ROI is the Secret Sauce for Sustainable Digital Transformation

An article by Zachary Burke, Systems Engineer

Awd 1 to site

The Problem

Digital Transformation is a broad term that can mean different things to different stakeholders – hence, it’s difficult to lay out a universally applicable sequence of steps that’s right for everyone. Even within the same industry, different organizations will likely need to go about this slightly differently. Even more confounding is that an organization may even be comprised of individual segments at different points along their digital transformation journey.

And this jagged progression isn’t necessarily an accident. Digital transformation needs to have an ROI associated with it to be sustainable. It’s not hard to argue that one must spend money to make money, but the question of whether the money spent will be returned with earnings on top of it requires careful consideration of business priorities. In other words, what we’d describe as “digital” isn’t the end -  It’s the means towards an end.

So, how to assign an ROI to Digital Transformation? I think it is useful to talk about Digital Transformation in terms of levels, or stages of maturity.

Read about the 5 stages to go through this:

1.      The Experimental Stage

This often happens when an organization acquires some basic tooling but with no cohesive vision for what exactly it hopes to achieve. Solutions built during this stage are commonly created by intellectually curious individuals attempting to solve small-scale problems. Individual contributions may actually have characteristics of many later stages but won’t be scalable or repeatable.


2.  The Aggregation Stage

This level is characterized by a concentrated effort to gather data in one place. Even here you’ll see vastly different approaches between organizations and groups within these organizations. For instance, some want to collect as much data as they can and try to organize it for their own people to be able to do something with it. Others are a bit more focused on the target business area. To be of value, data at this stage shouldn’t just be gathered and dumped into a data lake. It really should have some context added to it so it’s readily apparent what different data streams reflect.


3. The KPI (Key Performance Indicator) Stage

When well designed, KPIs are reflections of how well an enterprise, division, site, area, or unit is doing. When KPIs are generated over longer time periods and calculated in real-time, they do two very important things. First, they guide strategy in the form of both long term and short term operations planning. The calculation for a given KPI acts as an optimization function, and this makes it clear where improvement efforts would provide the most value. Secondly, by setting tolerable ranges for KPI’s, they can form the basis of an alerting system. If a KPI has suddenly dropped below an acceptable limit (perhaps due to a valve failing to close or an instrumentation failure), systems can be built to spark an immediate intervention and minimize the loss.


4. The Prediction Stage

The performance of a KPI in relation to variables is relatively easy to predict because its calculation is known. Understanding how one variable impacts another is much harder, as operational data is highly susceptible to noise and confounding factors. However, with enough, well-organized data, models can be built that will predict important process variables in relation to others. This is the point where a lot of people want to say the “m” word, and they are not unjustified in doing so. When faced with the right problems and given the right data, machine learning can be an invaluable approach. However, people often underestimate how much can be gathered from just looking at the data if it can be visually represented well (visual analytics). In any case, this is the stage at which predictions can be made about how operating slightly different might affect key process variables that drive KPI’s. Some organizations go so far as to achieve these predictions through the incorporation of incoming data with process simulations (a highly mature form of a “digital twin”).


5. The Actuation Stage

Unlike its preceding levels of digital maturity, this stage focuses on “doing” rather than “knowing”. It is able to optimize for certain KPI’s based on the prediction strategy laid out in the previous step and is able to implement the results of these optimizations itself. An excellent example of this is Honeywell’s Advanced Process Controller (APC). Typical process control makes it so that an operator doesn’t have to worry about how to open a series of valves should be to reach a target flow. Instead, an operator specifies a set point and the control system opens valves accordingly. But with APC, the operator doesn’t even need to specify a set point. Instead, APC combines the data it’s seeing with the predictive models it’s been trained with to change a collection of setpoints itself in order to optimize the KPI’s it’s been instructed to.


Needless to say, these steps can be quite an undertaking. Many organizations will focus on one very small business segment to build its digital maturity. For instance, a popular tactic used by many industrial companies is to focus on one type of asset (such as compressors). In any case, it is unlikely that a single organization will be at the same level of maturity in its digital transformation journey across the board. Or, a company might assemble an in-house team to develop the tools necessary. However, the skills needed for such an effort can be hard to find. Even then, those skills when found internally are often in high demand and unable to meet rising strategic imperatives.  Another strategy is to team up with experience along this journey already. There is a rich ecosystem of vendors and service providers that provide the tools and the expertise needed to guide you along. A service provider, in particular, can help you answer questions regarding how business problems can be solved along the way and which tools are right for your needs.