Predictive analytics could be the future, but we must solve the data problem first

“Your voice is breaking up.” “We lost you for a minute there.”

How many times have we all heard or said these things? Or what about the white wheel of endless buffering? We’ve all experienced the endless glitches, outages and broken app experiences that impact us more than we might care to admit. 

For enterprises, the move to the cloud and reliance on SaaS apps has made the internet the corporate backbone. The internet is the digital supply chain that ensures users have a great digital experience. But there’s zero certainty in knowing if your app and its multitude of components distributed across multiple cloud environments are actually performing up to par. 

So for today’s organizations looking to be more proactive and automated in how they operate and manage their environments: Is delivering a predictable digital experience across an unpredictable internet environment even in the cards? I would argue the answer is yes, but only if you solve the data problem first. 

The data problem that is the internet

IT operations is an overwhelming place to be right now. In today’s connected world, where every business, application, and device relies on a digital connection every hour of every day, driving superior digital experiences is critical. But with apps running in the cloud and being accessed from many remote endpoints, the number of new blind spots has created massive challenges for anyone tasked to troubleshoot broken user experiences. This complexity creates a networking model plagued by reactive-based troubleshooting, and user experience is regularly degraded. 

Network professionals tell us that responding to disruptions and accommodating new business needs are their top two network challenges. For these businesses, the pursuit of predictive intelligence is all about the ability to move from reactive to preventative, thereby pinpointing issues before they begin to affect user experience. Forecasting and taking back control over what is happening across the cloud have now become core to the enterprise network.  

Predictive intelligence: Unlocking efficiency gains and opportunities

But predictive intelligence promises real productivity gains. For organizations with hybrid workforces, the gains can be significant. Predictively identifying a single service affecting fault and remediating it — such as by switching providers and paths that carry app traffic during peak periods — could save a single employee hours of downtime or degraded performance. Multiplied across the employee base, that number quickly becomes material. 

The same is true for satisfying consumer demand. In the age of exponential choice, proactively preventing any disruption is key to delivering the always-on digital experience buyers need and demand. In fact, expectations of digital experiences have soared.

Unlocking efficiency gains and opportunities to drive brand value is the real payback of predictive intelligence.

Sizing up the data-shaped challenge in predictive intelligence

Troubleshooting is a largely reactive endeavor based on analysis and informed decision-making to improve situations or highlight potential root causes of an active incident.

Determining what’s going, or has gone wrong, addresses an immediate need, but it doesn’t do anything to escape that cycle of users deserting your lagging application or unavailable cloud service.

That’s the promise of the predictive Internet: The ability to leverage a rich data set and visualizations to analyze historical patterns across a complex mesh of owned and third-party networks to predict outages or service degradation and take remedial actions before the effects are felt by users. 

Predictive intelligence at this level is both a data problem and a scale problem. Solving these is key to making it an implementable reality.

It takes an enormous amount of data to predict the beginnings of a degradation or performance deterioration with a high degree of accuracy. Although the volume of data needed to train a model has existed for some time, the data often wasn’t as clean as it needed to be. That caused flow–on effects in statistical models. Without good data, the models simply weren’t capable of producing granular assessments and actionable recommendations.

With the modelling technology now mature and supported by high-quality data collected from across a customer’s wide area network, predictive intelligence is firmly within reach.

A guiding hand

So what does predictive intelligence look like today? It starts with visibility and ends with trust. Data-driven visibility that provides insight into the cloud and internet environments that an organization doesn’t own — but that has become part of a corporate network and thereby critical as a delivery mechanism of digital experiences — is critical. And just as important is complementing that visibility with owned data from an analytics model that learns from past behavior and forecasts future events.

Third, and perhaps most importantly, is recommending what action to take based on data and insight of continuous performance measurement and assessment. Giving up control of IT infrastructure is an impossible ask without building trust first. Recommendations build trust. Trust that the data is right, and trust that the recommended action will provide the intended outcome. 

Predictive intelligence should be thought of as a guiding hand that helps businesses see and measure performance across all networks that impact the user experience, forecasts issues based on historical data and influences decision-making.

Mohit Lad is cofounder and GM of Cisco ThousandEyes.

DataDecisionMakers

Welcome to the VentureBeat community!

DataDecisionMakers is where experts, including the technical people doing data work, can share data-related insights and innovation.

If you want to read about cutting-edge ideas and up-to-date information, best practices, and the future of data and data tech, join us at DataDecisionMakers.

You might even consider contributing an article of your own!

Read More From DataDecisionMakers

Source