Analytics intertwined with every aspect of an organization is not an easy process. But if you do, the rewards are worth the time and effort.
The biggest obstacles to creating data-driven businesses are not technical; they are cultural. By getting buy-in from everyone in the organization, your biggest obstacle is gone. But how do you get everyone aligned, involved in the process and excited about changes that put analytics at the heart of your business?
Data-driven culture starts at the top
There are many things in life that are not measurable. Should you marry this person? Should you have a baby? What are we eating? There is no simple algorithm for the hard questions in life. But in an organization, almost everything is measurable with data and metrics.
Therefore, top managers must ensure that all decisions are data-driven. Having a culture of data-driven decisions rather than ad-hoc, arbitrary choices not only instills confidence in the data, but also provides the necessary justification.
This is a simple psychological concept called modeling: simply model the behavior you want to see in others. Others will observe and imitate. This is the only time the trickle-down theory works.
Also see: A call to end-to-end supply chains for data analytics
Don’t think AI is about profitability
Used properly, artificial intelligence (AI) can increase revenue and reduce overheads. While this will make a company more money, short-term thinking in this way results in small changes that don’t yield long-term gains. Adding a one-time analysis of customer information and finding incorrect email addresses fixes a process for a quarter, but doesn’t make significant changes or adds value to the business in the long run.
Successful AI-driven companies see the purpose of AI as transformational. It is an end-to-end process, where automation and AI are applied to every step of daily processes. It should be seen as an opportunity to create a continuous loop of change, feedback and improvement. This will help increase profitability, but this should not be the ‘why’ of AI projects.
Don’t rain on your data scientist’s parade
Organizations often have an ‘IT department’ and ‘everyone’. Within that IT department, there are data scientists who hold the key to business success. The best data scientists have a lot of domain knowledge and an implicit understanding of business goals.
If you don’t give data scientists the chance to interact with staff and analytics, your analytics won’t be as strong as they could be. To eliminate the boundaries between the company, you can take the experts to meetings or walk them through other roles in the company. In this way they can see the daily reality of the company.
By integrating the data scientists into the wider organization, they can identify gaps in processes and gain a deeper understanding of the business that executives cannot always capture. This not only better informs the data science team, but also helps to onboard staff when changes or decisions are made. When people are consulted about the processes they are involved in, it helps to refine and improve them, and then employees are more likely to accept the changes to the systems as they designed them.
By putting more trust in the data and the data scientists, employees can clearly see the positive changes. This in turn will make them more confident in the analysts and will create a positive feedback loop where they can suggest more points for improvement.
This process also inspires psychological ownership, where employees feel involved in the organization and its results. This increases motivation and loyalty and helps keep your people engaged in refining and improving business processes.
Make sure everyone has access to the data they need and the ability to use it
The democratization of analytics is one of the biggest challenges in organizations. Despite efforts to silo data and make it available to everyone, there are still areas of businesses that are cut off from others. Without information, analysts cannot analyze, accountants cannot account, and marketing cannot quantify the results of their campaigns.
Instead of giving everyone access and arranging massive overhauls that never happen, organizations can open small parts of the business at once. The needs of a marketing department will be different from those of accounting, so giving employees easy access to the specific data they need will set the business up for success.
Business leaders must then train their teams to use the tools just before they need to, so they retain the information and are equipped to work effectively with their data. As we think about integrating data scientists more with the rest of the business, consider moving them from “employees” to “coaches” who can share data best practices across business units.
Every organization wants to be better, faster and stronger, but telling teams to just do better isn’t enough. Managers need teams that provide explicit, quantitative levels of uncertainty and outcomes. This does three things.
First, it challenges the data: is it reliable? It then gives analysts a much deeper understanding of their models when analysts need to evaluate uncertainty. Finally, it pushes companies past the shark nets by understanding uncertainty. Statistically rigorous experiments, controlled trials, and crossing your fingers is a good strategy to test before making large-scale changes.
Keep proof-of-concept simple and robust
Nothing but 87% of data science projects go into production† Analytics offers many promising ideas and less practical ones. However, it is not always clear which projects are feasible and which only after a proof of concept has been drawn up. And then, even if it’s doable, the team at the top can put the kibosh on something if it’s too expensive or takes eons to implement.
Employees see the recommendations, make projects, make a proof of concept and then it’s gone. It’s so daunting, especially when the data backs it up. Failure to act when there was a compelling opportunity to do so can negatively impact psychological safety, fuel mistrust of employers, and stop the feedback loop that is so essential to building a data-driven culture.
Instead, the focus should be on creating viable proofs of concept. Start with a simple project and it can be added later, expanding that existing system.
Use automated analytics to make employees’ lives easier
Implementing automated analytics can help businesses move forward in a more meaningful way. By bringing more automation to analytics, data scientists remove tedious and repetitive tasks, the need for rework, and free themselves for the creative, complex tasks.
The organization must not only make the data and decisions understandable to a wide audience, but also explain how people can use data for value in their work. When analytics become useful in an everyday context, people will embrace it. When a staff member can stop doing the absolutely boring data entry job they hate, they’ll love AI more.
Allow flexibility instead of consistency (short term only)
No matter how hard an organization tries, systems evolve. Departments will use tools in different ways, different programming languages, and different data organization systems. While consistency is the goal across the business, teams will feel frustrated by spending countless hours training and enforcing policies that will eventually change.
Short-term goals may one day lead to long-term goals and the systems will begin to coalesce over time. It is important to maintain flexibility and adapt to the technology and processes as they evolve.
In analysis, as with most problems in life, it is rare to have one correct answer. There are usually a few, each with different positives and negatives. What alternatives were considered? What were the considerations? Why was one approach chosen and the other discarded? Employees are more likely to trust and use a model or system if they understand how a decision was made.
Explaining choices made helps people understand why choices were made and helps to get buy-in. If discussed early enough in the process, it also gives people an opportunity to offer explanations, alternatives, or solutions. Make sure that feedback loop works with open and honest contributions.
It can also be a way of explaining why some predictions fail or break models. COVID-19 has changed everything about customer buying patterns and the economy in general. Having clearly visible, explainable AI means that models are easier to fix, or recommendations can be set aside because a model operates on faulty assumptions.
This also helps to suppress the long-term employees who have an attitude of ‘I know better than this’. The data can be used to verify (or discuss) their experiences and to inform them about changes. Data + real-life knowledge = success.
How do you know when analytics is part of the culture?
Here are seven signs when using AI and data for decision making is a habit at every level of the business:
- Analytics is represented at C-suite level
- A high-quality AI strategy that shows what percentage of the processes are automated
- Lowering the cost of analytics per use case
- AI-powered suggestions are implemented faster
- AI system changes are identified faster
- There is a wider range of easily accessible resources to use with AI
- AI is not tactical; it is transformational
Don’t treat AI like it’s the latest trend a company needs to talk about in meetings; make it a real, actionable strategy in every part of the organization. If analytics is just the way of life in an organization, you know it’s embedded and part of the culture.