Data-Driven Corporate Culture
Data-Driven Corporate Culture
The current crisis has changed the way we do business forever but in hindsight, it has also given us the time to reflect on our habits and lean more towards efficient business practices, towards a data-driven corporate culture.
Can corporates become data-driven?
Successful tech companies have to be data-driven. They only can survive in the market by measuring everything they do, everything their customers do, and everything else. Envious of these disruptors, traditional corporates have been wondering for a while if they could apply the same principles? Reach the same level of value creation? It created many digital transformation initiatives. But can traditional corporates become as obsessed with measuring as companies that started this way: Amazon, Google, Alibaba, Tencent? Organizations perpetuate the behavior that made them successful. When companies built themselves on the shoulders of charismatic, daredevil, intuitive leaders, they won’t easily let data take the place of these leaders' intuition and boldness. Being data-driven means to submit oneself to the power of data, the power of truth, the power of the market. These leaders must change their point of view, their mindset, their believes. The company leadership team must instill a data-driven culture. Otherwise, nothing will change. How many leaders will have the humility to be proven wrong by data? How many will recognize that the world has changed since they learned their trade? How many will become data-driven leaders? The answer to all these questions remains subjective, but from what I witness a movement has started. A movement towards data-driven culture. Data never lies. It is free from all reservations and human ego. It does not bind with assumptions or possibly cannot be corrupted and we can take advantage of that, by becoming more data driven. Now that we understand just how important data driven culture is, what many of you will be wondering is how many data scientists do you need to be data driven?
How many data scientists does it take…?
The first steps corporates take in their data-driven transformation are often to create a data science team, launch a data audit, a data warehouse or a data lake project, and install analytic tools on top. After that budget is gone, they look at what is left to do. In that scenario, what is left is to deliver value and start the transformation because an over skilled isolated data team cannot achieve much. Creating a dedicated data team will not change the culture of the company. On the opposite, it sends the signal that data is somebody’s else problem when it should be everybody’s concern. Having a team producing more analysis and more insights does not do any good if the decision-making process is not taking this new data into account. The answer is zero. You don’t need any data scientist to become data-driven. You have to raise awareness of the risk of relying on intuition to make decisions and the competitive advantage of backing every decision with hard evidence. First, work with the data you have. Later, you may need help from a data scientist to get into more advanced analytics and create new insights to make more informed decisions.
Another issue to be addressed is data ownership and trying not to kill the messenger along the way. Corporate politics often forces messengers of data to sugar coat the data to impress the clients. This selfish act of self-interest often results in long term failure in generating results for companies.
Killing the messenger vs. data ownership?
In a data driven culture, employees demonstrate a strong sense of data ownership. You need to know your data like the back of their hands, the sources of information, the manipulations, the business rules. The data must speak to you. Your manager must not be able to take you off-guard in a meeting by asking you: “why”? You must be able to explain why one of their key indicators is what it is, with reason, and hopefully, more data. The only reason to blame a data owner is if they don’t have an in-depth knowledge of her data. As for the data themselves, they are what they are. They have to be representative of reality. Too many corporates have a leadership culture of killing the messenger, confusing the data owner with the data. When you kill the messenger of bad news, nobody wants to present the bad news, but only the good news. That leads to manipulation of data to make it look good, distancing it from the underlying reality. A data driven culture aspires to be as close as possible to reality. Avoid behaviours that can directly or indirectly drive you further away from reality.
Should all decisions be data-driven?
When asked outside of context, the answer is yes. When asked in context, you may get different answers. There may be voices to consider other, non-quantifiable aspects, to rely on experience and intuition. It is a dangerous path:
- It is a way to give up on data-driven decisions without trying hard enough to support a decision with information. It is rarely easy to do so.
- While a rational decision-making process can be analyzed and improved in time, a mix data intuition process is the worst of both worlds.
There is one situation where one should not look for more data: if after careful analysis, you have two equally valid options. When there is no clear winner between two alternatives, looking for more data can be counter-productive. The best option in that scenario is a coin toss:
- If there were an obvious option, the first set of data would demonstrate that.
- More data most likely will not make much difference and only leads to analysis paralysis. • After choosing one of the options, there will be no comparison with the “road not taken.”
- After choosing one option, you will adjust the course based on how the execution unfolds.
A data-driven business has enough trust in its data and analysis to make and support a 50/50 decision.
Stop predicting the future
We live in unpredictable times. Remember the predictions you heard from experts a few months ago. Fortunately for them, experts are seldom paid on the accuracy of their predictions. The ones you see on major media are just paid by how credible, or incredible, their predictions sound at that time. But everybody else is judged on results. A prediction is usually a support for a decision. If it isn’t, it won’t have much consequence. We try to predict the future, to anticipate and position ourselves towards what we believe will happen. But, it seldom works. In times of uncertainty, it is even more of a folly. Furthermore, predictions take time. Accumulate data, run predictive models, assess the accuracy, list options, run alternate scenarios, puts you at risk of delaying your decision making until your original data is not relevant. A more productive approach is to make conditional decisions. That entails preparing several possible choices and the action plans associated with it and identified the trigger conditions of each of the scenarios. As situations unfold, we continuously collect information and evaluate the conditions to trigger the pre-assessed decision. This leads to rapid actions, tied to the most recent data.
Crimes against data
I realize that crime is a strong word and that it may seem inappropriate in the context, but it is. Data owners must protect data, or it is at risk of being abused. Being data-driven implies a form of respect for data and careful data manipulation. Too many companies are guilty of crimes against data, as they want to look data-driven but end up being the opposite. - Abduction: keeping critical data to oneself, as one perceives that data is power. Sharing data inside and outside the organization can create more value than keeping it locked down. - Torture: when one is trying to force the data to tell a different story, then it would naturally do. Too often, we approach data looking for confirmation of something we already believe. Let the data tell its story and then draw your conclusion. - Killing: one purposefully removes data from a report, an analysis, a statement. This type of manipulation can influence decision making in a direction that the full data would not allow. It frequently happens when the situation is not as good as one would like, and the boss has a history of killing the messenger. Respect and for care for data is a prerequisite to starting a data-driven culture. Have you witnessed any of these crimes recently? Committed some?
Why should you be data-driven?
It is obvious why companies want to be more data-driven, but what’s in it for you? Working for a data-driven company has several advantages for employees:
- Efficiency: you will be more efficient. As you build up decision making processed based on data, you gain in productivity.
- Accuracy: you will take, on average, less poor decisions. Nobody likes to make wrong decisions.
- Stress: the uncertainty of deciding without hard evidence can be stressful.
- Employability: if the rest of your industry is operating at a higher level than you, you lose attractiveness on the market.
- Growth: becoming data-driven is not easy. It will force you to develop new intellectual muscles, learn new skills, and, most importantly, change your mindset.
- Career: data-driven is the future of all functions: marketing, management, HR, information system, operations, finance. Whichever department you work in, your career will benefit from all of the above.
When corporates try to launch data initiatives, they often put forward the benefits for the firm. Adding the benefits for the employees has a better chance of inspiring their willingness to become data-driven.
We need less data
I often hear corporate telling us that they don’t have enough data. After a few discussions, we sometimes manage to convince them that they already have too much data that they are not leveraging. Here are the signs that you have too much data:
- Poor data quality: this is the number one reason for reducing the amount of information collected. Quality above quantity is a common mantra but rarely followed.
- Incomplete data: in a survey, or a registration, too many questions lead to less completion rate. When a piece of information is not mandatory in a form or a screen, why include it at all?
- Not used: if no report, no decision-making process, no segmentation, nobody is using a piece of information, discard it.
- Silos: it is easier to consolidate simple dataset then complex ones. We generate more insights from simple data models gathered from all the systems than trying to create comprehensive reconciliations between some systems
- Understanding: nobody understands the entire data model anymore. It takes a meeting with seven experts to explain a single end to end process.
- Less is more: removing data and processes can generate enormous value and open space for the modernization of information systems.
Making sense of the data.
The best-seller Moneyball quotes Bill James writing: “When the numbers acquire the significance of language, they acquire the power to do all of the things which language can do: to become fiction and drama and poetry.” A simpler version is that data must make sense. When leverage data with simple models, it is easy to rationalize and keep it under control. But trouble begins when the models are too sophisticated that nobody understands anymore what is happening. There is an invisible threshold, when models are blindly optimized for efficiency, without a sense of purpose.
In 2013, Andrew McFee published an article in Harvard Business Review titled: “Big Data’s Biggest Challenge? Convincing People NOT to Trust Their Judgment.” Often end-users of big data and machine learning don’t see the logic in the output and don’t follow the recommendation. That means there is missing information to make sense. So McFee’s prediction that “as the amount of data goes up, the importance of human judgment should go down” is both scary and wrong. Technologies are tools to augment human capabilities, not replace them. Using data, analytics, and machine learning must help us make sense of our environment, and enhance our decision making, not replace it.
Where do we start?
A data-driven culture starts at the top. Every element of a corporate culture trickles down from the management. Leading by example have a better chance of nudging employees to new behavior and gradually shift their mindset. Establishing a data-driven culture is no exception. I wouldn’t beat on a robust culture of data where the management continues to base decision making on their expertise, don’t question the information they see and don’t dig for insights themselves. I understand the uneasiness of changing the way one runs a department, division, or company.
A transformation towards a data-driven culture must come from an explicit and didactic change of behavior of the top management. With a data-driven CEO at the helm, there is be no tolerance for any other kind of leadership then a data-driven one. What does a data-driven CEO do? How to recognize one? How to hire, train, and coach a data-driven CEO? There are plenty of models: Jeff Bezos, Satya Nadella, Mark Zuckerberg, Tim Cook. I would love to attend one of these CEO’s staff meetings and one-on-ones with their managers. I would especially pay attention to the questions they ask before making a decision, and their reaction to surprising insights.