The Rival RTO Mandate Read: Bleeding, Holding, or Paper-Only

This is the read I run when a rival announces or tightens a return-to-office mandate and leadership wants to know, within two weeks, whether it is costing that rival senior people you could hire or whether it is a memo nobody enforces. It is for talent-intelligence analysts and strategy teams, and it works on one named rival and one function at a time. It scores three dimensions from public evidence and gives one of three verdicts - Bleeding, Holding, or Paper-Only - with the action each one triggers and the date to re-score.

What do Bleeding, Holding, and Paper-Only mean?

The verdict combines two kinds of evidence: is the mandate enforced, and are senior people leaving because of it. Bleeding means an enforced mandate is pushing senior people out to employers you can identify. Holding means the mandate is enforced but senior people are staying. Paper-Only means the policy exists in writing and nothing shows that anyone checks attendance.

There is a fourth pattern: senior departures with no enforcement. When you see that, something other than the mandate is driving the exits.

Terms used throughout:

• Return-to-office (RTO) mandate: a written policy setting a minimum number of in-office days, with or without exceptions and phases.
• Enforcement evidence: any public sign that attendance is measured and has consequences.
• Seniority-weighted departure lift: the change in the senior leaver rate for one function against that function's own pre-mandate baseline. Absolute senior headcount is always checked alongside the share.
• Destination clustering: leavers turning up repeatedly at the same new employers, especially employers with looser office policies.

Verdict rules

Each dimension gets a score from 0 to 2. The scoring rules are in the sections below.

• Bleeding: enforcement E1 or E2, and lift L2, or lift L1 with destinations D2.
• Holding: enforcement E1 or E2, and the lift and destination evidence fall short of Bleeding after the full window.
• Paper-Only: enforcement E0 and lift L0.
• Other cause: enforcement E0 with lift L1 or L2.

The two-week deadline creates an imbalance. On day 14 you can usually place the rival on the horizontal axis with confidence, because enforcement evidence is observable straight away. The vertical axis takes months to fill in. That is why a Paper-Only verdict can be close to final early on, while an early Bleeding verdict cannot.

Why are mandate strictness and enforcement separate questions?

Strictness tells you how many senior people could leave if the mandate bites. Enforcement tells you whether it bites at all. The published evidence supports both, and mixing them up is the commonest way this read goes wrong.

Strictness predicts the size of the senior loss

Van Dijcke, Gunsilius and Wright, in "Return to Office and the Tenure Distribution", used resume data from a range of tech companies. Coverage reported the dataset as 260 million resumes. The method is distributional synthetic controls: it builds a counterfactual, no-mandate version of each company from a weighted blend of comparable companies, then compares the whole tenure distribution against it. Together, Microsoft, Apple, and SpaceX employ more than 2% of the tech workforce and earn 30% of the industry's revenue.

The rows do not quite share a unit: Microsoft's change was reported in percentage points, the others as relative change. The direction is still clear. More required days came with a larger senior loss. The authors also found that several senior leavers moved to large, direct competitors without an RTO mandate. Microsoft disputed the findings.

The financial sector has strict cases too. JPMorgan Chase has required full-time office attendance since January 2025, and Truist did the same a year later. Ding, Ma, Xing, Yang and Jin, whose work is covered below, studied 54 large technology and financial firms.

A higher required-days number does not prove enforcement

The Flex Index tracks office policy at over 13,000 companies worldwide, more than 8,500 of them in the US. Its series shows required days rising much faster than attendance.

"WFH days stable" comes from Stanford's SWAA survey and the Census Household Pulse Survey. Over the same period, full-time-in-office requirements among Fortune 500 firms rose from 13% to 24%. A building-access occupancy barometer covering 2,600+ buildings measured a 56.3% weekly average against a 66.0% Tuesday peak. Policies got stricter while behaviour barely moved.

How do you tell if an RTO mandate is enforced?

A mandate is enforced when attendance is measured per person and has consequences: badge-swipe tracking, manager dashboards, and links to promotion or performance review. Mark each signal present, absent, or unknown, with a source. Required days and strong wording do not count.

Score the dimension:

• E2 Hard: at least one hard signal, with a source.
• E1 Soft: soft signals only, with at least two from different sources.
• E0 Paper: the memo and nothing else, or an official denial while postings stay flexible. Unknown counts as absent for scoring, but list every unknown so the re-score can close it.

Two documented cases set the bar for E2. Dell rolled out its mandate in three phases over about thirteen months: 39 office days per quarter first, then badge tracking about three months later, then five days a week. It also made non-compliant staff ineligible for promotion. Amazon gave managers a dashboard that pulled every badge tap into an eight-week rolling timeline, refreshed daily. Both connect attendance data to a decision someone makes about a person. That link is what you are looking for.

Reading job postings without being fooled by ghost listings

A ghost job is a posting with no active intent to hire. Ghost postings matter here because a ghost listing with an office-day field describes the policy for a job nobody is filling. On one large applicant-tracking platform, 18-22% of postings in any quarter are classified as ghost jobs. That classification includes frozen and cancelled requisitions as well as deliberate fakes, and construction reached 38%. The Congressional Research Service says there are no official statistics on the size of the problem, and no published method exists for discounting office-day fields. I apply three filters of my own:

1. Drop postings that are old relative to the function's normal time to fill.
2. Drop postings reposted repeatedly with unchanged text.
3. Give extra weight to postings where someone has visibly been hired into the role.

Then compare the surviving postings opened after the announcement with those opened before it. A switch from "remote" to a named office supports the enforcement score. Live postings that stay flexible after the announcement count against it.

How do you measure senior attrition after an RTO mandate?

Compare the senior leaver rate and the absolute senior headcount in the observation window against the function's own 4-8 quarter baseline. Only then compare against an external benchmark. The internal comparison is the evidence. The external benchmark is a sanity check.

Define "senior" once and write it down

The resume study read seniority from where a person sits in the tenure distribution. The exact rule it used to label someone senior from profile data is not given in the sources I could check, so treat it as unverified. Pick your own rule, either a title band or a tenure threshold, write it on the scoring sheet, and apply it identically to the baseline and the window. If the definition changes between the two, the lift you measure is meaningless.

Build a 4-8 quarter baseline

For each quarter before the announcement, record absolute senior headcount, senior share of the function, and senior leavers. The 4-8 quarter range is my inference, not a published standard. Fewer quarters leave too few events to see a pattern. Go much further back and you start measuring a different company. Counting a rival's senior band by title and employer is where Refolk saves the most time, because you can ask for the band in plain English and get the profiles back, rather than stitching together title variants by hand.

Set two observation windows and a lag buffer

Sources disagree on where the window should start. Ding et al. analysed RTO announcements, which catches people who leave in anticipation. Starting at the effective date catches people who leave once enforcement starts. Run both. Then add a lag buffer to the end of each window, because public profiles trail real moves. No rigorous published median exists for the gap. Career advice ranges from a couple of weeks to three months, and one modelled study puts cached databases 45-120 days behind a real job change.

Calculate the lift

Relative lift is the window's senior leaver rate minus the baseline senior leaver rate, divided by the baseline rate. Run it for both windows, then check absolute senior headcount, because share can move without anyone leaving.

To calibrate, use Ding et al., "Return to Office Mandates, Brain Drain and Gender Difference". Across 54 large tech and financial firms, average voluntary turnover rose 14% after mandates, time to fill rose 23%, and the hire rate fell 17%. The rise was larger for women (20%), skilled workers (18%), and top managers (19%). Sources disagree on units: one describes the lift as roughly 13-14 percentage points, while most report a 14% relative rise. I treat it as relative.

External benchmarks, for the sanity check:

A low baseline hides a real effect. A 14% relative lift on a 7.5% engineering baseline gives 8.55%. On a 13.0% US baseline it gives 14.8%. In a 200-person engineering function, the first case is about 2 extra leavers a year, which a small team cannot tell apart from noise. One European benchmark puts average engineering tenure at 1,084 days, about three years, so the quarterly counts are small to begin with.

Score the dimension:

• L2: senior leaver rate up at least 14% relative to baseline, absolute senior headcount down, and no layoff in the window.
• L1: rate up but below that, share down while absolute count holds, or the window has not closed.
• L0: at or below baseline after the full window and lag buffer.

Where do the leavers land, and why does it matter more than the rate?

Destinations are the strongest evidence in this read. When senior leavers cluster at employers with looser office policies, that ties the exits to the mandate in a way a rate cannot. It also tells you exactly who you are competing with for those people.

The resume study's finding that senior leavers moved to large direct competitors without a mandate is the pattern to look for. List every senior leaver in the window, group them by new employer, and check each destination's office policy against its Flex Index entry or official wording. Then look at location changes: a move from one named city to another supports a real relocation.

Score the dimension:

• D2: one or more named destinations recur, and their office policy is looser than the rival's.
• D1: leavers go mostly to looser-policy employers, but no destination recurs.
• D0: leavers scatter, or go to employers with equal or stricter policies.

What Refolk's index shows about remote signals

Do not shortcut this dimension by searching profiles for the word "remote".

Two of the 25 sampled remote-headline profiles list Microsoft as their current employer. That is an anecdote, not a finding. Departures you can observe are a better signal than a keyword. Pulling leavers by previous employer and recent job change is the slow part of this dimension when done by hand. In Refolk you describe the cohort in a sentence and get the profiles back, ready to group by destination.

The 14-day procedure

These steps are my synthesis. No single published source documents the full sequence, and the day ranges assume one analyst working one rival and one function.

Keep one sheet per rival and function. Put every source next to the line it supports.

Rival:
Function:
Senior definition (title band or tenure rule):
Announced:            Effective:            Required days:
Phases and exceptions:

ENFORCEMENT (present / absent / unknown + source)
Badge data tied to promotion or review:
Manager attendance dashboard:
Badge-swipe tracking:
Anchor days or manager check-ins:
Forum reports (number of separate threads):
Live postings with office-day field:
Enforcement score (E0 Paper / E1 Soft / E2 Hard):

LIFT
Baseline quarters used (4-8):
Baseline absolute senior headcount:      Baseline senior leaver rate:
Window A start (announcement):           Window B start (effective date):
Lag buffer added (days, 45-120):
Window absolute senior headcount:        Window senior leaver rate:
Relative lift vs internal baseline:
External benchmark used and rate:
Layoff news or WARN notice in window (yes / no + source):
Lift score (L0 / L1 / L2):

DESTINATIONS
Senior leavers found:
Destination :: count :: office policy
Destination :: count :: office policy
Destination score (D0 / D1 / D2):

VERDICT (Bleeding / Holding / Paper-Only / Other cause):
Provisional or final:
Action:
Re-score date:

How does this read go wrong?

Most wrong verdicts come from crediting the mandate with something it did not cause, or from trusting a signal that only looks like evidence. Run every one of these checks before you sign the sheet.

1. Layoffs look like mandate exits. A senior share that falls during a layoff is not voluntary attrition. The resume study's mandates came before the large tech layoffs that began in 2022, which is how it kept the two apart. Check: search for layoff news and WARN notices inside both observation windows. If you find either, drop to Other cause until you can separate the two.
2. Share can fall without anyone leaving. At Microsoft, the share of positions held by sub-senior employees rose by over four percent. A wave of junior hiring alone produces a falling senior share. Check: count senior people in absolute numbers, not just as a share.
3. Profile lag hides or invents a spike. A short window undercounts leavers who have not updated yet. NBER work cited in a vendor freshness study found 19.7% of LinkedIn users go back and edit a past job, which can move a departure into the wrong quarter. Check: compare each leaver's stated end date with the date you first saw the change.
4. Ghost postings describe a policy for roles nobody is filling. Evergreen listings with office-day fields make the policy look live. Check: posting age, repost count, and whether anyone has actually been hired into the role.
5. The remote keyword is too rare to use. Only 0.13% of US senior software engineers in Refolk's index put "remote" in their headline. Check: never treat finding no keyword as evidence of anything. Use observed departures instead.
6. A country-only location is not a move. Changing a location to just "United States" may be a privacy choice, not a relocation. In a 25-profile sample from Refolk's index, 12% of US profiles and 12% of UK profiles showed a country-only location, rising to 20% in the remote subset. That sample is indicative only. Check: require a named city both before and after the change.
7. Forum reports skew. Threads on employee forums such as Blind come from people who chose to post. Check: treat them as a soft sign of enforcement, never proof, and count separate threads rather than replies in one thread.
8. The company says there is no mandate. Microsoft said it was "inaccurate to say we have a return to office mandate". Check: read the exact wording of the official policy and its Flex Index entry, then score enforcement as usual. A denial does not lower the score, and coverage calling it a mandate does not raise it.

There is a ninth failure that belongs to the analyst rather than the data: calling a final Bleeding on day 14. In the first fortnight, the lift and destination dimensions are mostly empty, however vivid the leaked memo is. Write "provisional" on the verdict and put the re-score date in the same sentence.

What each verdict triggers, and how to keep the read current

Each verdict comes with one default action and one re-score date. A practitioner tracking tool recommends following a mandate for the two quarters after it lands, so the default re-score is the end of the second quarter after the effective date.

Actions by verdict

• Bleeding: open targeted outreach to the affected senior band in that function, led by your own office flexibility. Prioritise people at or near the destination clusters, because those employers are your competition for them. When you build the slate, Refolk can pull the cohort by previous employer and change date in one request.
• Holding: run no campaign. Keep a watchlist of the senior band and re-score on schedule. An enforced mandate that people tolerate today can tip at the next phase, as Dell's three-step rollout shows.
• Paper-Only: spend nothing on a mandate-themed pitch. Send strategy a note saying the rival's policy is not a differentiator for now. Re-score as soon as an enforcement signal appears, such as reported badge tracking or a review link.
• Other cause: find the real driver first, whether a layoff, reorg, or junior hiring wave, then decide whether it opens a different recruiting angle.

Keeping the read current

Benchmarks drift. One European tech attrition series went from 27.0% to 17.4% to 11.1% across three consecutive reports, and the same source puts the UK at 12.1% and Germany at 15.3%. Re-pull every external benchmark before each re-score rather than reusing the last one. Re-check the rival's Flex Index entry too, because mandates are often phased. Rebuild the internal baseline only if your definition of senior changes, and if it does, rescore every earlier window with the new definition.

Rival and function: [rival], [function]
Verdict: [Bleeding / Holding / Paper-Only / Other cause], [provisional / final]
Why, in one line each:
- Enforcement: [E score] because [strongest sourced signal]
- Senior lift: [L score], [window rate] vs [baseline rate], absolute senior headcount [up / flat / down]
- Destinations: [D score], top destination [employer] with [count] leavers
What it does not show yet: [the dimension still open and why]
Action: [the default action for this verdict]
Re-score date: [end of second quarter after effective date]