Every day, we generate approximately 2.5 quintillion bytes of data. The more data we generate, the more we have to work with to make better decisions for our businesses. When we can support new ideas with solid evidence, things tend to run smoother. But, simply having the big data isn’t enough. You must have data scientists and tools to help break the data down into actionable insights before it offers any real business value or competitive advantage.
Today’s companies are investing more time, effort, and money into data analysis in the hopes of increasing customer satisfaction, making business processes and operations more efficient and becoming clear on business strategy. but for many of them, a data-driven culture still evades them.
Why is this the case?
The truth is many obstacles to developing a data-driven business aren’t technical in nature. They have more to do with your company culture. It’s easy to describe how to include data into your decision-making processes, but it is far more difficult to make this the norm for your employees.
It requires a shift in mindset which presents a major challenge. Let’s look at how you can create a data-driven culture in your organization in 10 steps.
Start at the Top
Companies that have a solid data-driven culture generally have top managers who set the tone. They expect that all decisions must be anchored in data as the norm. They lead by example. Executive leaders at one retail bank go through evidence from controlled market trials to decide on their product launches. At a leading tech firm, senior executives spend half an hour at the start of meetings reading detailed proposal summaries and their supporting data so they can take action based on evidence.
These practices move downward all the way to entry-level employees. It works because employees who wish to be taken seriously must communicate with managers on their terms in their language. By setting an example at the top, you can foster a shift in what becomes normal company-wide. If you want your team to employ the use of data in everything they do, you should do it, too.
Carefully Choose Metrics and KPIs
Leaders can heavily influence behavior by choosing what they want to measure and the metrics and key performance indicators they want employees to use. If your company can profit by anticipating your competitors’ price moves, you could use predictive accuracy through time as a metric. Your data team should continuously make predictions about the extent and direction of these moves as well as the quality of the predictions.
For instance, a popular telecommunications operator wanted to make sure that its network provided customers with the best possible user experience. However, they only collected aggregated statistics on network performance. Because of this, they didn’t know much about what end-users were receiving and the service quality they got. By developing detailed metrics on their customers’ use data and experiences, the company could then create a quantitative analysis of the consumer impact of network upgrades. To accomplish this, the company had to have a better grip on the provenance and consumption of its data than is generally the case.
Don’t Segregate Your Data Scientists
Oftentimes, companies sequester the data scientists within the organization. This results in data scientists and business leaders not knowing enough about each other. Analytics won’t provide value or survive if it operates independently from the rest of the business. You’ll create silos, which will make it hard to vet the data quality and put it to good use. If your organization suffers from an issue, you can address it in two ways.
First, remove boundaries between the business and the data scientists. For example, a global insurance company rotates staff from their centers of excellence into line roles where they scale up proofs-of-concept. Then, they return to the centers. One global commodities trading firm has developed new roles in various functional areas of the business to support analytical sophistication. These roles have direct relationships with the centers of excellence. The particulars themselves don’t matter as much as the principal. You must find new ways to connect domain and technical knowledge if you want to put the company’s data to good use and improve customer experience.
Leading companies use another tactic. Beyond bringing data science closer to the core of the business, they push the business towards data science by insisting that employees understand code and are fluent in the conceptual aspects of quantitative topics. Your senior leaders don’t need to be retrained as machine learning engineers, but leaders of data-centric organizations must have some data literacy.
Ensure Data Access is Open Where Necessary
One of the most common business problems organizations struggle with is the ability to obtain basic data. This situation persists in spite of a variety of efforts to democratize access to data within a company. Deprived of information, analysts can’t do a lot of analysis so it’s impossible for a data-driven culture to develop let alone thrive. Democratization is crucial to optimize data management.
Top companies rely on a simple strategy to stop this issue. Instead of using large but slow programs to reorganize their data, they provide universal access to a few critical measures at the time. For instance, a leading global bank was trying to anticipate loan financing needs. They constructed a standard data layer for the marketing department which focused on the most relevant measures.
In this case, it was core data pertaining to loan balances, terms, and property information along with marketing channel data on how the loans were originated and how that data related to customers’ broader banking relationships. Regardless of the specific initiative, the best choice for the first data to make accessible is the metrics that are on the executive suite agenda. Requiring other numbers to eventually be tied to the data source will dramatically encourage its use, especially in functions like procurement.
Data is useful in all departments and aspects of a business. Sharing data across the entire company is critical.
Everyone knows that it is impossible to achieve absolute certainty. Yet most managers continue to ask their employees for answers without a corresponding measure of confidence. Requiring teams to be explicit and quantitative about their level of uncertainty will produce powerful effects.
It will force the decision-makers to directly deal with potential sources of uncertainty. It forces them to think about whether the data is reliable and if there are enough examples for a reliable model.
It will also ensure that data managers and analysts get a deeper understanding of their models when they have to rigorously evaluate uncertainty. For example, if you fail to adequately adjust to market trends, you should consider building an early warning system to take the trends into account and spot instances that would have otherwise been missed.
Placing emphasis on understanding uncertainty encourages companies to run experiments, which can provide additional guidance on what moves to make next.
Keep Proofs of Concept Simple
With advanced analytics, promising ideas are plenty and you’ll often find that there aren’t as many practical ones as you’d like. It’s typically not until a company tries to put proofs-of-concept into production that the difference is clear. Starting to implement something only to scrap the idea later isn’t good for morale.
It’s much better to engineer proofs-of-concept where a central part of the concept is viable in production. Start by building something that is incredibly simple, then work to increase the level of sophistication. Once your foundation is in place, work to improve each component independently.
Use Specialized Training Only When Needed
Many organizations invest in extensive training efforts only to find that employees rapidly forget what they learned because they don’t get to put it to use right away. Basic skills such as coding should always be a part of fundamental training but it is more effective to train your staff and specialized concepts just before the material is needed. By waiting until shortly before you need it to train your team on something such as the finer points of experimental design, the knowledge tends to stick better.
Remember Analytics Helps Employees, Too
With such a heavy focus on using analytics to help your customers, it’s easy to forget how data fluency can make employees happier. But by taking the time to empower your employees to bring gold out of themselves, you can enable them to take care of things themselves. If the idea of learning new skills to better handle data is presented as an abstract concept, few employees will be excited enough to work to revamp how they do things. But, if you frame it in such a way that immediate goals directly benefit them, such as reducing the amount of rework, collecting frequently needed information, and ultimately saving time, then that chore becomes a choice, ultimately fostering a stronger data culture.
Temporarily Trade Flexibility for Consistency
A lot of companies that depend on data have different data tribes. Each tribe may have its own preferred sources of information, metrics, and favorite programming languages. Across the organization as a whole, this is a recipe for disaster. It’s easy to waste hours trying to reconcile subtly different versions of a metric that needs to be universal.
Inconsistencies in how data modelers do their work will also take us home. If coding languages and standards vary across a business every move data analysts make will include retraining, which makes it harder for them to circulate. It can also be difficult to share ideas internally if everything requires translation.
Instead, focus on canonical metrics and programming languages. You can do this by insisting your new hires know how to code in your language of choice.
Explain Analytical Choices
In the majority of analytical problems, there’s rarely one correct approach. Instead, data scientists need to make choices with different trade-offs. It’s a great idea to ask your team how they approached an issue, the alternative they considered, what they understood the trade-off to be, and why they chose the approach they picked over another option.
Making this part of the process will give your teams a deeper understanding of the approaches and often make them think about a wider set of alternatives or completely rethink their fundamental assumptions.
Organizations as well as the divisions and individuals that comprise them Are more likely to fall back on habit because alternatives seem too risky. Quality data provides evidence to back up hypotheses that give managers confidence the make moves into uncharted territory with somewhat of a safety net. But it’s not enough to aspire to be a data-driven organization. For a company to be driven by data, you have to develop a data culture where this mindset can grow. Leaders need to promote this shift with an example, practice new habits, and develop expectations for what it means to base business decisions on data.