A data strategy must support your strategic goals, and ensure that data collection and utilization of this gives your company a competitive advantage. The data strategy must ensure that your entire data value chain is optimized so that it best supports your business and what you want to use data for. In this blog post I will give an insight into what a data strategy is and why it is so important.

Does your organization have a Data Strategy?
I have not yet met a company that has a strategy for how data should be handled. At best, they have a digitization strategy that focuses on IoT and Business Intelligence. Others have a Big Data strategy, which defines that one should have Big Data, but not how this is achieved and for what purpose.

The concept of data strategy as such is not new. But the execution of a data strategy and how it is written down is not as well defined as a business strategy. It can therefore be difficult to know how to best write down a data strategy, what it should contain, and how to ensure that it is executed and revised. In this blog post I will point out the importance of having a data strategy, and give examples of what a data strategy should contain.

Summary of a Data Strategy

Can you live with your supplier delivering components to you 10 days later than the time you need them? Probably not, which is why for many years you have been optimizing your value chain to be “lean” and efficient. This is exactly what you need to do with your data now. Develop a Data Value Chain that supports what you have strategically chosen to use data for.
It is therefore of no use that we collect data without knowing for what purpose. It’s like buying a new machine because we “may have to use it one day”.
It is also of no use to collect data from the right sources if it is processed too slowly so that you always get information 10 minutes late.
The data strategy should help you create a “lean” data value chain.

elements of a Data Strategy

A data strategy should:

  1. Support your business strategy
  2. Define what data is to be collected and how this is to provide value
  3. Define how data is to be stored
  4. Define how we ensure a high data quality
  5. Be supported by Data Governance
  6. Reflect on how to handle Business Continuity

Let’s' elaborate on the various elements that are in a data strategy and see some examples and pitfalls that we typically see among companies.

1. The data strategy should support your business strategy

The most important thing to keep in mind is that the data strategy must support your business strategy and the strategic goals you have for your business. In other words, you need to reevaluate your data strategy as often as you change your business strategy.
When you start working with sensor data or saving images for image analysis, it places different demands on your data platform than the classic operational data, which we e.g. know from ERP systems. As data begins to gain more power in corporate decision-making, it is important that the entire data value chain supports what you need your data for.

Data Strategy requires a good Business Strategy

A good data strategy is therefore strongly dependent on your company having made a good business strategy with clear and well-defined strategic goals.
Many companies we have advised in Neurospace have a goal of digitizing maintenance. But when has this goal been achieved? How far up the digitalisation ladder do we need to go before we are satisfied?

Digitization ladder

The digitization ladder is often drawn as five steps. If the goal defined in the business strategy is simply digitize maintenance, then is it step 1, 3 or 5? A better goal would therefore be: As maintenance manager, I want to be able to predict breakdowns on our water pump at least 15 days before it occurs, so that we can plan maintenance and order spare parts from the supplier. “ The strategic goal now also sounds more like a user story, which defines goals, desires and needs. With a well-defined strategic goal, you are already five steps ahead of making sure you have the right data to gain value.

2. The Data Strategy provides Right Data

A data strategy should help you select the right data, that supports your company’s strategic goals. You should therefore not start collecting a lot of data, because you might need it one day, but carefully think through what information and data can support your goals, and how these should be used and analyzed so that they create value.

A Data Strategy saves you both time and money

Many companies with a strategic goal of digitizing maintenance have already invested in sensors and completed the first step of the digitization ladder; data collection. However, if you are not already considering how to use this data, you may end up investing in the wrong sensors for the purpose. In the worst case, this means that several years of data collection have been wasted. If you have purchased sensors with too low frequency response you will be forced to invest in new sensors and the previously collected data will thus be useless.
My clear recommendation is therefore that you only collect data when you know for what purpose. In this way, you first and foremost ensure that you do not collect any incorrect data, but at the same time you do not spend resources on storing data that does not give your company any value.

Your data must provide value

Data collected for the purpose of “we may have to use it one day” has a purpose in the company, it is a cost, often a sunk cost. In my opinion, data needs to be collected because it needs to be used in a business to create insight and value. Therefore, you need to constantly think about how data can help your business achieve its strategic goals in a smarter and more efficient way. For example, it can be to be able to utilize the full life cycle of your components, so that you both ensure a longer shelf life on components, but also reduce the probability of breakdowns. If your company has a business goal of having a green profile, this is an important message to send, as you make sure you do not throw out well-functioning components, but at the same time do not compromise on your company’s productivity.
Another place where the green profile can be helped significantly with data and analysis is by using computer vision to perform quality control, and help reduce the amount of defective products that are discarded, and have a significant impact on the environment and your business economy.

3. Define how data is stored

Can you live with the fact that it takes 10 minutes to extract data from your data platform if you have to use this insight every 5 minutes? Hopefully you would be dissatisfied with such a solution.
A data strategy must therefore also take a position on how data is best stored, with the goal in mind. So it is not good that you only update your data once a day if you need a higher availability for your analyzes. You must therefore be aware of which data platforms best support your data type, and how fast this should be able to process your data from input to output.

One of the biggest mistakes I meet today is companies that enter sensor data into their classic relational database (Data Warehouse). A relational database is optimized to handle structure and relationships between the data sources. It is also not easy to scale, and can be expensive to change. Sensor data is time-based so a time series database will therefore be much more suitable, faster and more flexible than the classic relational database. However, as companies do not have a data strategy, previous choices are not reconsidered either, and new data sources are therefore adapted to an already existing system, without it being considered whether it is the right setup for future goals and challenges.
As a company, this will mean that you never develop a product, as this will lead to changes in your production.
Therefore, in your data strategy, you need to assess whether current technologies are sufficient or other technologies should be used.

4. Data Quality

It is one thing to collect data, it is another thing to make sure that the data is of such high quality that it can be used for something value-creating. Therefore, in your data strategy, you must also address how you want to ensure that you have a high data quality. When you start using data for decision making and reporting, this should be of high quality and be reliable.
If you make decisions based on poor data quality, it means you are making poor decisions.

Garbage in - Garbage out

You should therefore consider how you want to ensure:

  1. That data does not contain errors, such as missing values and error measurements
  2. That data can be used for the given problem you want to solve (defined in your business strategy)
  3. That data is used correctly when sufficient has been collected
  4. That data is reliable
  5. That data is collected in the same format (time zones, UTF-8, comma-separated or not, number of decimals)
  6. That data is collected at the same frequency (eg every 5 minutes)
  7. That data can easily be used for a given task (eg unique sensor naming)
  8. That data is not biased

Even if you have an external vendor to handle the data collection for you, it is still your job to ensure a high data quality.

5. Data Governance

Data Governance and Business Continuity are probably some of the biggest elements of the data strategy, and unfortunately often forgotten in companies. Data Governance deals with rules for policies, roles, responsibilities, standards and processes within data management and use. It should help you assess how to ensure that the right people have access to the right data at the right time. And that this data is not available to unauthorized persons. It is about data security, and must, among other things, answer questions such as requirements for data encryption in transit and at rest.

Can you live with another person owning data about your machines or customers?
Data governance should also help you assess the risk and requirements specification of an external provider of either sensors, your data platform or data analytics. One of the things you need to consider here is data ownership. If you do not have the ability to have these items in-house, set requirements for your supplier and make sure they live up to your standards.
Another example is if your company chooses to use an external company to manage your data platform. Here, it often becomes the external supplier’s task to ensure that you have a high data quality. But it requires you to set requirements for them, such as standards for dealing with missing values, frozen data and sensor errors.
My clear recommendation is that internal data is owned and controlled 100% by you. It is therefore your responsibility to ensure that you have a high data quality and that data is stored securely and correctly. You do this by setting requirements for your suppliers, and these requirements are written down in connection with your data strategy.

Business Continuity

Business Continuity is about securing your business in the event of unplanned disruptions such as unplanned downtime, in the event of a fire, or if the partnership ends. Let’s just rest on the latter. If you have an external partner who is responsible for all your data. So what happens to this data if at some point you want to change partner, or move it in-house? Or what happens to your data if your partner’s business ends or goes bankrupt? How do you ensure that you can continue to use your historical data in such situations? It is exactly these risk assessments that you need to make here, as well as form a plan for how to ensure your business’s continued operation in such situations.

Final thoughts

If your business does not yet have a data strategy, it is my clear recommendation to get started as soon as possible.

Neurospace has helped several companies in both utility and production to get started with a data strategy. If you would also like guidance on getting started with your data strategy, you can book a free and non-committal meeting in 45 min with our CEO Rasmus Steiniche.

// Maria Jensen, Machine Learning Engineer @ neurospace