LinkedIn Post Ideas for Heads of People

We ran exit interviews for years. I could have written most of the answers before the meeting: "new opportunity," "better fit," polite, scripted, useless for pattern-finding. Eighteen months ago we added a stay interview program instead — structured conversations with employees we specifically wanted to keep, run quarterly, independent of any performance cycle. The difference in what surfaced was immediate. Exit interviews are conducted by someone who has nothing left to lose being honest and often chooses not to be, because burning a bridge on the way out serves no one. Stay interviews are conducted with people who have every reason to want the friction fixed, because they're staying and living with it. What we learned that exits never told us: a specific manager's 1:1 cadence had quietly dropped to monthly across an entire department, and three of our strongest performers named it, unprompted, as the top thing they'd change. We fixed that manager's cadence within a month. Two of those three were flight risks we didn't know we had. The business case I took to our CEO: stay interviews cost roughly 40 hours of HR business partner time a quarter across the company. Retention of even two senior engineers we'd have otherwise lost pays for that many times over in replacement and ramp cost alone. Exit interviews measure why people left after it's too late to change anything. Stay interviews measure what's fixable while it still is.

We rolled out unlimited PTO three years ago, expecting it to be a clear win. Average vacation days taken actually dropped, from 16 days a year under our old accrual policy to 11 under unlimited. The mechanism, once we dug into exit and stay interview data: unlimited PTO removed the social permission structure that accrued days provided. When you have 15 banked days, taking a week off feels earned and visible — it's yours, use it. When PTO is "unlimited," nobody wants to be the person who takes the most, and there's no number signaling what's normal, so people default to less out of quiet anxiety about how it looks. We changed course a year ago: a minimum-PTO policy, 15 days mandatory, tracked, with a manager escalation if someone's on pace to take fewer than 10 by Q3. Not a maximum — a floor. Average days taken under the new policy: 17.5, higher than either previous system. Manager check-ins on the floor caught two cases of burnout-driving under-use in the first quarter alone that would have gone unnoticed under "unlimited." The lesson for other people leaders: unlimited-anything sounds generous and often functions as a guilt mechanism instead, because ambiguity doesn't remove pressure, it just removes the permission structure that used to counteract it. A floor works better than a ceiling that doesn't exist.

Our comp review cycle used to take six weeks and generate more manager anxiety than almost any other HR process we ran. Here's the mechanics that fixed it. Calibration sessions: every manager comes to a cross-functional calibration meeting with their proposed changes ranked, not just listed. Rankings force tradeoffs before the meeting instead of everyone defending their whole team as universally deserving, which is what happened when we skipped this step. Band transparency level: we publish full salary bands internally, visible to every employee, updated annually against market data. This was the single most contentious decision I've made in this role — three senior leaders predicted a wave of comp complaints. What actually happened: complaints dropped, because ambiguity, not the actual numbers, was driving most of the anxiety. Manager talking points: every manager gets a scripted framework for the comp conversation itself, not a script to read verbatim, but required talking points — where the employee sits in-band, what would move them up a band, timeline for the next review. Managers used to improvise these conversations, and inconsistency across managers was our biggest source of perceived unfairness. Result: cycle time down from six weeks to 12 days. Post-cycle survey scores on "I understand how my pay is determined" went from 54% agreement to 88%. Comp anxiety is mostly an information asymmetry problem. Fix the information, and most of the chaos goes with it.

91% response rate on our last engagement survey. I want to be specific about what actually drove that, because most surveys I see hover around 50-60% and everyone blames survey fatigue as if it's unfixable. What drove it: visible action on the previous survey's results, publicly tracked. We published, company-wide, the five action items from the last survey and their status — done, in progress, or explicitly declined with a reason. Employees had watched us follow through before being asked to fill out the next one. Survey brevity: 12 questions, six minutes, down from 40 questions and 20 minutes two cycles ago. We cut everything that wasn't driving a decision we'd actually act on. Leadership accountability for follow-through: every department head has one action item from the survey tied to their own quarterly goals, reviewed by me directly, not delegated to an HR business partner to chase down informally. The trick, if there is one: response rate isn't a survey design problem, it's a trust problem, and trust is built by what happened after the last survey, not by how the next one is worded. Teams that had seen zero follow-through from a previous cycle responded at rates in the 40s. Teams that had seen real action responded above 95%. If your response rate is low, look at what you did with the last set of answers before you touch the questions.

I handled a termination badly four years ago, and I still think about it more than almost any other decision in this role. What happened: I let the exit drag two weeks past the decision date because I was uncomfortable with the conversation, hoping performance would suddenly turn around and make the decision unnecessary. It didn't. Those two weeks were worse for the employee, not better — they could sense something was wrong and weren't given the dignity of a direct conversation, which is its own kind of cruelty dressed up as kindness. When the termination finally happened, I rushed the actual meeting to get through my own discomfort, which meant the employee left without clear information about severance timing or reference expectations, and had to email me twice afterward for basics I should have covered in the room. What I changed: I now hold myself to a rule — once a termination decision is final, the conversation happens within 48 hours, not when I feel ready. I built a mandatory checklist for every termination conversation covering the practical details, so discomfort can't cause me to skip them again. And I no longer allow myself to delay a clear decision hoping the problem resolves itself; that delay always costs the employee more than it costs the company. People leaders rarely discuss the terminations we got wrong. I think that silence is the actual problem — it means nobody learns from a mistake type that every one of us eventually makes.

The 30 days before a layoff announcement, from the inside, because employees only ever see the announcement itself and assume it happened overnight. Days 30-25: selection criteria debates. Not "who do we like" — a documented framework, reviewed by legal, scored against role criticality and performance data, specifically to avoid criteria that could look discriminatory in aggregate even if no individual decision was. Days 24-18: legal review of every individual selection, plus modeling the demographic composition of the affected group against the company overall, to catch unintended patterns before they became real ones. Days 17-10: manager scripting. Every manager delivering news got a written script and a live practice session, because an unscripted layoff conversation is where most of the dignity failures happen, not in the decision itself. Days 9-4: the support package fights. I pushed for four months of severance against an initial proposal of six weeks, and for extended healthcare coverage. I won three of four asks; the fourth, expanded outplacement services, I'm still not satisfied we got right. Day 0: the announcement, same day for everyone affected, no one finding out from a calendar invite before their manager could reach them. The invisible 30 days are where a layoff is either handled with real care or isn't. Employees only ever see day zero. The other 29 days are where a people leader's actual judgment shows.

I sat through a leveling framework presentation eighteen months ago, proposed by a well-meaning HR hire from a FAANG company, designed for an org of thousands. We had 80 employees. The framework had nine levels for individual contributors alone. It required a formal calibration committee, a documented promotion packet process, and a quarterly leveling review board. All of it made sense at a company of 100,000. At 80 people, it would have meant more process overhead than actual promotions in a given year. I killed it and built something right-sized: four levels, not nine. Promotion decisions made by a manager plus one skip-level, not a committee. A one-page promotion case, not a packet. Review twice a year, not quarterly. Same pattern with performance reviews — we'd been considering a FAANG-style stack ranking system. We built a simpler two-tier system instead: meeting expectations or needs a specific development plan, reviewed with the manager twice a year, no forced distribution. The instinct to copy practices from famous companies is understandable — those practices are well-documented and feel like a shortcut past building your own. But they're built for a specific scale and a specific set of constraints, usually thousands of managers who've never met each other, that most companies don't share. What FAANG practice have you seen crush a smaller org that adopted it wholesale?

Our AI resume screening tool ranked a candidate in the bottom quartile for a senior ops role last year. I overrode it on a hunch from the phone screen. She's been our strongest hire in that function in three years. What the algorithm was optimizing for: keyword match against the job description and a credential pattern learned from our historical top performers, most of whom had a specific degree pedigree and a specific title trajectory. She had neither — a nontraditional path, a title that undersold her actual scope at a smaller company, no elite school on the resume. What the algorithm couldn't see: in the phone screen, she described a process failure she'd caught at her previous company that had been quietly costing them money for a year, and the exact steps she took to fix it without authority to mandate the change — pure influence and initiative. None of that is visible in a resume's keyword structure. What we changed: any candidate the AI tool ranks in the bottom half now gets one guaranteed human phone screen before rejection, specifically to catch pattern-matching failures like hers. We're not removing the tool — it's still useful for triaging obvious mismatches at volume. But we stopped letting it make the final call on borderline candidates. The algorithm is only as good as the historical pattern it learned from, and if your historical top performers all look similar, it will keep selecting for similarity, not for the thing that actually made them good.

Four questions from our hiring and promotion data that predict manager quality better than anything in a standard interview loop. 1. "Describe a direct report you had to let go, in detail." Weak managers give a vague, victim-framed answer. Strong ones can walk through the performance timeline, what support they tried first, and what they'd do differently — accountability without cruelty toward the person. 2. "How do you talk about a report's contribution when they're not in the room?" We've started asking former reports this directly as a reference question. Managers who consistently share credit outward show up in retention data with meaningfully lower attrition on their teams, every time we've checked. 3. "Tell me about the last time you were wrong about someone's performance, in either direction." Managers who can't name one either haven't managed long enough or aren't being honest, and both are predictive of trouble later. 4. "How do you decide who's ready for more scope?" Vague answers ("I just know") correlate with promotion decisions that look like favoritism in the data later. Managers with an actual framework produce teams with more even promotion distribution. We started scoring these four questions explicitly in our manager hiring loop two years ago. Managers who scored well have, on average, 30% lower attrition on their teams than managers who scored poorly on this specific set, holding department and level constant.

What policy did you delete this year, not add? Mine: I killed our travel pre-approval chain, which required three signatures for any trip over $500. It existed to prevent a spending problem we'd had six years earlier, with people no longer at the company, in a completely different budget environment. It had outlived its reason by years and was still costing every employee real time and irritation on every trip. I also deleted our formal dress code for client-facing roles, replaced with one sentence: dress appropriately for who you're meeting. It hadn't been enforced in two years anyway; it existed only as a policy document nobody referenced until someone wanted to weaponize it in an unrelated dispute. What deleting both taught me: policies accumulate because adding one feels like progress and removing one feels like a risk, so the ratchet only goes one direction by default. I now review every standing policy over two years old annually and ask a specific question: if we didn't have this policy today, would we write it from scratch? If the honest answer is no, it gets deleted, not grandfathered. New policy announcements get attention because they signal the company is doing something. Deletion stories are rarer because they require admitting the old thing shouldn't have existed as long as it did. What did you delete?