How can an organization leverage data as a strategic asset? Data is expensive. You have to invest in data capture and cleaning, hosting and maintenance, data engineers, data analysts, risks, data quality, information presentation, etc.
If done well and in an orderly manner, an organization can harness information to thrive in its business. From MIT, Erik Brynjolfsson, comments that organizations based on data obtain a productivity of 5% - 6% higher than the competition. They also made their assets, return on capital and market value more efficient.
To become effective requires a global data culture that combines quality, access, knowledge and appropriate decision-making processes.
single data source
A single, centralized, quality-controlled data source should be the single point of enterprise data extraction. When métrixas are extracted from different systems, different results are inevitably generated.
This causes discrepancies that slow down decision-making and opposing visions in the scenarios, in addition to the lack of a single and faithful certainty about reality.
In addition, results may appear based on obsolete, inaccurate, or in general inappropriate data, which are being operated by good professionals but produce false results.
When that single source exists, it provides superior value to the organization: the decision makers. Data search times are streamlined throughout the organization and investment is made in their use.
Additionally, the data sources are better organized, documented, related and enriched. Therefore, users are better positioned to take advantage of the data and find the information.
Knowing where to get the data and providing quality results is essential for final decision making
For the data manager, a single source is also preferable. It facilitates documentation, avoids nomenclature conflicts in the structure (tables, fields), load and data quality controls and ensures that identifiers are consistent in all tables. It's also easier to provide flat, easy-to-work views of key relationships and entities that may have come from different sources.
In large organizations, there are often historical reasons why data is stored in silos.
For example, large organizations are more likely to acquire data systems through intercompany acquisitions, or by purchasing complementary data providers, or for regulatory issues resulting in standalone systems in addition to the core operating system.
Thus, a single source, while a large and complex investment, allows the central data team to provide official guidance, listing what is available, where it is, where there are multiple sources, the best place to get it, and other results that produce a great advantage for the organization.
Data Dictionary
Having a good data dictionary makes it easy for users to know what data fields and metrics mean.
This aspect is often an insurmountable obstacle for many organizations. When you don't have a clear list of metrics and their definitions, people make assumptions, which can differ from everyone else's and lead to divergences in results and opinions.
A company needs to generate a glossary with clear, unequivocal and agreed definitions. This requires collaboration with all key stakeholders and business experts.
First, you need to accept those official definitions; prevent teams from going rogue with their own, separate version of a metric.
Second, it's often not the core definition where people's understandings differ, but how to handle edge cases. For example, while everyone may have a common understanding of what an “orders placed” metric means, they may differ in how they want or expect to handle cancellations, split orders, or fraud.
Those scenarios need to be presented, discussed and resolved. The objective here is to reach consensus in two possible ways: either to combine multiple similar metrics into a single common metric, or to identify the situations in which it is really necessary to divide a metric into two or more separate metrics to obtain different perspectives.
Having high-quality, centralized data is not effective if it is not understandable by the organization
Specificity in well-chosen table and field names and unambiguous definitions with examples are key here. It is better to lean towards long but descriptive names, such as “orders_not_canceled” or “% conversion from journeys created to journeys completed” than shorter names that users think they understand and that produce personal interpretations, not common or accepted by all.
Access to information
Having centralized and quality data, with a clear and unified description, is of no use to the organization if people cannot access it. Organizations must ensure accessibility to data wherever they may be useful.
This does not mean giving access to all staff or not having policies in accordance with regulation and controls. It means harmonizing security with the needs of the organization from a holistic point of view.
Foster a culture so that people know what data is available and the security, compliance and regulatory criteria that impact it
It is the front-line staff (customer service for example dealing with a dissatisfied customer or a warehouse worker who finds some damaged stock) who take advantage of the data immediately to determine what actions to take next. With proper training, they are in the best position to resolve a situation, determine workflow changes, or handle a customer complaint.
Fostering culture among people involves training, a data dictionary, access to and understanding of KPIs, unified criteria for understanding data, availability, genuine and tangible internal use cases, as well as making them feel comfortable requesting both training and internal access processes.
Bureaucratic procedures must be streamlined so that, with an adequate approval and supervision process, people have access without too many delays or obstacles. Of course, revocation systems are also required when it is necessary to apply it.
data culture
As the organization is data-driven with broad access to data, people will be faced with key reports, dashboards, analytics and metrics, which they may have the opportunity to understand and analyze for themselves.
To do this effectively, there must be sufficient knowledge of the information, where it comes from, what it means and the impacts on the business and its function.
Data culture is a multi-pronged effort. It also encourages engaging both clients and employees at various skill levels and functions that require a personalized approach to each case.
The most widespread area is training in data science. This includes an introduction to the most advanced and computational machine learning and data mining approaches for extracting insights from data, as well as creating data products such as recommendation engines and other predictive models. This involves focusing on the top of the skills pyramid for more technologically advanced users.
One of the most effective aspects for the organization and customers is to train people with standard user technology skills to become data scientists.
For example, the pharmaceutical and financial sectors have business analysts who are well versed in statistics but weaker in the computational domain. Technological profiles have talented programming profiles that lack business statistical rigor. Training statisticians in programming and programmers in business statistics produce quick benefits for the organization.
The data is not there to bolster or undermine existing decisions, but to help inform future ones.
Starting by improving basic skills in descriptive statistics is often the easiest and most effective. That is, the basic forms of constructing data summaries: mean, percentiles, range, standard deviation, etc., and detecting when these results are appropriate or not, give meaning to the data that make them up .
For example, when data is found to be highly skewed, such as home prices or income, the median is the appropriate metric to summarize the data, not the mean.
Empowering people to rely less on assumptions and focus on plotting and examining the data and the proper use of proper summary metrics has a significant payoff in time and proper decision making.
Another area of effectiveness is data visualization skills. Charts are often full of non-relevant information, unnecessary clutter, and annotations that detract from the key objective. Either you use inappropriate chart types, such as multiple pie charts, each with a large number of segments, or you choose a color scheme that makes it very difficult or impossible to interpret.
It produces great wear, at a personal and organizational level, to spend effort in the collection and analysis, to fail and reduce the impact at the end point of the presentation. A little data visualization training goes a long way and makes data clearer, more digestible, and ultimately more likely to be used appropriately.
A next level of complexity in culture is inferential statistics. These are the standard, objective statistical tests used to detect, for example, whether a trend or difference in website traffic from week to week is likely real or just random variation.
The point is not that a business manager or customer service agent can perform these tests, but rather to make them aware of how statistics can be useful: to understand correlation versus causation and to appreciate that forecasts always come with uncertainty. For decision makers and managers, this also gives them the skills to correct work quality or spot motivation when the data doesn't support their conclusions.
decision making
Above all, data has a real impact on the decision-making process. At this point, the organization can have quality, relevant data, trained analysts, and carefully crafted and presented reports. If the reports are not used or decisions are made without consulting them, regardless of what the data shows, then the entire investment is wasted.
HiPPOs, “highest paid person opinion”, are the antithesis of data management. They don't care what the data says, especially when it doesn't agree with their preconceived notions, and they're going to stick with their plan because they know what is best. In addition, they usually have a high command position.
HiPPOs are detrimental to companies, because they base their decisions on poorly understood metrics at best or mere guesswork. No smart tools to derive meaning from the full spectrum of customer interactions and assess the how, when, where, and why behind the actions. The HiPPO approach can be crippling for businesses.
Financial Times
HiPPO is the acronym for "Highest Paid Person's Opinion" is a term coined by Avinash Kaushik. It is used to describe the tendency of lower-paid employees to give in to employees with higher salaries and years of experience when a decision must be made. The term is also often used to describe an organization's reliance on human instinct rather than data in the decision-making process.
Quite often, organizations have a culture where intuition, free comment is valued, or there is a lack of accountability. In one survey, only 19 percent of respondents said decision makers are responsible for actions on those decisions. It is in these environments that HiPPOs thrive.
One way to counter HiPPOs is to cultivate a culture of objective experimentation, such as A/B testing. In those scenarios, whether it's a change in website design or marketing messaging, success metrics can be visualized and the minimum sample sizes. The key is to have a clear analysis plan and establish the success metrics and any predictions before the tests are run. In other words, prevent HiPPOs from selecting the results after the fact. The same goes for any pilot or proof of concept.
Conclusion
The data-driven culture is part of a multi-step process to consolidate both formally and over time in people's mindsets:
- Single, quality data source from which analytics can flow.
- A data dictionary agreed upon and assumed by the entire organization on the meaning of the data.
- Broad access to data to enable the application of collective business expertise in data analysis.
- Good training that helps promote and reinforce data literacy.
- The resulting reports must be in the hands of decision-makers who trust the data and affect decision-making.
