The gap between analytic aspirations and enterprise-scale ability is widening across a number of industries. Successful global market leaders are achieving returns above the cost of capital for their analytics investments, yet many companies are stuck in “pilot purgatory,” eking out small wins but failing to make an enterprise-wide difference.
The economic shock created by the pandemic and its recovery has already highlighted the competitive advantage of effective deployment of AI analytics. No longer does an enterprise’s physical scale translate to margins through procurement and operations. Instead, forward-leaning organizations are leveraging their data to drive margin and share, nurturing it as a strategic asset and applying it in ways that have a concrete impact on their business.
What’s more, the decentralized and matrixed nature of many large companies leads to a more sluggish progress in implementing large-scale analytics or technology transformation programs. While this type of infrastructure has led to marketing and product development prowess, it has hindered the ability to strategically invest in data and analytics platforms or to build the agile ways of working necessary to scale them.
Global market leaders who are winning in AI analytics have focused on executing in critical areas, including three that are particularly challenging:
- Developing the right talent base and operating model
- Building the right data and digital platforms
- Actioning learnings and implementing outputs into operations.
In this eWeek Data Points article, Ryan Grosso, US Head of Data Science at SparkBeyond, will discuss how to bridge the gap between analytic aspirations and ability.
Data Point 1: Develop in-house analytic talent
Many companies have a team of analysts who are well-placed for driving business insight (BI). Yet in order to ensure the success of an analytics project, data science expertise is required.
Driven by the data science skills shortage, new solutions are beginning to emerge that accelerate an analyst’s workflow by automating key activities such as root-cause analysis and model building. Such automation saves analysts from the painstaking process of searching for correlations that prove or disprove an individual hypothesis by allowing them to screen millions of hypotheses at once.
This also reduces the potential for bias as it is no longer incumbent on analysts to determine which datapoint to explore first, or on data scientists to determine which hypotheses to test; instead, they can concentrate on selecting the most relevant to use as insights or building blocks for a machine learning model.
These solutions also lower the technical barrier to entry for machine learning, enabling business analysts to take on more of a lead role – and bringing us a step closer to the democratization of AI.
Data Point 2: Create hybrid teams to foster collaboration
Analytics projects succeed when the process is infused with domain expertise, rather than an overloaded Analytics Center of Excellence (CoE) team running code in a siloed environment.
In order to achieve the critical mass of successful analytics projects across an organization, and reduce the burden on the CoE, training subject matter experts (SMEs) who can “speak data” to data scientists while “speaking business” to executives can be valuable additions to the teams working on analytics projects.
This helps to foster a culture of collaboration between data science experts and business users, enabling data scientists/analysts to focus more on advanced and complex processes while reducing time to access actionable insights for business users.
Data Point 3: Build the right data platform – on the cloud
Multinational companies need to reflect the diversity of the markets they operate within, but this poses a challenge: how do you build scalable AI solutions that support a decentralized business model?
For all AI use cases, data scientists rely upon a suite of tools and processes to ingest and transform data and insert it into a storage solution or app. Before cloud-based technologies reached maturity, building this infrastructure was both costly and time-consuming – local differences in the data stack had a knock-on effect on how data needed to be ingested, processed, and stored.
Enterprise-grade cloud solutions cut through these difficulties by enabling companies to spin up storage solutions as and when needed and make use of an ecosystem of third-party solutions to cope with diverse data ingestion and transformation needs. For example, migrating physical data centers to Microsoft’s Azure Platform helps pave the way for centralized Analytics Centers of Excellence to remain in-step with operational realities on the ground.
Data Point 4: Trust data-driven recommendations
Further downstream, analytics leaders are embracing automated driver discovery platforms to enable insights generation at scale, without sacrificing local market understanding. Previously, analysts and data scientists would search for correlations manually – a process that is slow, relies heavily on local market understanding and is difficult to refresh. By relying on a machine to surface potential drivers, AI solutions can be scaled quickly across geographies but still capture the individual dynamics of each local market.
Data Point 5: Action insights on the frontlines
Early adopters of AI analytics are already unlocking the benefits of operations-focused solutions. For example, a leading global snacks brand achieved a 1.5% uplift in convenience-store sales in a mature Latin American market, just by equipping its field team with store-level assortment recommendations. These insights require a dedicated data strategy that ensures the data stack continues to evolve over time as new sources of data emerge, enabling fast test-and-learn cycles.
Data Point 6: Democratize analytics to spark your analytic transformation
Making an enterprise-wide difference with analytics requires SMEs, CoE and business stakeholders to share, understand and collaborate across the entire analytic process.
AI analytics unlocks effective bottom-up decision-making, while frontline teams are likely to find themselves entrusted with even greater decision-making authority and influence, and an increasing share of investment. Using technology with a no-code/low-code environment and built-in explainability empowers teams to introduce rich, adaptive insights and ML models to business processes company-wide.
About the Author:
Ryan Grosso is US Head of Data Science at SparkBeyond