Background
The common commercial availability of artificial intelligence (AI) and machine learning (ML) technologies has and will continue to have great impact on any field including commercial endeavors. Although the most popular form of AI is large language models (LLMs) such as ChatGPT or Grok AI, machine learning techniques will most likely have greater impacts. One very promising ML technique is predictive analytics.
Predictive analytics is the use of historical data combined with current information to forecast future conditions. There are several techniques for doing this, including the use of statistical models and regression analysis. Predictive analytics can be applied (Cote, 2021) to forecasting product demand, predicting equipment malfunctions, estimating cash flow, etc.
Potential Consequences of Implementation
For sake of discussion, we limit the examples to predicting equipment failure and forecasting cash flow.
To predict equipment failure, detailed maintenance logs must be maintained, data must be extracted, and from this, valuable information can be determined, most importantly date of last failure and mean times between failures. From this, a schedule for performing preventative maintenance can be established. Doing this ensures that equipment downtime can be minimized.
This is of course beneficial to any company using this equipment since downtime for maintenance can be scheduled for off-hours (when the machine in question is not heavily used) whereas downtime due to failure can happen at the most inconvenient time - and usually will.
There are several problems with using maintenance logs to predict equipment failure (Yadav et al, 2024). First, with companies using modern supply chain methodology, it is quite possible that the equipment in question is not owned but rather leased. Further, the company using the equipment, the equipment manufacturer, and the technicians servicing the equipment may be three separate businesses. These arrangements are possible third-party logistics arrangements (Quigg, 2022, p. 164-169). Who should collect and track maintenance records?
Also, what level of detail should the maintenance records be maintained: the equipment as a whole, subsystems, individual components?
If the company using the equipment has only one of this type of equipment, there will not be sufficient data to make accurate predictions. The equipment manufacturer may not track their equipment once it is delivered to the customer. Finally, there may be multiple service companies that perform maintenance. Again, who collects the maintenance records and performs the predictive analytics?
The best solution would be to have the equipment manufacturer maintain relationships with the service companies. This often happens, and when it does, the companies providing service are called "authorized" or "certified" service technicians. Thus the equipment manufacturer would collect the data and perform the predictive analytics. The downside is that the company using the equipment must rely on the manufacturer for the predictions - which leaves open the possibility that mean time between failures are artificially shortened thereby inflating maintenance costs.
Unlike the prediction of equipment failure, the prediction of cash flows is very straightforward: there are only two possible places where data can be collected. The responsibility of collecting records of revenues and expenses lies either with the company itself, or with an external accounting firm. With the external accountants, no additional trust needs to be earned in order to perform cash flow forecasts (Quigg, 2022, p. 301-302).
Proposed Response
When considering whether to adopt predictive analytics, or any other technology, a company must consider the profit and loss associated with applying that technology.
By using predictive analytics to forecast cash flow, there really are no costs since all tools used for accounting and finance are capable of some form of predictive analytics. For example, Excel has the "Data Analysis ToolPak" for performing linear regression. The benefits of predicting cash flows is that this prediction allows for early business actions - it is not necessary to allow business savings to accumulate before purchasing an item (Wach et al, 2021).
The decision whether to use predictive analytics for anticipating equipment failures is not so clear. As indicated above, the primary company may not have access to the historical data needed to make such forecasts. Servicing companies can collect the data, but may not have complete data if the equipment manufacturer retains multiple servicing companies. The only company for which it is even possible to predict maintenance requirements is the equipment manufacturer, assuming qualified service companies send reports to that manufacturer. For the manufacturer, it does indeed make sense since the maintenance records can be used to improve product quality and can be touted as a benefit to using their machines. This benefit then carries over to authorized service providers and the company that uses the equipment and service technicians since it allows them to better anticipate operating costs.
References
Cote, C. (2021). "What is predictive analytics? 5 examples." Harvard Business School Online. Retrieved 3 October 2024 from https://online.hbs.edu/blog/post/predictive-analytics
Quigg, B. (2022). Supply Chain Management (1st ed). McGraw-Hill Create. https://bookshelf.vitalsource.com/books/9781307866025
Wach, M. & Chomiak-Osra, I." (2021). "The application of predictive analysis in decision-making processes on the example of mining company’s investment projects." Procedia Computer Science 192. https://doi.org/10.1016/j.procs.2021.09.284
Yadav, D., Kaushik, A., Yadav, N. (2024). "Predicting machine failures using machine learning and deep learning algorithms." Sustainable Manufacturing and Service Economics 3. https://doi.org/10.1016/j.smse.2024.100029
