TL;DR
Short answer: Yes, both can be true in the same company at the same moment. Engagement scores are averages, and averages exist to smooth out exactly the pattern that matters most: whether the people driving your revenue are already gone in their heads.
Key stats you need to know:
- Most HR functions still operate primarily as reporting shops, with the largest share of organizations stuck at basic or operational reporting levels rather than the analysis that changes decisions (SHRM).
- Even among organizations that call themselves advanced in people analytics, more than 70% still lean primarily on static reporting instead of analysis that changes a decision (Deloitte).
- In the case I walk through below, top performers in one sales organization left at three times the company-wide attrition rate, clustered around the 18-month mark, invisible inside a company-wide engagement score of 90%.
The leadership takeaway: If your people data has never told the room something it didn't want to hear, that's a reporting function wearing an analytics badge.
Why Do High Engagement Scores Miss Real Attrition Risk?
High engagement scores miss real attrition risk because they average sentiment across your entire population, and averages are built to hide exactly the pattern a business needs to see. A company-wide 90% can absorb a lot of damage. It can absorb a sales team losing top performers at three times the company rate. It can absorb an engineering org where the architects who hold the system in their heads are one recruiter call away from gone.
I watched this play out directly. Survey results came back strong, leadership pointed to them at every board meeting, and underneath that average, the sales organization was bleeding its best revenue-driving talent every single quarter. Nobody had segmented the data by role criticality, so nobody had a reason to question the green number.
This tracks with what shows up industry-wide: HR functions have invested heavily in dashboards over the past decade, yet most are still primarily reporting shops rather than analysis shops, with the largest share of organizations still parked at basic or operational reporting rather than the advanced or predictive stages of the maturity curve (SHRM). Cornerstone has documented the same failure mode from the vendor side, calling it the trap where a strong headline number hides a retention crisis sitting one layer down (Cornerstone).
The dashboard answered exactly the question it was built to answer: how does everyone feel, on average, instead of are the people who drive our revenue about to leave. Get this wrong long enough and leadership starts asking why HR is in the room at all. This is the exact dynamic I covered when a CEO eliminated his entire HR function after deciding it wasn't earning its seat (my breakdown of the Bolt HR firing).
What Is the Dashboard Paradox, and Why Does It Matter?
The Dashboard Paradox is my term for the moment a company-wide metric looks healthy while a specific, high-value population inside it is already failing, and I use it to draw a hard line between reporting and analysis. Reporting describes activity: headcount, turnover rate, engagement score, time to fill. It tells you what happened, and it will never once change a decision on its own. Analysis tests a hypothesis that has a wrong answer built into it and then goes looking for evidence that could prove it false.
My test for which one you're running: has your people data ever challenged a decision leadership already wanted to make, with enough rigor that the room had to take it seriously? I've made a version of this argument before when I broke down how quickly the "AI is destroying jobs" narrative got walked back once anyone checked the data against it, including by the people who made the original claim (my breakdown of the walked-back AI job-loss narrative).
A convenient story survives right up until someone tests it. People data that has only ever confirmed what leadership already believed is corporate decoration wearing an analytics badge. In a boardroom, I'd put it this way: your dashboard is faithfully answering a question you never should have asked in the first place.
How Should CHROs Run a Causal-Inquiry Investigation?
CHROs should run a causal-inquiry investigation by segmenting the population before drawing any conclusion, then killing hypotheses instead of confirming them. Here's the four-step method from the case above.
- Segment before you conclude. Break the company-wide number down by function, role criticality, and tenure band before you say anything about it. That single cut is what turned a green dashboard red and surfaced top sales performers leaving at three times the company rate, clustered around 18 months.
- Form a hypothesis you can kill. The easy answer is almost always pay, so check it before you believe it. Comp data for the leavers came back competitive against market and against stayers. Hypothesis dead, and a dead hypothesis is progress because it stops you from spending money on the wrong fix.
- Go find the actual cause. Exit data, promotion records, and structured conversations with current top performers told one consistent story: high performers hit the ceiling of the promotion pathway within eighteen months, and the next step up either didn't exist or ran through a committee that met twice a year. So they left.
- Pilot before you roll out. Skip the company-wide "career pathing initiative" with a logo and a town hall. Fix the promotion pathway in one region first and measure it. Two quarters later, top-performer attrition in that region had dropped by half. Only then did it scale.
Notice what's absent from that sequence: another survey. You cannot survey your way to a causal answer. At some point you have to form a belief, test it, and be willing to watch it die.
What This Means for You as a Leader
Ask your own people data three questions this week. Where is attrition concentrated in your revenue-driving roles, not averaged across the company? Which leadership behaviors predict high team performance in your organization, not in a benchmark report someone else wrote? And what did you test this quarter, not just report?
Gallup's own research shows engagement is genuinely tied to turnover at the aggregate level, with high-engagement organizations seeing roughly half the turnover of low-engagement ones (Gallup). That correlation is real, and it's still the wrong altitude, because it tells you nothing about the specific team bleeding revenue right now.
If those three answers don't come quickly, the problem sits at altitude. You're reporting at ground level while the decisions that matter get made a level up, and I've made the case elsewhere that this is why so many CHROs feel locked out of the growth conversation despite doing competent work (my breakdown of why HR doesn't get a seat).
The fix comes down to a different mindset about what people data is for. Data that can't challenge a failing strategy leaves you decorating the walls of a room where decisions get made without you.
The Scientist is one of the five mindsets in my new book, The FutHRist: The 5 Mindsets of the Future HR Pro, out this fall. The full chapter walks through the causal-inquiry method step by step, including the exact case study above. If you want to know where you stand today, the free FutHRist self-assessment at trentcotton.com takes about ten minutes.
FAQ
Can engagement scores and high attrition both be true at the same company?
Yes. Engagement scores are population averages, so a strong company-wide number can sit directly on top of a specific, high-value group losing people fast. The average smooths the pattern out; it doesn't disprove it.
What's the actual difference between HR reporting and HR analysis?
Reporting describes what happened, using metrics like headcount, turnover, and engagement scores. Analysis starts from a hypothesis that could be wrong and goes looking for evidence to kill it. Most HR functions still operate primarily in the reporting mode (SHRM).
What is the Dashboard Paradox in simple terms?
It's the situation where a company-wide metric looks healthy while a specific, high-value population inside it is already failing, because the average absorbs the damage instead of revealing it.
How do I know if my HR function is doing real analysis instead of reporting?
Ask whether your people data has ever challenged a decision leadership already wanted to make, not just informed it. If your data has only ever confirmed existing beliefs, that's reporting wearing an analytics badge.
What's the first move to test the Dashboard Paradox on my own team?
Segment your current engagement or attrition number by function, role criticality, and tenure band before you say anything about it publicly. That single cut is usually what surfaces the real pattern.
Should I roll out a fix company-wide or pilot it first?
Pilot it first, in one region or team, and measure the result before scaling. Skipping straight to a company-wide initiative is how HR functions end up with expensive programs nobody can prove worked.
