I sat in a room once while a senior leader read a diversity pledge off a slide. Forty of us clapped. There was a stock photo of smiling people, none of whom worked in engineering. Then we went back to our desks and the next promotion round looked exactly like the last one.
Ruth Wooderson says diversity programs are failing because they are theater. She is right. The research has been saying the same thing for ten years, and most of us kept buying tickets to the show anyway.
The training does not work, and we have known for a decade
Frank Dobbin and Alexandra Kalev spent years studying three decades of data from hundreds of US firms. Their write-up in Harvard Business Review is blunt.
Five years after companies made diversity training mandatory for managers, they saw no improvement in the proportion of white women, Black men, or Hispanic employees in management. The share of Black women in management dropped by 9 percent. Asian American men and women shrank by 4 to 5 percent.
Read those numbers again. Not "less effective than hoped". Worse than doing nothing.
Nearly all Fortune 500 firms run this training. Nearly half of midsize companies do too. It is the most widely deployed intervention in the field, and the measured outcome is somewhere between zero and negative.
If a deploy pipeline produced results like this, you would rip it out inside a sprint. Nobody rips out the training. We book another session.
Theater is comfortable because theater is easy to count
A workshop has an attendance sheet. An employee resource group has a budget line and a lunch. A pledge has a press release and a nice photo for the annual report.
All three produce artifacts, and artifacts feel like progress. You put them on a slide. Someone senior nods. The item goes green on the dashboard.
Real outcomes are harder to look at.
Who got the interesting work last quarter? Who got handed the migration nobody wanted? Who left inside eighteen months, and did anyone ask them why before they went? Whose pull requests attract comments about tone rather than logic?
Those questions have no attendance sheet. They have answers sitting in your HR system and your git history right now, and almost nobody pulls them.
What the numbers in tech look like
Women hold roughly 26.7 percent of tech jobs and 25.1 percent of US software engineering roles. About one in five senior engineering positions. Around 10.9 percent of senior leadership. Meanwhile 11.3 percent of women in tech and STEM left the field last year, and 66 percent report no clear path forward. Those figures come from a 2025 compilation of Bureau of Labor Statistics, Statista and McKinsey data.
Twenty years of programs. Two and a half people in ten.
I'm not sure about this part: aggregator sites pull from mixed years and mixed definitions of "tech role". Treat the exact decimals as approximate. The shape of the picture holds up across every source I checked.
Then the programs got cancelled anyway
Between January and February 2025, Meta, Amazon, Google and Accenture all scaled back or scrapped their diversity work. Accenture dropped its global diversity goals outright. Google ended diversity-based hiring targets.
Here is the part worth sitting with. If those programs had been load-bearing, removing them would have broken something visible. Instead they came out with a memo and a quiet week on LinkedIn.
You do not delete the payments service in an afternoon. You delete the thing nobody depended on.
You would never run a system you refuse to instrument
Every engineering leader I know lives on dashboards. Error budgets. p99 latency. Deploy frequency. Change failure rate. We argue about the definition of an incident. We write postmortems and we mean them.
Then we run promotion, project allocation and retention entirely on vibes and hope the pledge covers it.
Five things worth instrumenting, starting this quarter:
- Promotion rate by group, per level. Especially the jump from senior engineer to staff, where the pipeline usually breaks.
- Attrition inside eighteen months, split by manager. If three people quit the same team, the team is the signal.
- Distribution of high-visibility work. Who gets the greenfield service and who gets the legacy maintenance rota.
- Code review tone. Pull a month of comments and read who gets nitpicked on style versus substance.
- Starting salary versus salary after two years. Gaps at hire compound. Gaps after two years are yours.
None of these need a consultant. They need a SQL query and the nerve to read the output.
Five things with evidence behind them
Dobbin and Kalev did not stop at the bad news. They measured what moves the numbers, across the same five-year windows:
- Mentoring programs: 9 to 24 percent increases in representation of women and minority groups in management.
- Diversity task forces: 9 to 30 percent increases, when the task force has real members with real jobs, not a comms function.
- Targeted college recruitment: roughly 10 percent more women in management, 8 to 9 percent more Black managers.
- Someone whose actual job this is: 7 to 18 percent increases across all underrepresented groups.
- Voluntary training: 9 to 13 percent increases, versus the negative effect of the mandatory version.
Look at the pattern. Everything on this list either puts managers in sustained contact with people they would otherwise overlook, or makes a named human accountable for a number. Nothing on this list is a lecture.
The mandatory workshop tells managers they are the problem, then sends them back to the same broken process. The mentoring program hands them a person and a year.
The manager is the entire program
My own research found 99.5 percent of survey respondents said they have had one or more types of bad boss. Not a rounding error. Near enough everyone.
Every diversity outcome you care about gets decided by a manager, one small decision at a time. Who to put on the interesting project. Whose interruption in a design review goes unchallenged. Whose promotion case gets written with effort and whose gets written the night before.
A pledge on a slide never reaches those moments. The manager's behaviour is the program. Everything else is set dressing. I write about this constantly over at Step It Up HR, because the bad boss problem and the diversity problem are the same problem wearing different clothes.
What to do on Monday
Stop defending the theater. Start with one thing:
Pull promotion and attrition numbers for the last two years, split by group and by manager. Do not present them. Do not build a deck. Read them yourself, alone, and see whether you flinch.
Then pick one manager whose numbers look wrong and go find out why. Not to punish them. To learn what your system rewards.
Cancel one ritual with no evidence behind it and put the money into mentoring instead. Give one named person responsibility for one number, and put it in their review.
Your team will forgive you for having a problem. They will not forgive you for staging a play about not having one.
So here is the question I would ask yourself before the next all-hands. If someone deleted your entire diversity program tonight, what would break by Friday?