Predictive and Prescriptive training in achieving business operational excellence

A case for applying Artificial Intelligence in training

Photo by Andrea Piacquadio from Pexels

Preface

This piece of work is part of the requirement for the University of Toronto course 2943–023 Maximizing value from Predictive analytics. The objective of this paper is to present a case for applying data analytics in eLearning for Predictive training. This paper discusses the training industry’s current state, gaps, and suggests a high-level solution to address the business need. Outlining the actual implementation is not the scope of this paper, the intent is to design an architecture on which a solution and/or service could be developed to help businesses achieve greater levels of success with their training initiative.

Overview

Training is the key to any business success — a better-trained workforce produces better results and has a greater impact on a business’s bottom line. In the service industry, continuous training is required to maintain service quality, whereas, in the production industry, it is pivotal to maintain the quality of the product and reduce the wastage.

Businesses view training as an investment and are keen in maximizing the ROI. They want highly effective training which is optimized to prepare a new member in the shortest possible time to be effective.

Training typically comprises of;

  • Mandatory Courses/subjects for compliance;
  • Operation courses to prepare team member to perform and execute the everyday tasks to achieve business objectives;
  • Business orientation and policies

Standard industry practice is to develop a library of courseware and deliver using offline and online platforms as part of a new team member onboarding process.

Gaps and challenges

Although the training is targeted meaning, a certain set of training modules are predefined for specific operational roles in the organization. For example, a team member hired to work as a cashier doesn’t need to know how the logistics work and vice versa. In this system, there is no flexibility of adjusting the learning path in real-time, every learner has to go over the same material regardless of their existing knowledge of the subject.

The other problem with this training model is that the system treats all learners as if they have the same level of knowledge and baseline education. Let’s take an example where a business has a course library which is 3 hours in length and the courses comprise daily operation, personal hygiene, and customer interactions. Every team member in a similar role has to go through the same 3 hours of training; regardless of their skill level in each area of knowledge. Out of three hours of training, one hour that talks about how to interact with customers are considered wasted time by the member who has an amazing customer relationship background.

The other challenge that business faces every day is how to implement a continuous improvement process! The traditional training platforms track team member training progress but once they are in the field there is no mechanism to trigger remedial training to fix the knowledge gaps that are identified by the team member progress.

Analytics opportunity and how predictive analytics can fill that gap

With the digital revolution in the past decade or so the Traditional Learning has evolved from classroom to online eLearning to blended medium where instructors teach students in both one-to-one and digital settings.

Modern e-Learning systems provide descriptive metrics which is a snapshot of what has already happened and fell short in understanding how descriptive and predictive metrics differ. Training administrators are responsible for setting goals and assigning training to each goal, they are not able to access the needs of each individual’s learning needs at depth. Here analytical models can help prescribe training based on learner’s attributes which are not possible with the traditional administrative system.

Organizations continuously collecting team member’s data and the data team members are generating as their digital footprint through social media is exponential. There are contents that are developed by the organization and there is content that is freely available and could be used to better train employees.

The great example is YouTube that is free and massively available — that does bring a challenge to professional content producers and we can also argue about the quality but still, there are tonnes of free resources available. According to some stats 300 hours of video are uploaded to YouTube every minute! YouTube may not have the right governance that one could have in a traditional learning environment but the amount of content that is available is incredible.

The velocity in which the data is being generated online is there for organizations to use and to better understand their workforce and to use it to their advantage to stay competitive. For some businesses, they have seconds to respond to stay competitive and need to make real-time decisions (up-sale, customer interaction, etc).

Of course, it is not easy to process data that is available in many forms — structured, unstructured, text, media, video, etc.

With predictive analytics methodologies, businesses can implement:

  • Recommended training — training module A has benefited a group and could also benefit other groups of learners with similar characteristics.
  • Predict effective time of training — not two individuals are the same. Some learners are more receptive to new knowledge in the morning and some later in the evening. A model that tracks the time when a learner was trained on a specific subject and compares the performance.
  • Suggest learners what it takes to win — how to sell the business/personal objective to the learner and what it takes to be successful — historical data with analytics could predict how high the chance of retaining the piece of information. Suggest correlation- if you do this there is a high chance of achieving that!

The primary goal of predictive analytics in training for a business is to achieve better performance, smarter decisions, and actionable insights that help drive business value.

Solution

Overview

Businesses’ objective is to have a well-trained work-force, who can deliver optimal operational excellence with the shortest possible onboarding process. An intelligent continuous performance improvement mechanism in-place that can react to issues that may arise due to the team member’s lack of knowledge. An adaptive workflow that is not only capable of identifying the root cause of the issue but also able to recommend remedial actions by providing targeted (predictive/prescriptive) training without impacting the on-going operation of the business.

Business Expectations from team members

  • Quality of the product/service
  • Delivery schedules (meeting deadlines)
  • Team Interactions (people issues)

How to achieve that — Predictive Training Curriculum using Team member’s dynamic profile and business objectives he/she is responsible for! (Fig.1)

Fig.1 — Workflow describing how a predictive training curriculum would be developed.

Analytic model

Predictive analytics data model builds on these attributes

  • Dynamic profile — that evolves with the learner
  • Past knowledge
  • Present activity in a current relationship with the organization
  • On-going social media activities
  • Business goals –
  • Assigned goals and KPIs
  • Training attributes
  • Subject relevance in the training material
  • Performance issues and their relationship with training material
  • Learner’s behavioral pattern
  • How a learner is interacting with the training — too quick to respond without reading the instructions (trying to guess) too slow to respond taking way too long to understand the problem.
  • Time of training — to better suggest dayparts when training is better retained by the learner.

Model Output

  • Generate knowledge gaps and predict completely dynamic curriculum
  • On-gong real-time Prescriptive training to rectify knowledge gap issues

Benefits and high-level implementation scenarios

Data collected during the hiring process comes from different sources — personality surveys, resume, data recorded during the interview process, and gathered from team member’s online presence.

A comprehensive training library has already been developed that comprises of custom training modules owned by the business; together with other online resources that supplement the training of the team members

Build an initial profile for training to be included for the onboarding process. Following profile, the machine will generate a training curriculum and a personalized learning path based on the objectives and learners’ current capabilities.

Update the learner’s knowledge profile for prescriptive training for the continuous improvement process (Fig.2). To achieve that objective a system has to be in a place that can provide real-time feedback to trigger continuous improvement. A system that tracks each individual performance in real-time and feeds data into the system. The model that is reading the data is continuously monitoring for any red-flags — the moment the system receives an issue identified as team member error, system updates the team members profile, model picks up the change, and immediately prescribes remedial training and finally, notify the learner and his supervisor.

Fig.2 — Workflow describing how the feedback helps predict remedial training to fill the knowledge gap.

An additional service can be developed that can actively monitor business’s other social platforms such as twitter/YouTube/Facebook etc. for team members updates and continuously run text/sentiment analysis. The results of that analysis will then feed into this workflow and then model predict the impact and corrective actions.

Learners A, B & C → trained using predicted curriculum -> Learner A perform well / Learner B performance average / Learner C performance not acceptable

Analyze data why Learner A performs better than the other two learners — the issue could be related to the content itself! (fig.2)

  • Quality of the content
  • Training format (visual versus text)
  • Training content (language) — Is this due to the fact that the training content was presented in the language which the learner is not too comfortable with.
  • Delivery of the content
  • What is learner’s prior/current understanding of the subject being trained
  • Are there certain times of the day where the learner is more receptive to new knowledge? Analyze the relationship between the time of training retained better than the other. Schedule training at predicted times.

Summary

Analytics are only valuable if they facilitate meaningful change and improvement. Corporates see training as an investment and with any investment, they are keen on maximizing the ROI.

Applying a predictive training model reduces the training duration by providing a targeted curriculum. Collecting feedback as learners go through training their reactions. Assessing what knowledge was gained and retained — behavioral changes and the impact on business results.

A real-time monitoring system that provides remedial solutions would enable businesses to react to the problem and help solve it quickly.