{
  "team": "MODEL ANSWERS \u00b7 all activities",
  "arc": "inverted",
  "conf": "Low",
  "fields": {
    "m8_view": "The turnover-by-unit chart.",
    "m8_remove": "1. The 3D effect, which distorts the bar heights and adds nothing.\n2. The legend, which repeats what the axis labels already say.\n3. The gridlines and the rainbow palette. Seven colours for one variable makes the eye work for no reason.",
    "m8_emph": "Units C and F in navy, every other unit in grey, and the units sorted by rate rather than alphabetically. The eye should land on the two units in trouble within a second. What we deliberately let recede: the five units that are fine. They are still visible, so the reader can see that most of the organization is not the problem.",
    "m8_honest": "The axis started at 8%, which made an 8% unit look like nothing and a 12% unit look like a crisis. Unit E effectively disappeared. We set the baseline to zero. The decision that changes: with a truncated axis you would fund a broad intervention because everything looks bad. At a true zero you can see the problem is two units, and that is a much cheaper and better-targeted response.",
    "m8_title": "\u201cTurnover is concentrated in two units, not spread across the organization\u201d (replacing \u201cQ3 Workforce Metrics\u201d, which made the reader find the finding themselves).",
    "m9_decision": "Do we intervene in the two high-turnover units now, investigate further before committing, or do nothing this cycle? The budget decision lands this quarter, so it is a real choice with a date on it, not a topic.",
    "m9_sit": "Workforce readiness has slipped for four straight quarters, and voluntary turnover is up from 12% to 16%.",
    "m9_ten": "The budget decision lands this quarter, and it funds either two units or a small across-the-board step. Not both. Doing nothing means the money goes elsewhere and we revisit this a year from now with the same problem and less credibility.",
    "m9_ev": "Turnover is concentrated in two units and falls steeply as supervisor tenure rises. Exit themes and engagement point the same way. It does not track pay position. Three independent sources converge.",
    "m9_rec": "Fund targeted supervisor support in units C and F this quarter. Do not spread the money across the organization.",
    "m9_if": "If we are right, we address the two units carrying most of the problem, at a fraction of the cost of a broad program, and we learn within two quarters whether supervisor support moves turnover.",
    "m9_ifnot": "If we do nothing, turnover in those units continues at 25 to 27% and the certification lapses that follow each departure keep compounding. If we spread the budget thinly instead, we spend the same money, help nobody measurably, and cannot tell afterwards whether it worked.",
    "m9_assume": "That supervisor tenure is a reasonable stand-in for supervisory quality. That the newest supervisors are not simply assigned to the hardest units, which would make this a unit problem wearing a supervision costume. That two quarters is long enough to see movement.",
    "m9_limit": "We cannot say supervision causes the turnover. We can say it travels with it and that pay position does not. We also do not know why those supervisors are struggling, which means we are funding support without knowing exactly what support is needed.",
    "cap_insight": "Most of the turnover problem sits in two units and moves with supervisor tenure, not with pay.",
    "cap_aud": "The executive who controls the budget line. They need a decision, not an education. They have about five minutes and they already believe it is a pay problem, so the recommendation goes first and the evidence supports it.",
    "cap_believes": "They believe it is pay. Leadership said so in the original request, and it is the explanation that requires no one to look at their own management. Anything we say that sounds like dismissing pay will be heard as defending supervisors.",
    "cap_tell": "We came in expecting a pay problem. That is what the request said, and it is what we would have tested if nobody had asked us to look wider.\n\nHere is what we found. Turnover is not an organization-wide problem. It sits in two units, at 25 and 27%, against 8 to 12% everywhere else. And it falls steeply as supervisor tenure rises: 24% under a supervisor with less than a year in role, 8% under one with three-plus.\n\nPay does not explain it. Turnover is flat across pay position, 13, 12, 11%, and recruiting still sees candidates accepting offers at the usual rate.\n\nSo we recommend funding targeted supervisor support in those two units this quarter, rather than a small step across everyone. Our confidence is moderate, not high, and there is one thing we could not establish: we do not know whether the newest supervisors were assigned to the hardest units. If they were, this is a staffing problem rather than a supervision one. That is the first thing we would check.",
    "cap_moves": "Contrast, once, in the insight: \u201cUnder a three-year supervisor, 8 percent. Under a first-year supervisor, 24 percent.\u201d Those two numbers next to each other do the persuading, so nothing else is needed there.\n\nAnd one analogy in the opening, to make the headcount feel operational rather than statistical: \u201cwe are losing people fastest in the two units we can least afford to lose them from.\u201d\n\nNothing else. The evidence is strong enough that adding a third move would look like we were compensating for something.",
    "cap_conf": "Moderate, because three independent sources converge and we actively looked for the disconfirming case, but we have no direct measure of supervisory quality and we have not ruled out assignment patterns.",
    "cap_limit": "We do not know why those supervisors are struggling, and we have not ruled out that they were assigned to the hardest units in the first place.",
    "m6_pattern": "Turnover is concentrated in two units (C at 27%, F at 25%) against 8 to 12% elsewhere, and falls steeply as supervisor tenure rises: 24% under one year, 8% at three-plus. It does not track pay position.",
    "m6_exp1": "Supervision quality",
    "m6_pred1": "If supervision drives it, turnover should cluster under particular supervisors, and engagement should be lower under them.",
    "m6_see1": "It does. 24% under supervisors with under a year, 8% at three-plus. Engagement 2.9 against 3.8. Exit interviews cite manager at 45%.",
    "m6_nofit1": "Supervisor tenure is a proxy for quality, not a measure of it. We have no direct assessment.",
    "m6_need1": "Where the newest supervisors are assigned. If they are concentrated in the hardest units, we cannot separate supervision from the unit.",
    "m6_exp2": "Workload and overtime",
    "m6_pred2": "If workload drives it, turnover should rise with overtime hours and be highest where staffing is thinnest.",
    "m6_see2": "Workload is the second exit theme at 30%. Both affected units are small, so each departure raises load on those who remain.",
    "m6_nofit2": "We have no overtime or staffing-ratio data by unit, so the prediction is untested rather than confirmed.",
    "m6_need2": "Overtime hours and vacancy rates for C and F against the rest.",
    "m6_exp3": "Compensation",
    "m6_pred3": "If pay drives it, we should see more exits where pay sits furthest below market, regardless of unit.",
    "m6_see3": "Turnover is flat across pay position: 13%, 12%, 11%. Recruiting also reports offers accepted at the usual rate.",
    "m6_nofit3": "The prediction failed. The slope it requires is not there.",
    "m6_need3": "Pay position inside C and F specifically. The flat line is an organization-wide average.",
    "m6_exp4": "Training and readiness gaps",
    "m6_pred4": "If inadequate training drives exits, turnover should be concentrated among newer or less-trained employees, and the affected units should be less trained than the rest.",
    "m6_see4": "Training completion is 82% and flat across every unit. Exposure is the same everywhere.",
    "m6_nofit4": "The prediction failed. If training were the driver, the affected units would differ. They do not. Separately, certification lapses begin AFTER the departures, so the lapses are a consequence rather than a cause.",
    "m6_need4": "Nothing further. Flat exposure rules it out as the explanation for the unit differences.",
    "m6_setaside": "Compensation. It was the explanation everyone started with, so it deserved a real test rather than a dismissal. Its prediction was a slope across pay position. There is none, and recruiting sees no change in offer acceptance. We are not saying pay does not matter; we are saying it does not explain the concentration in two units.",
    "m6_best": "Supervision is the strongest contributor, at moderate confidence, amplified by workload in the two understaffed units. Compensation is ruled out as the explanation for the unit differences, and the training backlog is a symptom rather than a cause. Not proven: we have no direct measure of supervisory quality. Next step: act narrowly and gather. Check supervisor assignment patterns in C and F before committing anything wider.",
    "m9_conf": "Moderate, because three independent sources agree and we looked for the disconfirming case. Not high: we have no direct measure of supervisory quality.",
    "m7_f1": "Turnover is concentrated in two units: C at 27% and F at 25%, against 8 to 12% everywhere else.",
    "m7_q1": "How does turnover compare across the seven units?",
    "m7_t1": "Comparison. Turnover rate is a percentage with a true zero, so bars are legitimate.",
    "m7_c1": "A horizontal bar chart, sorted by rate rather than alphabetically, with C and F in navy and the rest in grey. Not a line: these are seven separate units, not a sequence. Baseline at zero, because truncating it would make the 8% units look like a crisis too.",
    "m7_f2": "Turnover falls as supervisor tenure rises: 24% under one year, 14% at one to three, 8% at three-plus.",
    "m7_q2": "Does turnover move with how long the supervisor has been in role?",
    "m7_t2": "Relationship, shown as an ordered comparison. Turnover has a true zero; supervisor tenure is banded rather than continuous.",
    "m7_c2": "A simple three-bar chart in tenure order, with the gradient visible. A scatter would be better if we had supervisor-level data, but we only have three bands, so a scatter would imply a precision we do not have.",
    "m7_f3": "Turnover is flat across pay position: 13%, 12%, 11% from below market to above market.",
    "m7_q3": "Does turnover differ by where pay sits against market?",
    "m7_t3": "Comparison, and the point is the absence of a difference. True zero applies.",
    "m7_c3": "Bars at a true zero, deliberately. This is the one chart where the zero baseline carries the argument: at zero the three bars look almost identical, which is the finding. Truncated at 10% they would look like a steep decline and say the opposite.",
    "lad_belief": "That workforce readiness is slipping because our pay is not competitive.",
    "lad_q1": "Request a market pay adjustment for the affected units, and ask for it this budget cycle.",
    "lad_q2": "That people leave for money, and that pay is the lever we control that would change it.",
    "lad_q3": "That we have a compensation problem. I concluded it from a turnover number that rose from 12% to 16%, and from leadership saying so in the original request.",
    "lad_q4": "That what leavers say on the way out is why they actually went. And underneath that, that the reason people give in an exit interview is a reason rather than a socially safe answer.",
    "lad_q5": "That 16% is high. I never said compared to what: not to last year in the same units, not to comparable organizations, not to any figure at all. It felt high because it had risen.",
    "lad_q6": "I looked at turnover and pay. I did not look at supervisor tenure, workload, or who gets assigned where. Supervisor assignment data existed on the first morning and I never opened it.",
    "lad_test": "Necessary: yes. If leavers do not leave primarily for money, the whole recommendation falls over. Supported: weakly. One source, and it is self-reported by people with a reason to be diplomatic. Testable: yes. Turnover by pay position would show a slope if pay were driving it, and it does not.",
    "lad_next": "Write it into the assumptions field of the next brief, as a stated assumption rather than a background belief, and pull the pay-position cut before recommending anything.",
    "m2_rev": "Reversible in principle, but not cheaply. A pay adjustment cannot be withdrawn once granted, and a supervisor support program can be stopped after a quarter. So the two options on the table sit at opposite ends of this.",
    "m2_stakes": "Money, credibility and people. Roughly $2.4M of one-time funds, the standing of the Human Capital office if we spend it on the wrong thing, and continued loss of mission-critical staff if we do nothing.",
    "m2_unc": "High. We have a request, a belief and a dashboard. We have not established why anybody is leaving, and we have never done this analysis before.",
    "m2_bar": "A focused check on one or two things. Not a quick read, because the money is significant and partly irreversible. Not the full investigation, because the funds expire this quarter and a six-month study would answer a decision nobody is still making.",
    "m2_decision": "Whether to fund a targeted intervention in the units that are actually losing people, a small across-the-board pay step, or nothing this cycle.",
    "m2_bounds": "The budget is fixed and expires at end of fiscal year. Headcount cannot be increased. Pay bands are set outside this directorate. Any intervention has to start within the quarter.",
    "m2_success": "We will know this worked if turnover in the affected population falls over the following two to three quarters, and if we can say afterwards which of our reasons was right.",
    "m2_assume": "That the readiness index actually measures readiness. That people who left would have stayed if the cause were removed. That leadership's stated belief about pay came from evidence rather than from what leavers said on the way out.",
    "m2_bq": "Why is voluntary turnover rising in mission-critical roles, and what would actually reduce it?",
    "m2_aq": "Is turnover evenly distributed across units, or concentrated? Does turnover vary with pay position against market? Does turnover vary with any characteristic of the work environment, such as supervisor tenure?",
    "m2_hyp": "Turnover is driven primarily by pay falling behind market, and it will be highest where pay sits furthest below market.",
    "m2_dis": "If turnover is flat across pay position, or if it clusters somewhere pay does not explain, the pay hypothesis is in serious trouble. That is the first thing we will check.",
    "m2_eb": "Turnover by unit. Turnover by pay position against market. Turnover by supervisor tenure. Headcount by unit so we can compute rates rather than counts. Nothing else yet.",
    "m3_place": "The dashboard sits at level 3, multiple observations from one source or method. It is a real measurement across the whole population, but one system and one metric.\n\nThe senior manager sits at level 2, single observation or source. \u201cI hear it constantly\u201d is a recollection, not a count.\n\nRecruiting sits at level 4, independent measure or source. It comes from a different system, a different population and a different process.",
    "m3_can": "The dashboard can establish that turnover rose. It cannot establish why anybody left.\n\nThe manager can establish that pay is what departing people say. It cannot establish that pay is why they went.\n\nRecruiting can establish that candidates still accept our offers at the usual rate. It cannot establish what our pay is worth to somebody two years in.",
    "m3_why": "Recruiting is the strongest, and that surprises people because it sounds like the smallest. It is the only source whose failure mode is unrelated to the other two. If the dashboard is wrong it is wrong about definitions or timing; if the manager is wrong it is about small samples and what people find easy to say. Offer acceptance fails in neither of those ways, so if it agrees with something, the agreement means something.",
    "m3_sentence": "Current evidence is mixed and weak. Our confidence that this is a compensation problem is low, and the action that earns is to gather. Specifically: turnover broken out by pay position, which would settle the manager's claim, and by unit and supervisor tenure, which is where the remaining explanations live.",
    "m4_metric": "Engagement score: 3.4 out of 5, down from 3.7.",
    "m4_u": "What is engagement here? It is a survey index, and indexes are constructed. Two organizations both reporting 3.4 may be measuring different things.\n\nWho answered? A 3.4 among 40% of staff and a 3.4 among 90% are different numbers, and the least engaged people are the least likely to respond. Nothing on the dashboard gives us the response rate.",
    "m4_i": "Engagement genuinely fell across the workforce.\n\nOr the composition of respondents changed. If more disengaged people answered this time, or a group that previously skewed high stopped answering, the average moves without anybody's experience changing.",
    "m4_e": "Enough to justify looking further. Not enough to name a cause or fund a response. Confidence: low to moderate.\n\nIt is a real signal from a real instrument, and it is a blended average from an unknown population with no distribution behind it.",
    "m4_line": "The data stops at \u201cthe mean of the responses we received fell from 3.7 to 3.4.\u201d Everything after that is assumption: that respondents represent the workforce, that the instrument did not change, and that 0.3 is meaningful rather than noise.",
    "m4_q": "What was the response rate, and did it change between the two surveys? If it moved, the score may not have.",
    "m5_moves": "Check the rate, not just the count. Unit D has 45 leavers, more than anywhere else, at a 10% rate. Unit C has 40 at 27%. By headcount D looks worst; by rate C is.\n\nSegment before you generalize. Turnover is 8 to 12% almost everywhere and 25 to 27% in two units. There is no organization-wide turnover problem.\n\nLook for what moves together. Turnover falls steeply as supervisor tenure rises: 24%, 14%, 8%.\n\nCompare before you conclude. Turnover is flat across pay position: 13%, 12%, 11%. If pay drove this, that line would slope.",
    "m5_p1": "Turnover is concentrated in two small units rather than spread across the organization. C at 27% and F at 25%, against 8 to 12% in the other five.",
    "m5_p2": "Turnover falls as supervisor tenure rises, and does not vary with pay position. 24% under a supervisor with less than a year, 8% under one with three years or more, while pay position is nearly flat.",
    "m5_q": "What is different about those two units? Sharper version: do C and F also have the newest supervisors? That is the question this data invites and cannot answer."
  },
  "zones": {
    "pool": [],
    "cut": [
      "f0",
      "f1",
      "f2",
      "f8",
      "f9",
      "f10"
    ],
    "keep": [],
    "arc0": [
      "f3"
    ],
    "arc1": [
      "f5"
    ],
    "arc2": [],
    "arc3": [
      "f4"
    ],
    "arc4": [
      "f7",
      "f6"
    ],
    "arc5": [
      "f11"
    ],
    "arc6": []
  }
}