8 Ways to Get More Value From Your Data
Practical tips for getting started
Many organizations struggle to unlock the full potential of their data. There’s a desire to do more, but obstacles keep blocking the path —analytical and engineering talent shortages, historical data of murky quality, fragmented efforts, not enough hours in the day. Or in many cases, it’s just hard to figure out the best place to start.
If you’re stuck or dissatisfied with your progress, here are eight ways you can start to capitalize on the value of the data you’re already holding, where there’s a significant ROI to be found with even a moderate level of investment.
1. Extracting data from silos
For practical or political reasons, data is kept in functional silos in many organizations. Marketing, customer service, digital, finance…each has its own data sets, and they’re often not brought together to “connect the dots” and to understand macro trends or to gain a holistic view of a customer’s experience.
Surfacing those trends goes beyond acquiring a Tableau, Domo or Looker dashboard to bring the data onto the same screen — it requires analytical and technical effort to align and unify the data, to craft common metrics, and to surface trends which are hard to see with a human eye. But the value that can be extracted by taking this step makes it well worth the effort.
2. Expanding beyond first-party data
First-party data (which a company collects and owns itself) is like what gets illuminated by the headlights of a car — critical territory, but absent a whole lot of broader context. There can be tremendous value in finding relevant, complementary third-party data to enrich and expand your existing data and provide a more comprehensive view.
3. Engaging a statistician
It’s quite common to extract insights from the aggregated metrics available in reporting tools like Google Analytics. Unfortunately, these analyses are often conducted without a proper foundation in statistics. A common mistake is to interpret the output of these reporting tools as evidence of correlation or a causal relationship — and to take action based on those findings, when those patterns are actually fleeting and inconsequential. Having a team member or resource with the proper background in statistics is critical when interpreting data and understanding which insights you’re going to act on.
4. Developing data modeling capabilities
Moving beyond ad-hoc reporting and dashboards, it’s becoming increasingly critical to take the next step towards building data models that can continuously and automatically ingest a wide variety of granular data streams and deliver insights about the performance of a digital product or the broader business it supports.
Simple models using regression analyses, clustering, and other time-honored techniques can deliver a new level of insights that go beyond your current reporting, enabling you to project trends into the future, instead of remaining stuck in the past. Initially, the output of these models can be used to support managerial decision-making; next, the models can be “productionized” to automate those decisions at scale. This brings us to the next point…
5. Experimenting with machine learning and artificial intelligence
Machine Learning and AI — including Natural Language Processing, Computer Vision, Deep Learning, and other techniques — have the potential to extract value from a vast array of content that may have been out of reach previously. As a starting point, advanced ML/AI tools are being made available via relatively simple “plug and play” APIs so you can leverage the technology that Amazon, Google, Microsoft and others have developed. A great way to start is to explore an initial proof-of-concept project to identify what kind of opportunities might exist and shape your ML/AI strategy moving forward.
6. Being intentional about data capture
Frequently, historical decisions over what data to capture were made in fragmented ways by different people at different times. There may be a checkered history of data gaps, tagging errors, inconsistent metrics, and other impediments to getting maximum use from the data. A critical foundational step is to audit existing data and to ensure that future data collection is carried out strategically. As a minimum, this kind of audit can provide greater confidence that you possess the proper raw material for future analytics and data science initiatives, and help you to steer clear of privacy risks and other liabilities that can arise from over-collection.
7. Restoring historical data
When an audit reveals compromised or challenging historical data — due to incorrect tagging of digital interactions, file corruption, inconsistently applied rules, hard-to-decipher Excel files, arcane legacy formats, or otherwise — don’t rush to write that data off. In many cases, the data can be cleaned and restored with a reasonable amount of effort, and it may prove critical in future analysis efforts.
8. Creating a unified data strategy
Finally, and perhaps most importantly, within many organizations, a ‘mixed bag’ of data initiatives has yet to be unified by a cohesive and holistic strategy for how data can support business goals and create a competitive advantage. Without a clearly defined strategy, it’s difficult to develop widespread data literacy, to establish a common vocabulary and protocols, and to unlock the full potential of data initiatives.
Most organizations have real value trapped in their existing data. Considering these possibilities and deciding which is most applicable to your particular situation will help you unlock some of that value.