Simplify your Data Strategy
Bridging the Gap between Defense and Offense with a Feature-First Approach
In 2017, HBR introduced a data strategy framework, defense vs offense, to elucidate the complexity – read chaos – that can ensue if an organization doesn’t know which side of the ball it’s on, or if it doesn’t know that its position may change based on business objectives or other industry moves.
Defensive data strategy is represented as a single source of truth (SSOT). This data strategy is focused on data quality, integrity, compliance, and security. It’s exactly the type of data you’d need to face an audit. Defensive data is centered around control.
Offensive data strategy is characterized by different cuts of data, versioned from the SSOT, that can be used to derive insight and gain competitive edge. This is the type of data used for analytics or AI/ML, and is tailored to a specific business need. Offensive data is centered around flexibility.
As you can imagine, most organizations are under pressure to deliver on both sides, but traditional data storage and access methods typically do one or the other, not both.
Flexibility vs Control
In many organizations, data infrastructure and integrity is governed by IT departments and requires deep technical expertise of Data Engineers to properly build and maintain. There are typically multiple environments used to develop and test all the data pipelines to make sure everything is working as expected before it is pushed out into production – which is the SSOT.
Data Scientists and other analysts need access to this SSOT, but the data is typically not in a format that is optimal for analysis. Broad access to production data creates risk. To get around this access issue, analysts are often forced to work with pre-aggregated data (check out our blog on pre-aggregation for more info) or to copy data into a development environment in order to preserve the SSOT. Both solutions result in complexity and latency that drastically delay actionable business insights.
Molecula can simplify your data strategy by bridging the gap between defensive and offensive data – without the need for pre-aggregation, copying data, or putting production data at risk – through its product FeatureBase.
FeatureBase: Think Feature-First to Simplify Your Data Strategy
FeatureBase will translate your source data into a highly optimized format that will reduce latency up to 1000x and storage up to 90x – all without copying, moving, or modifying the source data. This optimized format converts raw data into features – which are informative data points to input into models. This process is typically done manually by data scientists. With FeatureBase, source data is automatically converted to a feature-oriented format. Adopting a feature-first approach to your data strategy will give your offensive business the flexibility it needs to drive business outcomes without disturbing the SSOT that your defensive business relies on.
The Benefits of Balance
A balanced approach to data strategy results in a simplified path from data to decision. In addition to tangible cost savings from reductions in storage and processing, your organization can shorten project timelines and ease collaboration – reducing wait, waste, and rework for data engineers, data scientists, and other data-centric roles.
Request a Demo to learn how FeatureBase can supercharge your data strategy
|Key Objectives||Data security, privacy, integrity, quality, regulatory compliance, and governance||Improve competitive position||Sophistication, not brute force|
|Core Activities||Optimize data extraction, standardization, storage, and access||Optimize data analytics and enrichment||Optimize data for analytics and AI/ML with proprietary data extraction, standardization, and storage techniques – without disturbing the underlying data|
|Data-Management Orientation||Control||Flexibility||Control + Flexibility|
|Enabling Architecture||Single Source of Truth||Multiple Versions of Truth||Single Source of Truth that powers multiple use cases|
Abstracted From “WHAT’S YOUR DATA STRATEGY?” BY LEANDRO DALLEMULE AND THOMAS H. DAVENPORT, MAY–JUNE 2017