In the COVID-19 reality, healthcare providers scramble to test and treat patients. They need access to new lab data and they must obtain ongoing input from patients about their symptoms. Some have implemented dashboards that show changes in the conditions of each patient with mild to moderate symptoms through patient self-reporting. When conditions worsen, a virtual consultation is performed. With this information, treatment plans are being rapidly customized. These life-saving activities rely on data and the ability to generate a quick answer to needs and questions that change daily.
While these operations are growing out of unprecedented healthcare demands of the pandemic, they are not new. In the past decade, this type of data-focused activity, known as ad hoc analytics, has grown considerably.
Through ad hoc analytics, leading business intelligence (BI) vendors, and an expanding pool of startup companies, have introduced solutions that generate insight to specific questions using relatively straightforward interfaces. What is common to these solutions is that they allow non-technical users, like nurses and physicians, to interact with the data without necessarily knowing database queries or the technical infrastructure underlying their work.
But another important aspect of ad hoc analytics is that it does not fit well with the common notion of how analytics should develop in a company.
The Linear Vision of Analytics Doesn’t Quite Match Up
If you are focusing on transitioning your company’s operation toward analytics, you have likely seen some version of Gartner’s analytics ascendency model. It has become somewhat of an axiom in the analytics space.
The idea is that you start with rudimentary descriptive analytics, which gives you a basic sense of your data by way of descriptive statistics and dashboards. You then progress the analytics operations by engaging at a higher-level with the data with diagnostic analytics. This helps you identify root causes and develop actionable business strategies. Moving forward, using the data and measures developed in the previous stages, you apply statistical models to make predictions that inform your planning. Finally, you reach the most mature type of analytics, one that is both complicated and financially rewarding, by way of algorithms that prescribe optimal courses of action.
Undoubtedly, there has been a tremendous effort in companies to adopt the idea undergirding this vision of analytics. While never easy (or cheap), developing capabilities along this linear evolution in analytics generally pays off.
But this notion of analytics ignores other types that often grow outside of the data science area in different parts of the company, like ad hoc analytics. Not paying enough attention may mean that insufficient resources are being directed to these developments. More concerning, insights gained in other parts of the company may be excluded from the analytics objectives of the company.
Why Ad Hoc Analytics Has a Bad Reputation
Being removed from the typical data science thinking of analytics, ad hoc analytics is often considered a low-end application — baggage from the old ways of analyzing data haphazardly.
There is validity to this criticism. Users applying ad hoc analytics often keep the knowledge gained from their work to themselves. For instance, being able to generate dashboards and reports more easily also means that ad hoc analytics can discourage communication between users and the data science team. Therefore, the risk is further removing these otherwise isolated activities from the analytics apparatus of the company.
Still, by simplifying access to data and insight generation, this method slashes the time it takes to derive insight from data. It also removes significant overhead from the data science team, so that they can focus more closely on the more complex (linear) application in analytics.
Importantly, ad hoc analytics brings users from different parts of the company closer to working with data and being a contributing part of the analytics apparatus.
Revisiting the healthcare provider example, for instance: Learning from results collected tentatively by client-facing users that seek to address a difficult challenge, the data science team of a hospital can help predict patient demands and prescribe ways to optimize supplies and staff that can be used in the company during, as well as after, the pandemic.
There are several ways a company can successfully implement ad hoc analytics.
- Simplify access to data. The ability to perform ad hoc analytics requires upfront involvement of technical staff to generate queries and connect the BI platform to data assets. Invest time and resources needed to make this work. To benefit even more, implement data lake architecture to ensure that users have transparent, yet secure, access to as many data assets as possible.
- Invest in training. You should not expect everyone to become a data scientist (that’s kind of the idea behind ad hoc analytics). But exposing some of the complexities under the hood tends to help staff understand the potential and limitations of what they can do. It brings them closer to the data and helps them communicate better with more linear developments of analytics in the company.
- Communicate analytics. Silos suffocate insight and innovation. True, some ad hoc reports are generated to answer a very specific question only one individual can use. But many other examples that arrive from ever-changing needs of staff can affect how you think about your analytics priorities more broadly. It is important to facilitate communication about analytics across units to address those changing priorities, especially with the data science team.
- Cultivate analytics liaisons. Identify individuals who are most passionate about data in a department (you can find them during training, for example). Nurture their knowledge in analytics by offering more training, and letting them provide input on the larger, more linear data science pipeline. These invaluable individuals can help their colleagues implement ad hoc analytics and be conduits of knowledge about analytics across the company.
Ultimately, the more users that partake in data-centered work, the higher appreciation and support they have to investments in data and analytics. And when done properly, both the more linear data science apparatus and the more nonlinear ad hoc analytics can complement each other.