In our last post, we discussed how implementing analytics as a business process could help improve its success rate. Today, we will dive deeper into what an organization needs to run analytics initiatives successfully. We call that framework the five pillars of analytics. You can think of it as the five elements that you need to get right for your analytics to give the expected value.
Culture
Yes, culture is the number 1 requirement for analytics initiatives to succeed. Not data, not software, culture. Unfortunately, it is also the most underestimated of our five pillars, which probably explains a good proportion of analytics’ failure rate.
What makes an organization culturally analytical? Based on our 100+ projects, we found the following to be the defining lines:
- The organization knows how to identify problems and look for solutions in a scientific way. It usually starts with good reporting processes but goes beyond that. Many organizations still operate in their own ways because “we’ve always done it like that”. However, the findings from analytics initiatives will often challenge the status quo and highlight long-lasting pain points within an organization.
Culture eats strategy for breakfast… data for lunch, and “stacks” for dinner
- The organization knows the value of (good) data and is open to learn from modeling and testing. Collecting and maintaining good data is costly and organizations that don’t see the value of data will often fail to make the necessary investments. Once the data is in place, the organization also needs a culture of testing assumptions and learning from the results.
- The organization understands the need to build long-term solutions at scale. This is where Culture meets Technology. Conversely, many analytics initiatives are treated as one-off projects and their solutions never get the chance to be fully implemented. This also speaks to analytics as a process, not an event.
Process
Business is, first and foremost, a process. This statement also applies to the various components of any given business: Finance & Accounting is a process, Marketing is a process, Sales is a process, etc. Strangely, in most businesses, and especially the ones outside of the tech industry, this state of affairs does not apply to Analytics. However, analytics is most valuable when it is consistently and repeatedly executed over time, just like the other business processes.
What does it mean in practice? Well, the answer would deserve a book but here are some general guiding principles.
- Create a process whereby the needs of the business get channeled to the Analytics team regularly. Think of how businesses run marketing campaigns: in windows of 6 to 8 weeks, defined several months ahead, and based on what the business needs. A good place to start: list all the projects your analytics team has worked on over the past 12 months and identify the most repetitive ones.
Without processes, no results
- Develop an in-house process for your Analytics team to follow every time a request comes in. It includes everything a professional analytics team is supposed to do: locate and collect the data, build and fine-tune a model, generate and interpret results.
- Set up an implementation process that channels the output from the analytical work back into the business. This is as much on the analytics teams as it is on businesses themselves: how do you “translate” the output from an analytics workflow into something that can be implemented? Again, a good place to start is with the projects that your analytics team is already running repeatedly.
Data
The biggest mistake when it comes to data is to confuse quantity and quality. Many organizations believe that they have “good” data because they have terabytes of it. If only!
More important than having a lot of data is having the right data, which means:
- Relevant data. We often see organizations (and their executives) getting frustrated when they find out that their terabytes of data are, for the most part, not very helpful in solving their problems. The better approach is to start with the problem and take stock of the data available. If the data turns out to be unsuitable, this approach saves the business a lot of time and effort.
Having the right data: the starting point, not the end game
- Clean and reliable data. This has been a growing challenge over the past decade, with the rise of data warehouses and data lakes. The data no longer comes directly from the source systems, which causes additional risks of data corruption. Data validation is and remains a necessary step in every project.
- Ongoing data. The true value of analytics comes from its repetition over time, not from one-and-done projects. From that standpoint, the ability to accumulate data on an ongoing basis and have it flow through the analytics process becomes critical, even more so in industries increasingly reliant on real-time data to optimize their decisions.
Skills
When it comes to staffing an analytics team, there are generally two types of organizations: the ones that under-invest, and the ones that over-expect. The first group gets put off by the (comparatively) high salaries paid to the profession; the second one chases the mythical “unicorns” with 15 years of experience in deep learning (a field that made it out of academic labs barely 10 years ago).
In reality, skilled analytics professionals are not that expensive or difficult to find. All you need is:
- Sound business understanding and a good ability to “translate” business problems into analytics workflows. Some of these skills can be acquired in a theoretical framework via the plethora of business courses available online and in business schools. However, the specifics of your organization can only be acquired by way of internal training, including by sending your analytics team “on the field” (may it be a store, a factory, or a warehouse) for a few weeks.
Invest in your analytics team, or perish by it
- Technical skills. Forget about unicorns. Trained analytics professionals are now relatively easy to find, especially at the junior level. There are plenty of master’s programs in analytics to recruit from. These programs equip their graduates with the necessary toolkit to perform in a business environment.
- Ability to think practically about the solution’s future implementation. This is usually the area where analytics professionals struggle the most. A pragmatic solution is to assemble a cross-functional team, involving Analytics but also IT, Finance and Operations, specifically charged with implementing analytics recommendations.
Technology
Technology acts as a magnifier: it will turn your analytics from good to great. It also acts as an enabler for the other pillars: with the right systems, processes can be made more effective, data is easier to collect and maintain, and your analytics team is more efficient. Here is what you need:
- Reporting/BI/Visualization. A good analytics workflow starts with an accurate understanding of the (business) problem, which is usually triggered by solid reporting. This is where BI platforms and other visualization/dashboarding software come into play. Of course, the value of these tools is intertwined with the quality of the data they are built upon.
Yes, you do need an “analytics stack”
- Statistical modeling. If you are serious about analytics, you need to equip your team with the right tools for the job. And no, spreadsheets software do not qualify here. However, there is, here as well, a plethora of options available, from the free and open sources ones to the enterprise solutions. Choosing the right one for your organization depends primarily on your needs.
- “Connectors” to other systems for implementation. This has traditionally been the weakest link for most organizations, primarily for a lack of solutions available. However, the shift to cloud computing has now made it feasible to create full “pipelines” going from analytics models to enterprise systems relatively easily.
Where should you start?
Hopefully, by now you feel confident in your ability to run analytics successfully. You might be wondering where to start: out of the five pillars of analytics, is there one that should be implemented first? Obviously, your specific roadmap will depend on what you already have in place (maybe you have good data but no team to analyze it; or maybe you invested in your systems, but your data is still a mess). However, it is never too early (or too late!) to start building an analytical culture within your organization.
Thank you for reading!
