Modern enterprise growth is heavily reliant on an organization's ability to assess past IT events to then look forward, anticipate and prevent service failures from happening. This is the crux of predictive analytics. Today's fast-moving enterprises have data and expertise locked up in siloed organizations, making it difficult to extract actionable insights, which inevitably impacts the scale, size and speed of a company's growth.
There are two parts to successfully implementing a predictive analytics practice. The first is tech enabled: an open, scalable and unified way to collect, search, aggregate and analyze millions of metrics and logs across networks, infrastructure and applications. The second is talent enabled: finding the right skills to develop insights with domain-specific context for each business unit to carry forward.
The second part is one that most organizations struggle with. Do we build a dedicated, highly specialized team? Do we use a consultant or develop in-house talent? Do we train a cohort of non-IT employees on a licensed platform or do we reserve those capabilities solely for the IT department?
The following are steps to build the best predictive analytics team:
Identify the right representation of expertise
Just like building an app requires a marriage of design experts and full stack developers, a good analytics team needs to start with a mix of domain experts and data scientists. Once they have established some of the basic parameters, you can integrate developers into the mix. Depending on how accessible the data is, developers could provide data scientists with the tools to run the most optimal machine learning algorithms.
Use existing in-house talent and build additional support
There isn't always a need for a separate "predictive analytics team." Integrate data scientists into the functional teams and provide them with access to the data and ensure they can consult with domain experts. If you have multiple groups across the company, create an overlay team (or guild) of data scientists so that they have a forum for knowledge sharing, especially on the latest developments in AI/ML.
Recruit unconventional talent
Data scientists are in high demand, and while it's easier to find them in the hot tech markets of Silicon Valley and Boston, you will face tough competition in attracting and retaining new talent. So, rather than looking in traditional fields like Computer Science and Statistics, cast a wider net to include quantitative fields like Physics, Chemistry, Economics, Biostatistics, etc.
Set a framework to iterate upon
1. Assess the problem.
2. Compile and correlate all relevant data to the said problem.
3. Identify all needed tools for analyzing the data (machine learning, etc.)
4. Make data available to domain experts so you can evaluate the results and iterate on the assumptions.
Include employees outside of the tech walls
Once you have initial results, take the time to share and validate it across the organization. Input from support and services will go a long way in validating the results and gaining further insight into the data. For example, if a company is seeking to improve customer renewal health, share the dashboards/results that you generate with the customer success team so that they utilize this knowledge to improve renewals. Validate the outcome and make tweaks to improve the end-to-end process.
A notable value of predictive analytics is the ability to identify trends and patterns and to formulate different questions. These outputs will inevitably require more analysis and lead you down the path of discovery. So, the more cohesive and responsive your predictive analytics team is, the more poised your company is for dynamic growth.
