You Only Keep What You Measure
A company rolls out a new quoting process. 3 months in, the report looks great. Adoption is up, the team completed the training, the tool shows activity. Everyone agrees it went well. Then someone actually pulls a sample of recent quotes and finds that half of them went out the old way, off a spreadsheet on somebody's desktop, because the old way was faster and no one was watching for it. The change was declared a success by a measurement that never checked the one thing that mattered.
This is the part of implementation that gets skipped, and it gets skipped because it is the least satisfying part. The decision felt good. The rollout felt like progress. Measuring whether the change actually took hold, months later, when everyone has moved on to the next thing, has none of that energy. So it does not happen, and the business slowly slides back to where it started while the paperwork says otherwise.
The failure rate is real, even if the famous number is made up
You have probably heard that 70 percent of change programs fail. It gets attributed to McKinsey and repeated everywhere. What is less well known is that even people who study this admit the exact figure is closer to folklore than research. Nobody can quite source it. What holds up is the direction. Across the serious estimates from firms like Bain, BCG, and others, somewhere between half and two thirds of change efforts produce disappointing results by most reasonable definitions of success.
And a dataset of more than 20,000 real strategic plans, tracked from 2017 through 2024, found that roughly 84 percent of strategic projects never reached completion at all. Not failed on the merits. Never finished. They were started, worked on, and then absorbed back into the daily grind before they crossed the line. The single most common outcome for a business initiative is that it fades before it is done, and the business rarely notices the moment it happened.
You do not close that gap with more willpower. You close it by deciding, before you start, how you will know the change actually stuck, and then by watching that number long after the excitement is gone.
Measure the behavior you are trying to replace, not the activity around it
Here is the distinction that separates a change that holds from one that quietly reverses. Activity metrics tell you people are doing things near the change. Trainings completed. Logins. Meetings held. Those numbers go up early and they feel like proof, and they are almost worthless, because a busy team can generate all of them while the actual behavior underneath never moves.
The number that matters is the one that tracks the old way going away. If the point was to move quoting onto the new system, do not measure how many people were trained. Measure what percentage of quotes last week ran through it, and go hunting for the ones that did not. If the point was faster approvals, do not measure how many people know the new policy. Measure how long approvals actually took, this month, in real days. The right metric is usually the one that would embarrass you if the change had not really happened, which is exactly why it is the one people avoid picking.
Set it before you begin, not after. A metric chosen at the start is a target you committed to in daylight. A metric chosen at the end is a story you tell about whatever the data happened to show. Owners who wait until the review to decide what success looks like will always find a number that flatters the effort, because there is always some number that went up.
Then keep looking well past the point it stopped being interesting. The dangerous window for any change is not the launch. It is month four, month five, when attention has moved on and the old behavior comes creeping back through the side door, one busy afternoon at a time. If nobody is still measuring by then, the reversal is silent and total, and the next time you propose something your people will remember, correctly, that the last change did not really change anything. A decision only becomes a change the business keeps when someone is still counting after everyone else has stopped caring.