The phrase ‘data is the new oil’ is increasingly cited, yet the engines they are supposed to fuel never leave the shop: 87 percent of data science projects fail to make it into production. Similar to any digital transformation initiative, many of these stalled projects consume valuable resources and fall victim to pilot purgatory.
Organizations cannot afford to let this data – a valuable internal resource – go to waste in today’s increasingly data-driven world.
Why are companies struggling to extract value from this increasingly prevalent digital gold? The answer lies with challenges around data maturity and how a company sources, standardizes, and analyzes data. As they turn to cutting-edge technologies like artificial intelligence, the stakes to tap into this lucrative data and compete in this ‘data race’ run high.
Below we’ll explore some data best practices – in context to their applicability in artificial intelligence – that should be considered when embarking on a data-driven journey.
Data maturity varies greatly with different degrees of formalized data management practices across an organization’s products, processes, and people. Many attempt to store all enterprise data in data lakes, but research shows 80% do not include effective management capabilities.
What’s more is poor data management practices compound on employees: data professionals spend the majority of their time searching and preparing data and only 27% spent analyzing insights. If companies are going to make data-driven actions, they will need to internally harness this enterprise data.
This is an important prerequisite to implementing any data-oriented project, but unique considerations do arise for artificial intelligence. Many see AI as the answer to all of their data and business problems; however, this mindset ultimately leads many down the path of failed data projects. Being honest about data maturity will diminish unrealistic expectations and favorably position the company for future growth.
While AI may not be the answer to everything, it can be the answer to pressing business problems. An AI data strategy should be oriented to solving that business problem and requires relevant data to do so. To be clear, this is does not mean sourcing the entirety of company data collected since its inception.
Establishing a scope for data sourcing is a critical first step, but there are still challenges as data takes many shapes and forms in the modern enterprise. For example, only 20% of companies have created a common data model or data architecture plan across their IT and OT data.
With increasingly ubiquitous computing to harness this data from advancements in the public cloud, processing power, and other technological innovations, AI is increasingly a cost-effective endeavor for enterprises. It becomes not cost efficient when companies scope creep and run AI models across bevies of different data. This can take many different forms for industrial enterprises such as unstructured video, images, and voice data and more structured CAD files, ERP systems, and HR documents.
Capturing, sourcing, and processing a vast amount of historical data for training and validation is required to build an AI model. This AI data can vary greatly in accessibility, usability, and impact. For instance, an asset efficiency use case may require historical systems of record, time-series data from industrial systems, or real-time performance and health IoT-generated data.
It may be challenging but overcoming barriers to obtaining this data and standardizing it for the use case will pay dividends for running analytics further downstream in the process.
The ultimate goal of aggregating this data is to create an analytics model trained to make predictions in the real world.
The traditional journey for industrial companies in their analytics program starts in descriptive analytics and pushes increasingly to predictive. On this analytics maturity scale, using logic for descriptive analytics describes ‘what has happened’. These models analyze data fluctuations and anomalies from user-defined calculations and thresholds to create alerts, such as sensor readings for temperature, pressure, or power, common in factory equipment.
Diagnostic analytics drive further down into the root cause of an issue and ask ‘why did it happen?’. This could include a service technician drilling-down into a deployed machine’s behavioral data to determine why it is malfunctioning and offer potential remediating service actions.
The evolution of analytics has increasingly pushed into the predictive era where businesses are enthralled with answering ‘what will happen next’. This is where the field of AI is becoming increasingly pervasive. These forms of AI logic mostly pertain to those that can learn without explicit programming and recognize patterns in vast sums of data to predict outcomes.
Industrial companies are using these models today to answer long-standing questions for ‘what will happen next’ for complex industrial equipment in their production environment. Using ML models for predictive maintenance with real-time failure and risk analysis can save millions in downtime costs.
With massive digitization of information across the world there is less ambiguity. This information to make decisions is now more readily available to consumers and business alike. With this onslaught of data entering the world, the next decade will largely benefit those who harness and abstract insights from it.
Nine out of 10 companies claim data-driven decision making is important to their business, yet only 57% are actively doing it.
This gap between data aspirations and capabilities will inevitably close and minimize ‘judgement calls’ for business worldwide. AI is the most promising path to make these decisions, glean insights, and minimize business risks.
AI will deliver transformational changes for companies in the near-future; AI spend will grow from $50.1 billion in 2020 to more than $110 billion in 2024. Those looking to benefit from AI must consider their current data maturity, source and standardize the relevant data to their business problem and DX program, and apply the right logic that’ll shine data-driven insights through previous clouds of gut judgments.
Companies successful in undertaking these data and AI initiatives today will be the future industry leaders for the next decade.