The Trackers Logged 185,894 Exits. Microsoft, ServiceNow, and Uber Added 12,000 More Off-Book.
Layoff trackers miss voluntary buyouts, no-backfill attrition, and "reorg" exits. Here's how to find the hidden talent pool they ignore.
If your sourcing pipeline starts with a layoff tracker, you're reading the wrong dataset. Skillsyncer's June 19 update logs 267 events and 185,894 workers cut in 2026, roughly 1,093 a day. TrueUp's count for the same period sits at 150,096. Challenger says 123,653. The three biggest trackers disagree with each other by 25 to 30 percent, and even the highest number is missing the three most interesting talent events of Q2.
The trackers are counting one thing, and only one thing
Skillsyncer, TrueUp, Layoffs.fyi, Crunchbase, and Challenger Gray & Christmas all index the same shape of event: an announced, involuntary, WARN-style reduction. Crunchbase says so in its own methodology disclosure: "Layoff and workforce figures are best estimates based on reporting. We source the layoffs from media reports, our own reporting, social media posts and layoffs.fyi." If it isn't reported, it isn't counted.
That definition assumes three trigger conditions. A WARN filing (which requires involuntary separations above a threshold). A press announcement (which requires the company to want the story out). An SEC 8-K (which requires a material event). If a company can move headcount without firing on any of those three triggers, the event is invisible to every tracker on the internet.
In Q2 2026, three companies did exactly that, and together they moved more senior engineers out of their orgs than the entire month of May's tracked total at Challenger.
Microsoft's Rule of 70: 8,750 senior exits, zero tracker rows
On April 23, Microsoft announced its first-ever Voluntary Retirement Program. CPO Amy Coleman's memo offered a buyout to U.S. employees at level 67 (senior director and below, but including the entire IC ladder up to Principal) whose age plus tenure summed to 70 or higher. About 8,750 people, 7 percent of the U.S. workforce, were eligible.
The package: eight to 39 weeks of base pay as a single lump sum, equity left intact, separation date July 2, 2026. Decisions closed around June 6.
Read that again. These are not RIF victims. They are senior-tenured engineers (the median Rule-of-70 candidate is 50-plus with 15+ years at Microsoft), they chose to leave, they walked out with eight to 39 weeks of cash, and they are not flooding LinkedIn with "Open to Work" banners. Nothing about this event triggered a WARN. There's no SEC disclosure beyond a passing mention in earnings. Skillsyncer's tracker does not have a row for it. Neither does TrueUp.
If you're searching for "tech layoffs tracker 2026" to find your next pipeline, you are not seeing any of these 8,750 people.
Why this pool is better than a RIF cohort
RIF cohorts are noisy. They flood Indeed, they all have the same "impacted by the recent reduction" headline on LinkedIn, and they get hit by the same fifteen recruiters in the first week. The signal-to-noise ratio collapses by day three.
Voluntary buyout cohorts are the opposite. They're shoppable, not desperate. They're not broadcasting. Most are weighing "fully retire" against "do something interesting for two years." A targeted, specific outreach in week one of July, before they decompress, lands. A LinkedIn blast does not.
The problem is finding them. They don't self-identify. There's no list. You have to reconstruct the pool: U.S.-based, level-67-or-below, Microsoft tenure plus age above 70, by team and location. This is the kind of query that breaks Boolean. It's a plain-English question with a structural answer, which is why we built Refolk: you describe the cohort the way you'd describe it to a colleague, and you get a ranked shortlist. Our index returns roughly 9,460 senior and principal Microsoft-affiliated engineers in the U.S. concentrated in Greater Seattle, the SF Bay, and the Redmond/Bellevue corridor, which is a workable shadow of the Rule-of-70 eligible pool before any tracker catches up.
ServiceNow's "no backfill": 3,000 to 4,000 invisible exits per year
On the April 22 earnings call, ServiceNow CEO Bill McDermott told investors something most analysts heard as boilerplate cost discipline. It wasn't. It was the new layoff playbook.
As you have attrition in the company, you don't have to backfill it.
McDermott guided that ServiceNow expects to start 2027 with the same headcount it started 2026 with, even after closing the $7.75B Armis acquisition. The mechanism is attrition. Nobody gets fired. People leave on their own timeline, and the seats simply don't come back.
At roughly 27,000 staff and normal tech attrition of 13 to 15 percent, ServiceNow no backfill implies 3,000 to 4,000 roles per year that quietly disappear. None of them trigger a WARN. None of them generate a press announcement. None of them show up on Layoffs.fyi.
Layered on top: an internal April 14 memo eliminated the dedicated Quality Engineering function entirely, and the Armis close created a second wave of overlap exits, particularly on the Now Platform side. None of this is in any tracker. The only public surface for these people is LinkedIn job-change activity and GitHub commit drop-off.
This is the structural piece that the trackers can never fix. The events are designed to be invisible. They evade WARN because there's no involuntary separation. They evade press because there's no announcement. They evade SEC disclosure because attrition isn't material.
Expect this to spread
McDermott just gave every other CFO the script. Workday, Salesforce, Atlassian, and Snowflake all have the same pressure on headcount and the same investor patience for "flat year on year." Watch for ServiceNow no backfill language to appear in Q3 earnings calls under phrases like "natural attrition," "operational leverage," and "letting the org breathe."
The corollary: "headcount-flat" is never "hiring-flat." ServiceNow is still hiring aggressively into AI/ML, customer excellence, and sales while freezing engineering. The internal redeployment-driven exit market inside a flat company is 10 to 15 percent of headcount a year. If you only watch announced layoffs, you miss every one of those people.
Uber's "reorg": 23 percent of a function, framed off the books
On June 3, Uber cut roughly 23 percent of its People & Places division, including recruitment and HR. A spokesperson told CNBC the cuts accounted for "well under 1 percent" of Uber's 34,000 employees, which puts the absolute number around 340. Dara Khosrowshahi's memo called it streamlining. The company specifically said the cuts were unrelated to AI.
Skillsyncer logged this as one event. It did not log it as a 23 percent reduction of a function. The headline number, around 340, is small enough that most trackers treated it as a footnote. But if you're hiring senior TA or HRBP leadership in the next 90 days, that footnote is your entire candidate pool. Nikki Krishnamurthy, the outgoing CPO, departed May 11. Jill Hazelbaker was promoted to president and chief corporate affairs officer around the same time. The senior closers in that division are in motion right now, and they are not broadcasting it because the company framed the whole thing as a reorg.
This is the third evasion pattern. The "reorg" label is a deliberate tracker-evasion tool. When you see "streamlining," "delayering," "operational excellence," "modernizing," or "rebalancing" in a CEO memo, you're looking at a layoff that doesn't want to be one. Sourcers who keyword-monitor "layoffs" miss every one. Sourcers who monitor the euphemism set catch them in the first 48 hours.
The 40 percent figure, where it comes from
Stack the three events. Microsoft Rule of 70: ~8,750. ServiceNow no backfill: ~3,000 to 4,000 in 2026 alone, plus the QE elimination and Armis overlap. Uber People & Places: ~340 senior. That's roughly 12,000 to 13,000 high-quality exits from three companies in a single quarter that appear in zero rows on Skillsyncer.
Pulled forward across the full year and across the ten or fifteen other large-cap tech companies running similar playbooks (Google's Platforms & Devices buyout in January 2025 was the template, Oracle's ~30,000 cuts framed as AI reinvestment, Meta's phased 8,000, Intuit's 3,000), the off-book exit pool is conservatively 40 percent the size of the tracked pool. In senior-IC and senior-manager segments, it's higher. The tracked layoff dataset is junior-heavy because RIFs are junior-heavy. The off-book exit dataset is senior-heavy because buyouts, no-backfill, and reorgs all target tenure.
The trackers aren't wrong. They're counting what they can count. They cannot see voluntary buyout sourcing pools because there is no public artifact to scrape. The fix is not a better tracker. The fix is a different primitive: search the people, not the events.
What to do instead
Stop starting your week with the Skillsyncer feed. Start with the earnings calls. McDermott's "you don't have to backfill it" was the entire ServiceNow no backfill story, on tape, on April 22, six weeks before any LinkedIn job-change signal showed up. Coleman's Rule-of-70 memo was public April 23. Khosrowshahi's "streamlining" language was June 3.
Then convert the signal to a search. Don't search "ex-Microsoft." Search the eligible cohort: U.S., level 67 and below, age plus tenure over 70, in the teams you care about. Don't search "ex-ServiceNow." Search Now Platform engineers and QE leads who changed roles in the last 60 days but haven't surfaced on a tracker. Don't search "ex-Uber People & Places." Search senior TA and HRBP leadership who reported into Krishnamurthy's org.
This is the part that Boolean breaks on, because the qualifying constraints are semantic, not keyword-shaped. Asking Refolk in plain English ("ex-Microsoft principal engineers in Seattle who likely took the Rule-of-70 buyout, not currently posting Open to Work") collapses a four-hour LinkedIn session into a ranked shortlist, and crucially it surfaces the hidden talent pool layoffs trackers can't see because the trackers were never looking.
The 1,093-a-day number is the floor of the talent flux, not the ceiling. The interesting pool is the off-book one. Get there first.
FAQ
Why don't the layoff trackers count voluntary buyouts?
Trackers depend on three triggers: WARN filings, press announcements, and SEC 8-K disclosures. Voluntary buyouts like Microsoft's Rule of 70 evade all three. There's no involuntary separation, so no WARN. The company doesn't want a "8,750 leaving" headline, so no press push. And a voluntary retirement program isn't a material event for SEC purposes. The result is a structurally invisible exit cohort that the trackers cannot reach by design, not by oversight.
How is "no backfill" different from a layoff for sourcing purposes?
Mechanically, it moves the same seats out of the company. Behaviorally, the exits are slow-motion and self-paced, so they don't generate a cohort moment. There's no week-one flood of "impacted" posts on LinkedIn. People leave on their own timeline, often six to twelve months apart, which means competitors and recruiters don't notice the pattern. ServiceNow's no-backfill posture is moving 3,000 to 4,000 seats a year without a single tracker row, and the same playbook is about to spread.
What language signals a "reorg" that's actually a layoff?
Watch for "streamlining," "delayering," "operational excellence," "modernizing," "rebalancing," and "right-sizing the org." When a CEO memo pairs any of these with a specific function name and a percentage (Uber's "23 percent of People & Places"), it's a layoff that's been framed for tracker evasion. The "<1 percent of total headcount" framing is the giveaway, because it admits the cut while making the absolute number look immaterial.
How do I find people from a voluntary buyout cohort if they aren't broadcasting?
Don't search the event, search the eligibility profile. For Microsoft Rule of 70, that's U.S.-based, level 67 or below, tenure plus age above 70, in the teams you want. Cross-reference with GitHub commit drop-off around the July 2 separation date, and with LinkedIn role-change timestamps in the surrounding eight weeks. This is exactly the kind of multi-signal, plain-English query Refolk is built for, and it's the only reliable way to reach a voluntary buyout sourcing pool before everyone else figures out it exists.