Control-group ethics in coaching measurement.
What you can withhold, delay, or randomize without turning employees into an experiment. Waitlists and staggered pilots can be ethical. Denying coaching to a struggling manager to prove ROI is not.
Control-group ethics in coaching measurement is what you can withhold, delay, or randomize without turning employees into an experiment. Finance or People analytics wants a comparison group. Waitlists and staggered pilots can be ethical. Denying coaching to a struggling manager to prove ROI is not.
What to measure, and the behavior worksheet, already live on coaching program metrics. The finance-shaped catalog is measuring coaching ROI. This page stays on the experimental design. If you came for a metric list, leave.
A clean chart is not worth a manager you chose not to help.
When Finance wants a control group.
They want to know the change is not just time passing. That is a fair question. It is not a license to run a lab on people who report to you. Answer with a design that still gets everyone in the eligible population a seat, plus a date. Take that design to the board talent-committee pack if they asked for science. Do not promise a withheld group of people in trouble.
CFO-facing metric language, still not this ethics question, is coaching ROI metrics for a CFO. Category ROI essays stay on leadership development ROI.
What you can delay or randomize.
A dated waitlist inside one eligible population
Everyone who meets the criteria will get a seat. Start dates are staggered because you can only operate one wave. Randomize the order if analytics needs it. Publish the date.
A staggered site or function rollout
Site B was always going second because of a local calendar. You can look at Site A vs Site B for a quarter. You cannot keep Site B off coaching because they are "the control."
Opt-in delay for people not in distress
A director wave that is stretch, not rescue, can start in two cohorts. Tell them why. Do not call the second cohort a control in the invite.
Waitlists as access policy (not as science) are on nomination vs open enrollment. A first buy that wants a comparison should write the stagger into the pilot charter before anyone is left out by accident.
What you cannot withhold.
| Design | Why it fails ethically |
|---|---|
| Keep struggling managers off coaching so they stay "the control" | You already know they need help. The chart is not worth it |
| A permanent never-eligible group inside the same job | That is a caste, not a method |
| Hide the stagger so people do not "contaminate" the data | Employees are not subjects. They are colleagues |
| Use session content to assign people to groups | That is surveillance plus an experiment |
Distress breaks the design. If someone in a delayed wave starts to fail in role, you pull them forward. You do not protect the p-value.
Waitlists and staggered pilots.
Recommendation, not a methods paper we published. Write the eligible population. Split it into Wave A and Wave B by a rule you can say in public (random start date, site calendar, capacity). Give Wave B a date. Measure behavior on both, using the worksheet, not session content. When Wave B starts, you still have a before/after for them. You do not owe Finance a forever untreated group.
SCALE can run both waves as 1:1. See SCALE. The ethics do not change because a platform is fast. Speed is not consent.
This is not another ROI article.
We will not list leading indicators, vanity metrics, or a payback story here. Those pages exist. If a paragraph on this route starts to explain how to calculate ROI, delete it. The only number that matters on this page is zero: zero people left without help so a slide can look scientific.
Design the stagger without a human control.
Bring the Finance ask. We will write a dated waitlist or a site stagger, not a withheld group of struggling managers.
Book a strategy callFrequently asked questions
Is it ethical to use a control group in a coaching program?
Sometimes. A waitlist or a staggered pilot can be ethical when everyone in the eligible population will get access, the delay has a date, and you are not withholding help from someone already in trouble to prove a number. A permanent denied group of struggling managers is not a control. It is a choice to leave them without support.
What can you withhold or delay when measuring coaching?
You can delay a wave that is not in distress, randomize start dates inside a waitlist, or stagger sites that already have a rollout calendar. You cannot withhold coaching from a manager you already know is failing in order to keep your statistics clean.
How is this different from measuring coaching ROI?
ROI and program-metrics pages tell you what to measure. This page is the experimental-design ethics: who you may leave out, for how long, and why. If you want the metric catalog, leave this page.
Do you need a control group to know coaching worked?
No. Pre/post behavior on a named wave, observed by people who work with the manager, is enough for most operating decisions. A control group is a Finance or analytics request. Meet it with a waitlist design, not a human sacrifice.
Should employees know they are in a measurement design?
Yes, at the level of the design. "We are rolling this out in two waves. Wave two starts in Q3. We will look at both groups." They do not need a methods appendix. They do need to know they are not the withheld prize.