Introduction
The ability to sustain change in large organizations is more important than it has ever been. The increasing pace of technologies and their application in society increase the complexity of those changes, requiring us to search for new tools and strategies that can keep pace. Artificial intelligence, machine learning, and language modeling are trending capabilities that have promising effects on automation, decision-making, data aggregation, and trends analysis, to name a few examples. In this series of blog posts, we apply these technologies to the change management demand of quantifying change saturation in an enterprise organization. Going beyond the common theoretical discussion, we intend to show a specific application of machine learning and invite our community of readers to join us in this conversation.
The Potential of AI and Machine Learning in Change Management
Beyond just implementing new tools and processes into organizations, change management is also about understanding stakeholder groups that are affected to maximize their likelihood of adopting change. Machine learning and AI are powerful capabilities for aggregating stakeholder data and delivering insights for specific stakeholder groups while significantly reducing the manual labor that a practitioner would otherwise undertake for those insights. ChangeAnalytics has been aggressively developing these capabilities. Leveraging reputable datasets like the ONET job description database, we have developed advanced machine learning models and algorithms to associate occupational data with custom-created stakeholder groups.
What is Change Saturation?
Many companies talk to us about tackling the “change saturation problem”. We hear varying definitions of change saturation and have created a working definition ourselves.
Change Saturation is the optimal threshold or limit of simultaneous changes that can be successfully comprehended and adopted by stakeholders while maintaining performance and productivity in their core business function.
Ultimately, leaders want to foresee the quantity of changes that a group of people are going to experience vs. the available capacity that group has for change.
The Benefits of Change Saturation
Change Saturation is crucial in that it serves as a leading metric for large enterprises to innovate and adapt as efficiently as possible while maintaining productivity, minimizing resistance & risk, and ensuring sustainability for its people. There are several benefits of tracking change saturation:
- Minimizing Resistance: Large enterprise organizations often have numerous stakeholder groups with varying processes, cultures, and norms. Too many changes at one time can overwhelm the stakeholder and create resistance. Managing change saturation fosters a pace of change that stakeholders can more realistically adapt to.
- Maintaining Productivity: Employees may struggle to keep up with new ways and means when they are introduced too often. The affected teams may experience a decline in productivity and an increase in errors.
- Risk Mitigation: The risk of error generally increases alongside the quantity of changes implemented simultaneously in an organization. When errors do occur, it can prove difficult to locate the root cause of issues because of the saturation of new changes.
- Quality Assurance: Large enterprises often have strict quality control measures. Introducing change gradually ensures that proper testing and validation can occur, thereby reinforcing quality standards.
- Cultural Adaptation: Organizational culture can be disrupted when too many changes are introduced, leading to stakeholder resistance and even resentment.
- Resource Allocation: Resources are always finite, and too many changes at once can strain those resources, especially time and budget. A sustainable saturation of change should equivalate to a sustainable resource demand.
- Learning and Development: Most change requires a certain amount of training and development. L&D efforts can create structured and thoughtful approaches when change saturation is managed effectively.
- Feedback and Iteration: Incremental change fosters rich feedback loops in the organization. Gradual change can allow time to gather feedback and make iterative adjustments before the next phase of change.
- Change Adoption: Adoption is about more than just successful implementation; it's also about ensuring that a change sticks and is accepted as a new normal. Gradual implementation allows for reinforcement and habituation of new practices.
- Maintaining Customer Experience: For businesses that interact with customers, experiencing too many simultaneous changes can disrupt the customer experience. It's crucial to maintain a level of consistency while introducing improvements.
Quantifying Change Saturation
Ok, so we’ve unpacked the theory and value of change saturation, but how do we quantify and measure it? People generally understand that change saturation is when the demand for change has exceeded the capacity of change. We also believe that saturation can exist if you are near or at capacity for an extended period. Through years of experience in delivering OCM methodology and digitizing existing change methodologies, we find that there are two schools of thought in assessing change saturation. These aren’t mutually exclusive schools of thought; rather, we find them complimentary.
School of Thought 1: As a data aggregator, we provide you with organizational demand data. It's up to you, as the leaders and change managers, to assess whether this is too little or too much in terms of capacity. In other words, you know capacity and we provide a visual of your organizational demand. You then decide whether people groups have the necessary bandwidth.
There is a significant benefit in this school of thought: Just like employee satisfaction, if someone tells you that they are unsatisfied, it is irrelevant what your formula tells you. An individual’s direct feedback will trump the calculus of a satisfaction equation. This is effectively a change saturation assessment. Any formulaic method creates some level of generalization. With this approach, there is less generalization as we receive direct, reliable feedback. The leaders in your company are always going to know their team’s saturation levels better than any equation will…but scaling and standardizing this is difficult.
There are some limiting factors in this approach: Most of all, it is often impractical to routinely ask frontline leaders and employees what the capacity is for every single system:
- You cannot calculate capacity because all variables are not quantifiable. You must rely on individual feedback from employees and managers
- This is hard to scale for organizations that have tens of thousands of employees (surveys)
School of Thought 2: Capacity can be calculated using a mathematic formula to account for most variability. The system can give you both the quantifiable capacity and demand data, so that you can make an educated assessment. This highlights a need in our industry for measuring what capacity will be. This approach, too, isn’t without its challenges:
- This approach forces you to try to say, “What is the unit of measurement?” Naturally, we think of hours of time, but not everything lends itself nicely to hours of time. What about the loss in hours due to resistance, or a productivity dip? I can’t simply say that because we've planned 40 hours of activities on this high impact initiative, the impact is 40 hours.
- There remains a hidden constant: politics, issues with the system, the user’s struggle to perform their job in a new way, and the build in frustration natural resistance. How do you quantify saturation as best as you can while making some level of estimate based on a multiplier of the magnitude of impact to this workgroup? This is purely subjective.
Both schools of thought are sure to be something a leader will confront when measuring organizational saturation. And when we say they are complimentary, we mean that just because we provide you with a calculative method for saturation, that doesn’t replace the essential practice of check-ins with your stakeholder groups. A formula cannot replace EQ, and AI shall not replace the human experience. After all, our formula doesn’t account for what people have occurring personally or professionally. We are positioning our approach to tell the story of “Accountants (an occupation organized within an org structure) will have a couple of hours of availability for changes next week”, instead of “Steve (an individual) has a couple of hours of availability for changes next week”. At ChangeAnalytics, we have traditionally followed the first school of thought. ChangeAnalytics is a great platform for beginning to understand the overall demand for changes. Through unlimited usership, you allow all relevant teams, leaders, and practitioners accessing the tool to see what their demands will be across various time scales. But we want to move the needle and challenge ourselves, our clients, and the change industry to truly quantify capacity in a way that is standardized and communicated. The source data, the associations & connections, and the calculative capacity requirements are daunting. This is where the advent of machine learning allows us to realistically approach this problem, and we are excited to share a tangible application for utilizing machine learning models to answer the change saturation question.
Finding a Reliable People Capacity Database
Our first stop in building a robust change capacity formula is realizing that we need a database of capacity based on the context of an occupation. https://www.onetonline.org/ features a detailed database of occupations and their work context. We want to use the O*NET database to determine the hours of change capacity an employee has per week based on their job family and job title. Through the context of this database, change capacity means: The number of hours an employee has available per week to prepare for and adopt changes that an organization is implementing. From our experience, an individual’s work context is the largest contributing factor to forecast change capacity.
The O*NET work contexts that impact change capacity include:
- Time pressure: How often does this job require the worker to meet strict deadlines? Employees with jobs that constantly have them under tight deadlines or immediate pressures would have less capacity to deal with organizational change.
- Consequence of Error: How serious would the result usually be if the worker made a mistake that was not readily correctable? If an error in their daily tasks could lead to significant consequences, it can be assumed that these employees would be more resistant to change that might increase their risk of making errors. They would be more focused on minimizing errors in their regular work.
- Deal With External Customers: How important is it to work with external customers or the public in this job? If a significant portion of an employee's time is spent dealing with external customers, their capacity for internal organizational change might be limited. They would prioritize customer-related tasks over internal change processes.
- Impact of Decisions on Co-workers or Company Results: What results do your decisions usually have on other people or the image or reputation or financial resources of your employer? Employees whose decisions have a wide-reaching impact might be more cautious about changes, as they would need to consider how those changes affect their decision-making processes and the outcomes of those decisions.
- Importance of Being Exact of Accurate: How important is being very exact or highly accurate in performing this job? Like 'Consequence of Error', roles that require a high degree of accuracy might be more resistant to changes that could disrupt their routine or introduce new elements that they're not familiar with.
- Do they work indoors or outdoors? How frequently is this job exposed to outdoor conditions, or a consistent a comfortable work environment? Available change capacity can fluctuate as they are exposed to a more demanding physical environment.
According to ONET, these physical and social factors influence an individual’s nature of work:
- Interpersonal Relationships: This category describes the context of the job in terms of human interaction processes.
- Physical Work Conditions: This category describes the work context as it relates to the interactions between the worker and the physical job environment.
- Structural Job Characteristics: This category involves the relationships or interactions between the worker and the structural characteristics of the job.
The aggregation of these variables produces a “work context modifier”. The higher this modifier, the less hours someone has available for change initiatives. The work context modifier can be between 0-1, and the multiplier = 1-work context modifier.
For example, if your work context score is 0.5, that means you will have 2 hours of change capacity in a week instead of 4:
- ((4*(1-0.5)). => 4*0.5=2
Let’s continue in showing you how we arrived here.
Orders of Magnitude
After collecting our capacity variables, our next step was to assign an order of magnitude, or weights, to each of those variables according to how much they would affect an occupation’s change capacity. We are going to use a Customer Service Representative (CSR) as an example for the rest of this paper. A CSR’s capacity variable weights are:
- Consequence of Error (E) = 0.25
- Deal With External Customers (C) = 0.05
- Impact of Decisions on Co-workers or Company Results (D) = 0.25
- Importance of Being Exact or Accurate (A) = 0.15
- In an Enclosed Vehicle or Equipment ( V{enc)) = 0.1
- In an Open Vehicle or Equipment (V{open}) = 0.1
- Outdoors, Exposed to Weather (O{exp}) = 0.1
- Outdoors, Under Cover (O{cov}) = 0.1
- Time Pressure (T) = 0.2
Modifier
Given the O*NET score range, we can refine our modifier based on the scores provided for the CSR. First, let's normalize the O*NET scores to lie between 0 and 1, which will make our computations more intuitive. We can normalize by dividing each score by 5. Using the example scores for CSR:
- Consequence of Error (E) = 2.16/5 = 0.432
- Deal With External Customers (C) = 4.79/5 = 0.958
- Impact of Decisions on Co-workers or Company Results (D) = 3.86/5 = 0.772
- Importance of Being Exact or Accurate (A) = 4.53/5 = 0.906
- In an Enclosed Vehicle or Equipment (Venc) = 1.15/5 = 0.23
- In an Open Vehicle or Equipment (Vopen) = 1.12/5 = 0.224
- Outdoors, Exposed to Weather (Oexp) = 1.29/5 = 0.258
- Outdoors, Under Cover (Ocov) = 1.12/5 = 0.224
- Time Pressure (T) = 4.55/5 = 0.91
Let's compute (O) (representing the average of outdoor/equipment-related scores):
O will range from 0 (no impact) to 1 (maximum amount of impact)
O = (Venc) + (Vopen} + (Oexp) + (Ocov) /4
O = (0.23) + (0.224) + (0.258) + (0.224) / 44
O = 0.234
Now, given the previous hypothetical weights (and adjusting them slightly):
- D = 0.25
- E = 0.25
- A = 0.15
- T = 0.2
- C = 0.05
- O = 0.1
Your modifier (M) becomes:
M = 0.25(0.772) + 0.25(0.432) + 0.15(0.906) + 0.2(0.91) + 0.05(0.958) + 0.1(0.234)
M = approx. 0.727
Solving for Capacity
Weekly Job Role Capacity = (estimated hours per week) x (1 - [work context score])
Our adjusted change capacity (CC) for a Customer Service Representative is:
CC = 4(1 - 0.727) => CC = approx. 1.092
Given this calculation, a Customer Service Representative would have just over one hour (1.092 hours) of change capacity per week.
This method allows for a dynamic evaluation of change capacity based on a role's O*NET scores. This formula adjusts the change capacity down based on the impact of the work contexts. A higher modifier (M) reduces available capacity of an occupation due to multiplication with the factor (1-M). Adjusting weights or adding new contexts would be the primary method to refine the model further. Always remember to validate with real-world feedback to ensure the approach is robust and meaningful.
Caveats and Considerations
Our calculation assumes all employees are available up to 4 hours every week outside of their day job to participate in change work. We take our work context algorithm into account to reduce this to something more tailored for this individual’s job. For example, unlike a customer service agent, an airline pilot is not available for 4 hours per week. The more your variables increase, the less hours of change capacity an individual has per week.
Because of O*NET’s extensive database of job occupations, one can match almost every job title to the occupation database to retrieve the work context score. Work context can be searched for using a database like O*NET. And while standard occupations have an associated O*NET mapping work context, we realize there are differences between organizations, teams, and individuals that will create a numeric variance in available starting capacity, variable weighting, and in the overall calculation itself. So, while O*NET provides a great industry average rooted in research and expert feedback, organizations will want to tailor their variables for roles specifically in their organizations.
Our equation also doesn’t account for individuals’ actual capacity. This forecasting is generally based on a group of people, but it doesn’t translate to the statement that “Bob Doe has 2 hours of capacity next week”. For example, Bob could be on PTO, and Jane on maternity leave. This is a general forecast for a group of people based on their role, but individuals have many actual realities that will change this narrative.
Remember, while data from O*NET provides an excellent starting point, change management is multifaceted, and human elements can often introduce unpredictability. Your model might give a general idea, but on-the-ground observations and continuous feedback loops are crucial for a holistic understanding.
Where Do We Go from Here?
The ability to forecast saturation allows one to avoid change saturation in the future. Right now, saturation is looked at to answer the question of, “is there too much happening”. But we want to answer the question of “how can we MINIMIZE or AVOID saturation from occurring say, 2 quarters from now?” The spirit behind our effort is to upgrade the concept of saturation management from ‘firefighting’, to ‘fire prevention’.
This forces saturation to be spoken about at a job role level. Instead of saying, “What business units in my organization are saturated”? we challenge you to say, “I want to see the jobs that are saturated”. You could aggregate these and say, “If more than 10% of jobs in a unit are saturated, this organization is saturated”, but we take the stand that saturation is determined by job roles.
Closing Remarks
At ChangeAnalytics, we work with organizations that have upwards of 5,000 job titles. It would be daunting to manually map this many job titles,incorporate work context, and assign scoring to each title.Enter machine learning. In our next blog on this topic, we will describe how we are fine-tuning pre-trained natural language processing models by leveraging tensor flow to perform text classification between job titles and O*NET occupations.
We Want to Hear Your Opinion! Please consider taking our survey HERE and tell us your thoughts on this approach to change saturation.
“Work Context.” O*NET OnLine, National Center for O*NET Development, www.onetonline.org/find/descriptor/result/4.C.2.a.1.c. Accessed 1 Nov. 2023.