It moved from volunteer effort to funded line item because the lead stopped describing the program and started describing a cost. She built circles that ran on a fixed calendar, tracked who was in them, waited two cycles, and then asked HRIS one narrow question: of the women who completed a circle, how many moved up a level or laterally into a higher band within twelve months, compared with a matched group who did not participate.
The finance conversation changed when the ask was framed against replacement cost rather than against goodwill. The number the CFO cared about was not engagement. It was what it costs to backfill a senior individual contributor who leaves, and whether a few hundred dollars per participant moved the odds.
What follows is the structure of that program as it was actually run, including the year two failure, so another lead can copy the parts that worked.
The starting point: an unfunded volunteer mentoring group
The first version was one to one matching by spreadsheet. Someone posted a form, sixty women signed up as mentees, nineteen signed up as mentors, and the lead spent two weekends pairing them by hand. There was no end date. Pairs were told to meet monthly.
By month four, roughly half the pairs had stopped meeting and nobody could say which half. There was no budget, so there was no reporting obligation, so there was no data. That is the trap. Unfunded programs feel low risk because nothing is on the line, and they generate nothing you can take to a funding conversation later.
Two structural problems were doing the damage. Mentor supply was the binding constraint, and open ended commitments have no natural moment to check in or exit.
Keep reading: What should I check before booking an outside speaker for our women's history month event?
Designing circles of six on a fixed twelve week arc
The redesign fixed both problems at once. Group format solved the supply problem: one mentor to five mentees means nineteen mentors can serve ninety five people instead of nineteen. A fixed arc solved the drift problem, because a program with a stated end date has a completion rate, and a completion rate is a metric.
The shape:
- Six people per circle: one facilitator at director level or above, five participants.
- Twelve weeks, six sessions, every other week, ninety minutes, calendar held from day one.
- Two cycles a year, spring and fall, with the summer used for recruiting facilitators.
- A published session arc so nobody has to invent an agenda: career narrative, visibility and sponsorship, negotiating scope, navigating a difficult stakeholder, an outside guest session, and a closing session where each participant states a specific next move.
- A single ground rule stated at session one: what is said in the circle stays in the circle, and no performance information travels to anyone's manager.
Ninety minutes is not arbitrary. Sixty minutes with six people gives each person ten minutes and no discussion. Two hours does not survive contact with a calendar.
Matching criteria that avoided pairing people with their own manager
The rule that mattered most was negative: no participant is placed in a circle with anyone in her direct management chain, up or down. Nobody speaks candidly about scope, pay or a difficult stakeholder in front of the person who writes her review.
The matching logic ran in this order, and it is worth copying:
- Exclude any circle placement inside the participant's reporting chain, in either direction, at any distance.
- Exclude same immediate team, because the point is a view from outside the function.
- Balance each circle across at least three different departments.
- Keep each circle inside a two level band range, so a new analyst is not sitting with a senior principal and saying nothing.
- Honor time zone as a hard constraint before anything soft, because a circle spanning Pacific and Eastern will lose people by week six.
Chain exclusions cannot be done by eye. That check needs the reporting hierarchy, which is the first place HRIS becomes necessary rather than nice to have.
What was measured each cycle and what was deliberately not
Four things were collected, all of them cheap:
| Measure | How it was captured | Why it earned its place |
|---|---|---|
| Session attendance | Facilitator marks six sessions | Defines completion; four of six sessions counts as complete |
| Completion rate per circle | Derived from attendance | Identifies facilitators who need support, not blame |
| A three question pulse at week one and week twelve | Short survey, same wording both times | Movement on the same question is readable; a one time survey is not |
| Stated next move at the closing session | One line, written by the participant | Turns a feeling into something checkable six months later |
The pulse asked three things, unchanged between cycles: whether she could name a person who would advocate for her in a room she is not in, whether she understood what the next grade up required of her, and whether she intended to still be at the company in twelve months.
What was deliberately not measured: session content, individual survey answers shared with managers, mentor quality ratings, and anything resembling a performance signal. Every one of those would have broken the confidentiality rule, and the program's entire value rested on that rule holding.
Keep reading: Why does our ERG membership number look great while attendance keeps falling every quarter?
Working with HRIS to connect participation to internal mobility data
This is the step most programs skip, and it is the step that made funding possible.
The lead did not ask for HRIS access. She handed over a participant roster with employee IDs and cycle dates, and asked the people analytics team to run the join on their side and return aggregate results only. That framing matters: she was requesting a report, not personal data.
The question was narrow. For participants who completed a cycle, what share had a level change, a band change or an internal move within twelve months of completion, compared with a comparison group matched on department, level and tenure who did not participate?
Two conditions came back from analytics, and both were reasonable. Results would be suppressed for any cell below a minimum group size, to prevent re-identification in small departments. And the comparison would be described as an association, not a causal effect, because participants self selected. Accepting those conditions is what made the number credible rather than something the CFO could dismiss.
The funding ask: cost per participant against replacement cost
Here is the arithmetic she brought, with the assumptions stated as assumptions.
Assume 96 participants across two cycles. Assume direct costs of $4,800 for a facilitator development half day, $2,400 for an outside guest speaker across both cycles, $1,800 for materials and light catering at kickoff and close, and $1,000 for software and administration. That is $10,000, or roughly $104 per participant.
Assume the population is mid level professionals with a fully loaded cost of $140,000. Assume replacement cost is conservatively half of annual compensation once recruiting, ramp time and lost productivity are counted, so about $70,000 per departure. These are planning assumptions, not measured figures.
At $104 per participant, the program pays for itself if it prevents a single departure across 673 participants. With 96 participants a year, the program has to influence the retention of roughly one person every seven years to break even. That framing is what ended the debate. The precise retention effect stopped mattering, because the break even bar was so low that arguing about the size of the effect was not worth the meeting time.
She also brought the number the CFO had not asked for: the fully loaded value of facilitator time, sixteen directors giving nine hours each. That is the real cost of the program, and naming it before finance found it built credibility she used later.
See how ERGCircle handles this for employee resource group programs
What broke in year two and how the model was adjusted
Demand tripled after the first cycle results circulated. Facilitator supply did not. The lead accepted anyone willing, and three circles ended up with facilitators who had never run a group discussion. Two of those three had completion rates under half. Participants in them reported the sessions felt like an unstructured meeting.
Three fixes went in for cycle three:
- A required two hour facilitator preparation session, with the session arc walked through and two hard moments rehearsed: the participant who dominates, and the participant who goes silent after week two.
- A cap on circles per cycle set by facilitator supply, with a waitlist, publicly explained. Turning people away with a reason preserved more trust than admitting everyone and running weak circles.
- A mid cycle check at week six, one question to each facilitator: is this circle working. Two circles were merged that year on the strength of that check.
The waitlist itself became an argument. A visible queue of named employees asking for something is more persuasive in a budget review than any satisfaction score.
What another lead could copy directly
The transferable pieces, in order of how much they matter: the fixed twelve week arc with a published session plan, the reporting chain exclusion rule, the same three pulse questions asked at week one and week twelve, the aggregate only HRIS request, and the break even framing against replacement cost.
What is not transferable is the number. Your population, your attrition and your replacement cost are different. Run the same arithmetic with your own assumptions and show your work in the room.
Running it without the spreadsheet sprawl
Every piece of this depends on records kept consistently across cycles: rosters, attendance for six sessions per circle, two pulse waves, a budget line, and a report at year end that ties them together. The first version failed partly because that record did not exist.
ERGCircle holds those pieces in one place: event attendance by session, budget tracking against the line you were funded for, member pulse surveys you can run at the same two points every cycle, and an annual impact report you can hand to finance. Start the next cycle with the tracking already in place, and the funding conversation twelve months from now writes itself.
