concise story showcasing the most relevant visualizations and a high-quality insights narrative.
RocketBots allow companies to operationalize data science on demand. After a user asks a question, RocketBots get to work, automating the analytics process so that business people can focus on taking action (instead of waiting for answers or trying to divulge meaning from static dashboards).
Data scientists can also take advantage of this technology by launching their own ML algorithms within AnswerRocket’s platform and adjusting RocketBots based on their deep understanding of the business and its data.
RAPIDS is a software suite for GPU-accelerated data analytics and machine learning. Pioneered by NVIDIA, RAPIDS enables faster, deeper data insights powered by ML and AI. Because it’s open source, software developers can leverage this technology for their custom needs.
The NVIDIA accelerated computing technology has enabled breakthroughs in AI across industries— from driving intelligent retail to optimizing content creation workflows. It’s no surprise that NVIDIA would power the next wave of disruptive AI-driven analytics.
RAPIDS allows RocketBots to analyze an entire data warehouse and return the best answer at lightning speed. For companies with enormous amounts of data and complex analysis needs, RAPIDS provides unprecedented insights that would take a team of analysts days or weeks to uncover.
Together, RocketBots and RAPIDS can automate game-changing analysis.
For example, the Category Overview RocketBot performs a deep-dive analysis into CPG categories, providing insights tailored to category managers. With this one RocketBot, category managers can:
- Uncover the key contributors and detractors impacting their category.
- Compare category performance to competitors with comprehensive market share analysis.
- Gain insight into product attribute performance.
- Forecast future performance.
Essentially, RAPIDS makes sure that RocketBots leave no stone unturned because they have the processing power to evaluate every aspect of a question, every assumption in analysis, and every intention of the user.
What RocketBots and RAPIDS mean for the future
RocketBots and RAPIDS ultimately enable teams to achieve better business outcomes.
For companies that want to identify and drive growth, leaders need to know more than what they’ve gained or lost. They need to know which gains and losses most affected the end results and where to focus their attention to net the most growth.
When these insights are fast, exhaustive, and intelligent, business leaders can achieve an enormous competitive advantage. Understanding “why” in seconds means companies can move the needle in several respects.
First, companies can reduce analytics costs and better distribute resources to high-priority projects, instead of allocating data analysts to churn through routine reports (reports that often contain outdated insights by the time that they’re produced).
Second, companies can act proactively instead of reactively. While business leaders are waiting days to simply understand what happened last quarter, the market is moving. The ability to quickly gain a 360° picture of performance enables companies to move faster than their competition.
GPUs and augmented analytics can scale insights generation for companies, where traditional BI and analytics tools lack the power to do so.
By operationalizing data science, companies can gain the deep and fast insights that they need to achieve critical business outcomes.