Addressing Organizational Resistance to Analytics Projects & Transformation

Organizational Resistance is present, to a degree, in all analytics projects. It's not uncommon for resistance to choke a project and prevent it from ever being deployed. What are some proactive steps you can take to avoid this outcome?

Gauge the ‘analytic attitude’
Whenever my colleagues and I work with clients, allocate time to get the lay of the land before diving into the intricacies of any specific project. Sometimes there is so much enthusiasm around details and deadlines that no one asks these critical questions:
• Is analytics new to the organization?
• Has there been any initial successes or have efforts faced difficulty and frustration?
• Are analytics generally respected within the organization or do they raise fears and doubts?
• Is predictive analytics perceived as a potential profit center where projects return dividends - or is it perceived as a cost center?

Assemble the right team
How do potential end users of the resulting models feel about analytics and about the project? The biggest reason that organizational resistance occurs is that analytics is perceived as meeting an executive need with a purely technical solution. Both elements may be present, but the solution remains incomplete. In analytics, we are in the business of organizational change. We are introducing a new dynamic (a predictive model) into a system of humans and machines that are working together. We can’t forget the human element. We can’t wait until the end and simply pronounce that change is about to occur, then expect that everyone will willingly adjust. Those whose roles may change in the wake of model deployment must have a seat at the table from the earliest stages of the project.

Assess Potential Organizational Impact
After the model is built, focus should shift away from technical measures like R², area under the curve, lift, and other artificial metrics. Now is the time to determine what organizational impact the model will have when implemented and how the results will be translated for interpretation by leadership. If it is a predictive maintenance project the number of services performed each day might be anticipated to go down since your anticipating problems more accurately - but they could go up for the same reason. If the project is in healthcare, doctors and nurses may be concerned about a change in their day to day activities. So, assessing organizational impact is all about the logistical and operational impact and even perception of the impending model deployment.

Prepare Organization for Change
Once you’ve assessed the changes, you need to take action - proactively. You don’t want the end users of the model to have to request details. You will want to schedule briefings, prepare training, and almost certainly prepare a partial rollout. A full rollout, done too soon, is what often triggers a rejection of the model.

During the earlier phases you should be looking for allies within the organization and spotting where resistance may be the most intense. You will need this information to decide how long the partial rollout should be, who within the organization should participate and who will be waiting on the sidelines for tangible proof of success. This is at the heart of it.

If you assume that a model will be embraced merely because it is accurate you may be in for an unpleasant surprise. Evidence of effectiveness, on paper, will never be enough. You will need to use your powers of diplomacy and persuasion to successfully go live with your project’s model.


Keith McCormick will present two keynotes during the Data Warehousing & Business Intelligence Summit on March 27th and 28th 2019; Model Deployment for Production & Adoption – Why the Last Task Should be the First Discussed; and Addressing Organizational Resistance to Predictive Analytics and Machine Learning. Also afterwards he will present an unique post-conference workshop: Putting Machine Learning to Work.