File: 2026.10.23 - Kestrel Medical - 02 KPI Calculation Pack.md

Kestrel Medical: KPI Calculation Pack

Sample data. Every number below is calculated from the Sample data in 03 - Working Files/data/ by 03 - Working Files/calc_kpis.py (results in kpi_results.json) and 03 - Working Files/kewpie_substitutes.py (stand-ins, results in substitutes_results.json). Data runs to 2026-09-30. Nothing here is real Kestrel data: check it against real data before go-live.

Version: Calculation Pack v1.1 for Business Requirements v1.1. Date: 23-Oct-26. Author: Kewpie, KPI Creator Stage 2. Replaces v1.0.

What changed from v1.0

• Four KPIs are Phase 2 placeholders, as Adam decided: Gross margin per cent (K1), Gross margin by product group (K1.1), Monthly reports that match the agreed set of numbers (K5), Our price against the winning price on lost tenders (K4.1). Each shows 'Phase 2: waiting for data', with a labelled stand-in beside it. The calculation is still documented, and a named Phase 2 data source is recorded (section 4).
• Average discount off list price on big deals (K2): the target is now a straight-line path from 17.03% (Sep-26) to 14% (Dec-27), so pace needed moves from 0.00 to -0.20 points a month. Baseline unchanged at 17.03%. Status is Green today, but the run rate is flat, so it goes Amber from Oct-26 and Red from Feb-27 if nothing changes.
• Big tender win rate (K4): now a count over the last 5 big tenders: 2 of 5 (the same five tenders as before). The target is open (Deb sets it). Against Gary's starting proposal of 3 of the last 5, the next 2 big tenders must both be won.
• Revenue growth on the year (K3): numbers unchanged (-6.8%, pace +0.72, run rate +0.27). The 4% is now measured on 2026 full-year actuals: about $203.5m in 2027 (estimate), still 3.1% below 2025.
• Challenge notes are marked resolved or still open as Business Requirements v1.1 records them (section 5).

1. Summary: the Top 5

Status uses Gary's thresholds. Pace needed is (target - current) / months remaining, with months remaining counted from Sep-26 to Dec-27 (15). A Phase 2 placeholder shows no number, status colour, sparkline or pace needed.

Stand-ins beside the placeholders (not KPIs)

Each stand-in is labelled 'Stand-in, not the KPI' and 'Sample data'. It has no target and no status colour, and never uses the KPI's name or finance's stated 41%.

2. The calculations

Gross margin per cent (K1)

• Result of the check on Gary's draft: Confirmed (Phase 2 KPI: placeholder tile on the dashboard)
• Plain English: Net sales less cost of sales, over net sales.
• Calculation: (SUM(net_value_aud) - SUM(cost_of_sales)) / SUM(net_value_aud)
• Sources and why: sales_order_lines.net_value_aud (the sales system of record). Cost of sales: no field found. Phase 2: cost of sales per shipped order line from SAP product costing and billing documents, joined on order and line number.
• Grain: Order line (each order has one line)
• Filters: Shipped lines, January to December. Must reconcile to finance's 41% before use. Placeholder tile until then; the stated 41% may appear only as a text note on the placeholder, never in the headline slot and never on the stand-in.
• Baseline: Phase 2 placeholder: waiting for cost of sales from finance. No baseline can be calculated from the Sample data: no cost of sales field exists in any of the 11 files. Finance's stated figure of about 41% is a text note on the placeholder only, not calculated. Period: Phase 2 (the stated figure is the current year, from the sales leader). Evidence: Working Hypothesis (a stated figure, not reproducible)
• Pace needed: Not applicable until Phase 2: the placeholder shows no pace needed
• Run rate: Not available: no history without cost of sales
• Reading it: Phase 2: waiting for data. Cannot say whether on pace until cost of sales is supplied
• Counter-metric: Net sales value year to date (Revenue growth on the year, K3), read alongside: a margin lift bought by dropping low-margin sales shows as falling net sales. Net sales Jan to Sep 2026 are $149.36m, -6.8% on the same months of 2025
• Notes: Phase 2 KPI. Formula confirmed as the right shape. Data mapping NOT confirmed: no cost field (checked every column in all 11 files; challenge note C1). Phase 2 needs cost of sales per shipped order line from SAP product costing, joined on order and line number, reconciled to finance's 41% before switch-on. First pass: placeholder tile plus the stand-in Net price realisation (Adam's choice), which never carries this KPI's name, target or status colour.
• Stand-in (first pass only, not the KPI): 87.36% of list price kept on shipped lines (net sales $149.36m), against 87.67% in the same months of 2025. By full year: 2021 89.72%, 2022 89.28%, 2023 89.28%, 2024 87.61%, 2025 87.22%. Excluding warranty replacement lines: 87.26% against 87.58%. Price side only: this is not gross margin. Period: January to September 2026 against the same months of 2025, ship date (Sample data). Evidence: Observed Fact for the calculation; Working Hypothesis that it moves with gross margin (no cost data to test the link). Calculation: SUM(net_value_aud) / SUM(list_price * qty), shipped lines, ship date

Average discount off list price on big deals (K2)

• Result of the check on Gary's draft: Corrected
• Plain English: One minus net value over list value, across big-deal lines, weighted by value.
• Calculation: 1 - SUM(net_value_aud) / SUM(list_price * qty), order_type = tender_contract
• Sources and why: sales_order_lines.list_price, .qty, .net_value_aud, .order_type, .order_date. discount_pct used as a cross-check only.
• Grain: Order line
• Filters: Order type tender_contract (placeholder until Sharon sets a size rule); period by order date; no lines excluded for missing list price (none).
• Baseline: 17.03% value-weighted discount off list on tender contract order lines (5,485 lines). Placeholder big-deal definition. Period: January to September 2026, order date. Evidence: Observed Fact (calculation); Working Hypothesis for which orders count as big deals
• Pace needed: -0.20 points a month: a straight-line path from 17.03% (Sep-26) to 14.0% (Dec-27) over 15 months. Path: Dec-26 16.42%, Jun-27 15.21%, Dec-27 14.00%
• Run rate: 0.000 points a month over the last 3 months (flat)
• Reading it: Green today (on the path at its starting point) but off pace: a flat discount goes Amber from Oct-26 and Red from Feb-27 (more than 1 point above the path). It has to fall about 0.20 points a month
• Counter-metric: Tender contract net sales and the Big tender win rate (K4): pulling back discount can simply lose the deals. Tender contract net sales Jan to Sep: $86.7m in 2026 against $95.1m in 2025
• Notes: Corrected (C4, accepted in v1.1): the draft placeholder 'orders linked to a contract_id' picks up service contract lines, not tenders. No order reaches $5m, so an order-total rule gives nothing. v1.1 changes the target to a path back towards about 14% (2023 level) by Dec-27, Gary's proposal for Sharon to confirm (C5). Discount recalculated from list_price, qty and net_value_aud agrees with discount_pct (99.9% of lines within 0.002 points). The 2026 reference is Jan to Sep until the year closes.

Target path (v1.1, Gary's proposal, for Sharon to confirm): straight line from 17.03% (Sep-26) to 14.0% (Dec-27), -0.20 points a month: Sep-26 17.03%, Dec-26 16.42%, Mar-27 15.82%, Jun-27 15.21%, Sep-27 14.61%, Dec-27 14.00%. Green at or below the path, Amber up to 1 point above, Red more than 1 point above. If the discount stays flat it is Amber from Oct-26 and Red from Feb-27.

Average discount on tender contract lines by year (order date): 2021: 14.04%, 2022: 14.04%, 2023: 14.05%, 2024: 17.03%, 2025: 17.03%, Jan to Sep 2026: 17.03%.

If 'big deal' meant something else (Jan to Sep 2026): tender contract and distributor lines 16.78%; the biggest 1% of lines by value ($87,101 or more) 4.14%. The baseline depends heavily on the definition, so Sharon's size rule (open question) matters.

Revenue growth on the year (K3)

• Result of the check on Gary's draft: Confirmed
• Plain English: Net sales shipped this year to date over net sales shipped in the same months last year, minus one.
• Calculation: SUM(net_value_aud, ship month 1..m, 2026) / SUM(net_value_aud, ship month 1..m, 2025) - 1
• Sources and why: sales_order_lines.net_value_aud and .actual_ship_date (the draft rule, until finance confirms its revenue rule); .order_type checked.
• Grain: Order line
• Filters: Shipped lines only; year to date against year to date.
• Baseline: -6.8% ($149.36m against $160.26m). Full year 2025 on 2024 was +0.4%, which matches 'flat last year'. Period: January to September 2026 against January to September 2025, ship date. Evidence: Observed Fact
• Pace needed: +0.72 points a month: from -6.8% to +4.0% over 15 months (Sep-26 to Dec-27). In v1.1 the 4% is measured on 2026 full-year actuals: about $203.5m in 2027 ($16.96m a month) against $16.6m a month so far in 2026
• Run rate: +0.27 points a month over the last 3 months (year-to-date growth moved from -7.6% in Jun to -6.8% in Sep)
• Reading it: Off pace: the run rate is about 2.7 times too slow. Even on the 2026 base, 2027 would be about 3.1% below 2025
• Counter-metric: Value-weighted average discount across all lines (Average discount, K2 context): growth bought with price shows here. All lines Jan to Sep 2026: 12.52%, against 12.84% for the full year 2025
• Notes: Confirmed on ship date. Numbers are unchanged from v1.0. Checked order types: there are no credit or return types; warranty replacement lines carry full list value at 0% discount ($1.8m in 2025) and removing them changes the baseline to -6.9%, so it is immaterial. On order date the baseline is -9.9%. 1,517 lines (1.26%, all 2026) have no ship date yet and are not counted. The 2027 measure cannot exist before 2027, so the 2026 year-to-date figure is the baseline. v1.1 sets the base for the 4% as 2026 full-year actuals (estimate $195.68m until the year closes); the base is still an open question for Sharon and finance (C6).

Year-to-date growth by month in 2026 (ship date): Jan -12.5%, Feb -9.1%, Mar -8.7%, Apr -7.8%, May -8.2%, Jun -7.6%, Jul -7.8%, Aug -7.3%, Sep -6.8%.

What 4% means on the v1.1 base (2026 full-year actuals; estimate, Working Hypothesis): full-year 2026 is estimated at $195.68m (Jan to Sep actual plus Q4 2025 moved by the year-to-date rate). 4% on that is $203.5m for 2027, which is -3.1% against full-year 2025 ($209.95m).

Big tender win rate (K4)

• Result of the check on Gary's draft: Corrected
• Plain English: Tenders won out of the last 5 decided tenders with a contract value of $5m or more (rolling five-tender window), with the calendar-year count alongside.
• Calculation: COUNT(outcome = won) over the last 5 tenders with outcome IN (won, lost) and contract value >= 5,000,000, ordered by close_date
• Sources and why: tender_register.outcome, .estimated_value, .contract_length, .close_date, .tender_reference. The bid manager's file, because the sales leader does not trust the CRM for win-loss.
• Grain: Tender
• Filters: Contract value $5m or more (annual amounts times years); exclude undecided, partial and cancelled; show the count next to the rate.
• Baseline: 2 of the last 5 big tenders (40%). Calendar year alongside: 2024 1 of 2, 2025 0 of 2, 2026 to Sep 1 of 1. Won Coastal Ambulance Service ($14.0m) and Capital Region Health ($5.6m); lost Southern Metro Health Network ($6.5m), Western Plains LHD ($7.2m) and Highlands Ambulance ($11.8m). Period: The last five tenders of $5m or more in the register, closing March 2024 to February 2026. Evidence: Observed Fact (small numbers)
• Pace needed: Target open (Q-1): Deb Harlow sets it. Against Gary's starting proposal of at least 3 of the last 5 by Dec-27, the next 2 big tenders must both be won: after one win the window still reads 2 of 5, after two it reads 3 of 5. At about 2 big tenders a year, about 2.5 are expected by Dec-27
• Run rate: Not meaningful: only 5 events, so no run rate over 3 periods
• Reading it: Not rated until Deb sets the target. If 3 of 5 is accepted it is reachable only if both of the next two big tenders are won, with little room for a loss
• Counter-metric: Number and value of big tenders bid, and the discount on tender contract orders (Average discount, K2): a count can be lifted by bidding only on easy tenders or cutting price. Big tenders bid: 2 (2024), 2 (2025), 1 (2026 to Sep)
• Notes: Corrected (C7, accepted in v1.1): 24 of 64 estimated values are stored as annual amounts ('$2.4m pa'), which must be multiplied by the contract length to compare with $5m. The same five tenders qualify either way. No partial or cancelled tender is $5m or more. v1.1 measures it as a count over the last 5 big tenders, not a yearly percentage (C2). At $4m or more the count is 5 of 10; across all sizes 28 of 53, so the result depends on the size line. Customer names vary in the register (Southern Metro and Southern Metropolitan; Western Plains LHD and Western Plains Health District).

What the count needs (Working Hypothesis, Gary's starting proposal of at least 3 of the last 5 by Dec-27): the next 2 big tenders both won. After the next win the window reads 2 of 5 (the March 2024 win drops out), and after the next two it reads 3 of 5. About 2.5 big tenders are expected by Dec-27 at the recent pace of 2 a year.

Size line sensitivity (won of decided): $5m 2 of 5, $4m 5 of 10, $3m 8 of 13, $2m 13 of 20, $1m 19 of 35, $0m 28 of 53

Monthly reports that match the agreed set of numbers (K5)

• Result of the check on Gary's draft: Confirmed (Phase 2 KPI: placeholder tile on the dashboard)
• Plain English: Phase 2: reports that carry a sales number and whose monthly net sales check against the governed sales table, over all reports that carry a sales number.
• Calculation: COUNT(reports whose month net sales agree with the governed sales table within tolerance) / COUNT(reports that carry a sales number)
• Sources and why: Phase 2: the monthly check results (report, month, report net sales, governed net sales, pass or fail), built by the data engineer from the 30 workbooks and the governed sales table. First pass stand-in only: monthly_reports_index.workbook, .sales_definition, .customer_grouping, .last_refreshed.
• Grain: Report, month
• Filters: Agreed definition and match tolerance set by finance (open); reports marked n/a reviewed one by one to see whether they carry a sales number. Placeholder tile until one full month's check has run.
• Baseline: Phase 2 placeholder: waiting for the agreed definitions from finance and the first monthly number check. No baseline can be calculated yet: the index records what owners say they use, not whether their numbers agree. Period: Phase 2. Evidence: Working Hypothesis (the agreed set is open)
• Pace needed: Not applicable until Phase 2: the placeholder shows no pace needed. Target open (C3), set by the finance report owner with the sales leader once the agreed definitions exist
• Run rate: Not available: the index holds one snapshot, no history
• Reading it: Phase 2: waiting for data
• Counter-metric: Reports refreshed in the last 30 days, and the total number of reports that carry a sales number: the rate could be lifted by retiring reports or relabelling the index. 22 of 30 were refreshed in the 30 days to 30-Sep-26; 8 were last refreshed in March 2026 or earlier
• Notes: Phase 2 KPI. v1.1 changes the measure from the self-reported index to Tom's check of each report's monthly net sales against a governed sales table (C3 resolved by the change; target still open). First pass: placeholder tile plus the stand-in Reports on the draft agreed set. Index facts (Sample data): sales definitions ship date 14, invoice date 5, order date 5, n/a 6; customer grouping n/a 9, parent where known 6, ship-to 6, manual mapping tab 5, sold-to 4. 12 reports have n/a in at least one column; 17 different combinations are in use.
• Stand-in (first pass only, not the KPI): 3 of 30 reports (10%) say they use the draft rule (ship date, customers grouped by parent where known); 17 different combinations of sales definition and customer grouping are in use. Shows what reports say they use, not whether their numbers agree. Period: Report index at 30-Sep-26 (Sample data). Evidence: Working Hypothesis (the agreed set is open and the index is self-reported). Calculation: COUNT(sales_definition = ship date AND customer_grouping = parent where known) / COUNT(reports), plus COUNT(DISTINCT sales_definition, customer_grouping)

Stand-in detail (Sample data): reports matching each candidate set (with parent where known): invoice date + parent where known 1, ship date + parent where known 3, order date + parent where known 1. Most common combinations: ship date with ship-to: 5; ship date with n/a: 4; n/a with n/a: 3. Reports on the draft set: 08 Work orders by tech, 12 Tender pipeline, 25 Rep targets. Reports last refreshed more than 30 days before the cutoff: 02 Sales by product group, 05 Open quotes, 09 Overdue PMs, 10 Service revenue, 13 Ambulance accounts, 23 Installed base (service), 25 Rep targets, 29 Quarterly review pack.

Sub-KPIs

Gross margin by product group (K1.1)

• Result of the check on Gary's draft: Confirmed (Phase 2 KPI: placeholder tile on the dashboard)
• Plain English: Gross margin per cent, split by product group.
• Calculation: K1 grouped by material_group
• Sources and why: As Gross margin per cent (K1), plus sales_order_lines.material_group
• Grain: Order line by product group
• Filters: As K1. Placeholder tile until the cost feed switches on; the groups must add up to the total.
• Baseline: Phase 2 placeholder: waiting for cost of sales from finance. No baseline can be calculated: no cost of sales field. Period: Phase 2. Evidence: Working Hypothesis
• Pace needed: Not applicable until Phase 2: the placeholder shows no pace needed
• Run rate: Not available
• Reading it: Phase 2: waiting for data. Switches on with Gross margin per cent (K1)
• Counter-metric: As Gross margin per cent (K1)
• Notes: Phase 2 KPI. Formula confirmed; data mapping blocked by the same missing cost of sales field (C1). First pass: placeholder tile plus the stand-in Net price realisation by product group.
• Stand-in (first pass only, not the KPI): Loading systems 82.91% (+0.06 pts on 2025, 8.4% of net sales); Powered stretchers 82.95% (+0.01 pts on 2025, 9.1% of net sales); Stair chairs 83.10% (-0.08 pts on 2025, 2.8% of net sales); Manual stretchers 83.12% (-0.15 pts on 2025, 3.8% of net sales); Defibrillators 84.17% (+0.12 pts on 2025, 19.6% of net sales); Beds 84.33% (-0.34 pts on 2025, 21.8% of net sales); ED stretchers 84.45% (+0.30 pts on 2025, 7.1% of net sales); Defibrillator consumables 95.00% (-0.00 pts on 2025, 8.7% of net sales); Parts 100.00% (+0.00 pts on 2025, 4.4% of net sales); Service 100.00% (+0.00 pts on 2025, 14.2% of net sales). Service and Parts show 100% because they are not discounted: that says nothing about their margin. Groups are not ranked best or worst. Period: January to September 2026 against the same months of 2025, ship date (Sample data). Evidence: Observed Fact for the calculation; Working Hypothesis that group margin follows group realisation. Calculation: Net price realisation grouped by material_group, with each group's share of net sales

Stand-in detail: Net price realisation by product group, Jan to Sep 2026 (Sample data, not gross margin):

Revenue growth by product group (K3.1)

• Result of the check on Gary's draft: Confirmed
• Plain English: Revenue growth, split by product group.
• Calculation: K3 grouped by material_group
• Sources and why: sales_order_lines.material_group, .net_value_aud, .actual_ship_date
• Grain: Order line by product group
• Filters: As K3
• Baseline: Service -20.8%; Defibrillators -9.2%; ED stretchers -7.4%; Stair chairs -7.1%; Powered stretchers -6.3%; Loading systems -5.0%; Parts -4.3%; Manual stretchers -2.3%; Defibrillator consumables -0.1%; Beds +2.6%. Period: January to September 2026 against 2025, ship date. Evidence: Observed Fact
• Pace needed: Service +1.65; Defibrillators +0.88; ED stretchers +0.76; Stair chairs +0.74; Powered stretchers +0.69; Loading systems +0.60; Parts +0.56; Manual stretchers +0.42; Defibrillator consumables +0.28; Beds +0.09 points a month to reach +4.0% in 15 months
• Run rate: Service +0.31; Defibrillators +0.04; ED stretchers +0.65; Stair chairs +0.89; Powered stretchers +2.54; Loading systems +0.31; Parts -0.64; Manual stretchers +0.86; Defibrillator consumables +0.45; Beds -0.89 points a month
• Reading it: Only Beds is growing (above 0%). Service is lumpy: 64% of its 2025 ship-date sales fall in January and July, so its year-to-date swings are mostly billing timing
• Counter-metric: Rolls up to Revenue growth on the year (K3); watch the Average discount counter-metric
• Notes: Confirmed. Group names are tidied from the material_group codes.

Revenue growth by rep (K3.2)

• Result of the check on Gary's draft: Confirmed
• Plain English: Revenue growth, split by rep.
• Calculation: K3 grouped by rep_id
• Sources and why: sales_order_lines.rep_id, .net_value_aud, .actual_ship_date
• Grain: Order line by rep
• Filters: As K3 (parent grouping makes no difference to a split by rep)
• Baseline: R10 -17.5%; R07 -10.6%; R05 -9.8%; R01 -9.2%; R14 -8.7%; R09 -8.0%; R02 -6.8%; R06 -6.1%; R08 -5.5%; R04 -4.8%; R13 -4.5%; R11 -1.5%; R12 +0.1%; R03 +35.4%. Period: January to September 2026 against 2025, ship date. Evidence: Observed Fact
• Pace needed: R10 +1.44; R07 +0.97; R05 +0.92; R01 +0.88; R14 +0.85; R09 +0.80; R02 +0.72; R06 +0.68; R08 +0.63; R04 +0.58; R13 +0.56; R11 +0.36; R12 +0.26; R03 -2.09 points a month to reach +4.0% in 15 months
• Run rate: R10 +0.36; R07 -0.89; R05 +0.58; R01 +2.05; R14 +0.07; R09 +3.85; R02 -1.11; R06 -0.83; R08 +0.29; R04 -4.76; R13 +0.08; R11 +1.28; R12 -3.17; R03 +21.20 points a month
• Reading it: Only R03 (small base) and R12 (flat) are at or above 0%
• Counter-metric: Rolls up to Revenue growth on the year (K3)
• Notes: Confirmed on rep_id. Gary's parent-grouping filter does not change a split by rep, because the rep sits on the order line. Reps are not tied to accounts in the data (median 4 different reps per sold-to customer). R03's +35% is on a small base ($1.5m in 2025).

Our price against the winning price on lost tenders (K4.1)

• Result of the check on Gary's draft: Corrected (Phase 2 KPI: placeholder tile on the dashboard)
• Plain English: Our submitted price over the winning price, minus one, for lost tenders.
• Calculation: our_submitted_price / winning_price - 1, outcome = lost
• Sources and why: tender_register.our_submitted_price, .winning_price, .outcome
• Grain: Tender
• Filters: Lost tenders with a known winning price; annual amounts converted to contract totals. Placeholder tile until winning prices are back-filled; show the count with a known winning price beside the result.
• Baseline: Phase 2 placeholder: waiting for winning prices from the debrief letters. Winning price is recorded on only 5 of 25 lost tenders (2 of the 3 big losses), too few to show as a baseline. Period: Phase 2. Evidence: Working Hypothesis (winning price missing on 80% of lost tenders)
• Pace needed: Not applicable until Phase 2: the placeholder shows no pace needed (evidence measure, no target)
• Run rate: Not applicable
• Reading it: Phase 2: waiting for data
• Counter-metric: Price score gap from the debrief scoring table, as a second source (see Why we lost, K4.2): on the 3 big losses our price score was behind by 0.6 and 1.2 points of 30 and ahead by 0.9
• Notes: Phase 2 KPI. Corrected (C7, accepted in v1.1): values stored as annual amounts are converted to contract totals. Winning price is missing on 20 of 25 lost tenders (C8). What the register holds today (not shown on the tile): Southern Metro Health Network +2.1%, Highlands Ambulance -1.5%, Western Plains LHD not recorded; average +0.9% across the 5. First pass: placeholder tile plus the stand-in Price score gap on lost tenders.
• Stand-in (first pass only, not the KPI): Behind on price in 11 of 20 lost tenders with a scoring table; average gap -0.03 points of 30, median +0.45. Big losses: Southern Metro Health Network +0.60, Western Plains LHD +1.20, Highlands Ambulance -0.90. Positive means the preferred supplier scored higher on price. Price score, not price. Period: Lost tenders, whole register (Sample data). Evidence: Strong Inference (scores parsed from debrief text; only 3 tenders to cross-check against real prices). Calculation: Weighted price points (out of 30) the preferred supplier scored above Kestrel, lost tenders with a scoring table; positive = behind on price

What the register holds today (documented, not shown on the dashboard):

Stand-in detail: Price score gap on lost tenders (Sample data): weighted points out of 30 that the preferred supplier scored above Kestrel on price (positive means we were behind). | Big loss | Price score gap (points of 30) | Gap across the other five areas (points of 70) |

Across 20 scored losses: behind on price in 11, level or ahead in 9, behind by 1 point or more in 5; average gap on the other five areas +1.32 points.

Why we lost, from the debriefs (K4.2)

• Result of the check on Gary's draft: Corrected
• Plain English: For each lost tender, the scoring area with the biggest gap to the preferred supplier, counted by area.
• Calculation: argmax over areas of weighted points (preferred - ours), outcome = lost
• Sources and why: tender_register.scoring_table, .debrief_notes, .debrief_received
• Grain: Tender
• Filters: Lost tenders with a scoring table; text parsed into six areas.
• Baseline: Biggest gap to the preferred supplier, of 20 lost tenders with a scoring table: Service and support 7, Clinical fit 5, Price 3, Sustainability 3, Supply reliability 2. All 3 big losses: Service and support. Price was the biggest gap in only 3 of 20. Period: Lost tenders, whole register. Evidence: Strong Inference (scores from the buyer's debrief, parsed from text)
• Pace needed: No target (evidence measure)
• Run rate: Not applicable
• Reading it: Debrief received on all 25 lost tenders: Green on Gary's coverage threshold
• Counter-metric: Debrief coverage: debrief received on 25 of 25 lost tenders; scoring table present on 20 of 25
• Notes: Corrected: the scoring text comes in three layouts and scales, so gaps are compared in weighted points out of 100 (weights 30, 25, 15, 10, 10, 10), not raw scores (C9). Five lost tenders have no scoring table and are excluded and counted.

Gaps are weighted points out of 100 that the preferred supplier scored above us (negative means we scored higher).

Who is winning the tenders we lose (K4.3)

• Result of the check on Gary's draft: Confirmed
• Plain English: Count and value of lost tenders by winning supplier.
• Calculation: COUNT, SUM(contract value) grouped by winning_supplier, outcome = lost
• Sources and why: tender_register.winning_supplier, .estimated_value, .contract_length
• Grain: Tender
• Filters: Lost tenders.
• Baseline: Lost tenders by winner: Allied Clinical Supply 7 ($15.2m); Meridian Biomedical 6 ($7.5m); Helix Healthcare 5 ($13.2m); Vantage Medical 4 ($10.6m); Norland Rescue Systems 3 ($13.9m). Big losses: Vantage Medical, Allied Clinical Supply (a distributor bundling brands), Norland Rescue Systems. Period: Lost tenders, whole register. Evidence: Observed Fact
• Pace needed: No target (evidence measure)
• Run rate: Not applicable
• Reading it: Not applicable
• Counter-metric: Rolls up to Big tender win rate (K4)
• Notes: Confirmed. Supplier names are consistent in the register (5 names, none missing on lost tenders). Values are contract totals (annual amounts multiplied by term). The big losses went to three different winners, so no single competitor explains them.

3. KPI tree

Gross margin per cent (K1)
  Gross margin by product group (K1.1)
Average discount off list price on big deals (K2)
Revenue growth on the year (K3)
  Revenue growth by product group (K3.1)
  Revenue growth by rep (K3.2)
Big tender win rate (K4)
  Our price against the winning price on lost tenders (K4.1)
  Why we lost, from the debriefs (K4.2)
  Who is winning the tenders we lose (K4.3)
Monthly reports that match the agreed set of numbers (K5)

4. What Phase 2 needs

Each Phase 2 KPI stays Must Have with its name, target and owner. Kewpie works with a data engineer to source the data. All four are placeholders until the switch-on test passes.

5. Challenge notes

No KPI has been added, dropped or reprioritised. Corrections to Gary's draft measures are marked. 'Status in v1.1' follows the Business Requirements change log.

6. Data quality log

7. Open questions that stay with Gary and the stakeholders

• Big-deal size rule (sales leader). Confirm the discount path towards 14% (sales leader). Base for the 4% (sales leader and finance). Win-rate count target, needed before Stage 4 is approved (bid manager). Cost of sales source and reconciliation tolerance, needed to start Phase 2 (finance). Agreed definitions, tolerance and target for the reports check, needed before the Phase 2 switch-on (finance, with the sales leader). Who is the data engineer, and when can they start (Adam).

File: 2026.10.23 - Kestrel Medical - 01 Business Requirements.md

Kestrel Medical: Business Requirements

Test case, Sample data. Kestrel Medical is a test case. Sharon was played by the case's stakeholder agent. Data named here is mapped to the Sample data in 03 - Working Files/data/, which must be checked against real data before go-live.

1. Document control

2. Problem Statement

Kestrel has lost three of its last five big tenders, each over five million dollars, and believes price is the reason but can't yet prove or disprove it. With gross margin at about 41 per cent against global's 43 per cent target for 2027, and thirty monthly reports that give different answers, the business needs one set of evidence before it asks global for more discount room.

The problem as stated

Sharon, in her words: lost three of the last five big tenders, each over five million dollars, "never happened to me in 22 years". Reps say Kestrel is being undercut 10 to 15 per cent. She wants a win-loss view and pricing against the competition, so she can take evidence to global and argue for more discount room. And she wants out of Excel: thirty monthly reports, and half of them say different things.

The problem as Gary understands it

• The ask and the target pull against each other. The ask is evidence for more discount. Global's target is more margin (about 41 to 43 per cent in 2027) and 4 per cent revenue growth, after a flat year
• The loss reasons are belief, not yet evidence. They come from reps and one procurement person. Sharon's own account of the three losses: one "was price" (Southern Metropolitan Health Network), one was "about the stretcher as much as anything" (Highlands Ambulance, a 15-year customer), and one went to a distributor that bundles several brands (Western Plains Health District, won by Allied Clinical Supply)
• The evidence exists but hasn't been used. Deb Harlow, the bid manager, keeps the tender file and the debrief letters
• Nobody agrees on the base numbers. Thirty monthly reports from ERP extracts disagree, and Sharon spends three days a month reconciling them

Gary's note: the question I'd put to Sharon is "What decision changes if this number moves?" Here, the answer is the discount case to global. If the debriefs say Kestrel lost on product fit or bundling, more discount won't win the next tender and will cost the margin target. That's my reading, not Sharon's words, and Sharon has not confirmed it.

3. Objectives

Each objective helps solve the Problem Statement: margin and growth are global's tests (O1, O2), the tender losses are the trigger (O3), and one agreed set of numbers makes the evidence credible (O4).

4. Audience and decisions

5. Business Requirements with Data Mapping

All measures, data mappings and filters below were drafted by Gary and confirmed or corrected by Kewpie at Stage 2 (Calculation Pack v1.0). Targets and thresholds marked "Gary's proposal" were not said by Sharon. Four KPIs are Phase 2 KPIs: see "Phase for each KPI" below.

Top 5

SME sign-off (all rows): Sharon (stakeholder) confirmed the playback with corrections, 23-Oct-26. Adam (acting subject matter expert, test mode) approved at the Stage 1 gate, 23-Oct-26, and approved the v1.1 changes at the Stage 2 gate, 23-Oct-26.

Measures for the Top 5 (confirmed or corrected by Kewpie)

Sub-KPIs (Must Have, explain a Top 5 measure)

Phase for each KPI

This first pass builds every Top 5 KPI and sub-KPI that the data supports today. Four can't be built yet because the data doesn't exist or isn't complete. They stay Must Have, keep their name, target and owner, and show as placeholder tiles until Phase 2, when Kewpie works with a data engineer to source the data. Every other KPI is marked 'First pass: built': Average discount off list price on big deals (K2), Revenue growth on the year (K3), Revenue growth by product group (K3.1), Revenue growth by rep (K3.2), Big tender win rate (K4), Why we lost, from the debriefs (K4.2) and Who is winning the tenders we lose (K4.3).

Stand-ins for the first pass (not KPIs)

A stand-in is a small, separate panel beside a placeholder tile. It carries its own plain name and the labels 'Stand-in, not the KPI' and 'Sample data' (until real data is loaded). It never takes the KPI's name, target, thresholds, pace needed, status colour or finance's stated 41 per cent. It has no target and no status colour of its own. It is removed, or moved to a context note, when its KPI switches on in Phase 2.

Adam chose, at the Stage 2 gate on 23-Oct-26, to show the Net price realisation stand-in beside the Gross margin per cent (K1) placeholder, with its labels, rather than the placeholder alone.

Placeholder tiles

A placeholder tile uses the KPI's exact name from these Business Requirements, its owner's role, and the text 'Phase 2: waiting for data', with one line on what's missing (for example 'Waiting for cost of sales from finance'). The headline slot shows no number, no status colour, no sparkline, no gap and no pace needed. The target may be shown as text ('Target: 43 per cent for 2027') so the reader knows what's coming. A placeholder tile sits in the same place the real tile will take, so the layout doesn't change at switch-on.

How validation treats Phase 2 KPIs

For Jennifer's validation checklist in this first pass:

1. A Phase 2 KPI counts as 'on the dashboard, calculated as documented' when its placeholder tile and stand-in are shown exactly as these Business Requirements document them, and is recorded as 'Phase 2 placeholder, as documented', not a fail
2. For Lift gross margin (O1), progress shows Average discount off list price on big deals (K2) with actual, target, gap and pace, and the margin target as text with 'Phase 2: waiting for data'. For One agreed set of numbers (O4), progress shows 'Phase 2: waiting for data'. Neither is a fail
3. Each Phase 2 owner can see 'Phase 2: waiting for data' at a glance, which counts as their status

It is a fail if a placeholder shows a number, a status colour, a sparkline or a pace needed, or if a stand-in takes the KPI's name, target, status colour or the stated 41 per cent. Kewpie's Stage 2 check, 'documented calculation that runs on the data', is met for these four by a documented calculation plus a named Phase 2 data source. Gary's proposal, approved by Adam; the process file itself is unchanged, which is Adam's call.

6. Backlog (Nice to Have)

7. Data inventory (as named in the interview)

Sharon said she isn't the one to ask about the data itself. Tom Nguyen (finance) and Deb Harlow (bid manager) hold the detail.

8. Constraints and chart preferences

• Must speak to how Kestrel's market actually works, not global's standard view
• Reports automated and consistent
• A pack for global first, then a screen in meetings with the reps. Not designed for phone
• Privacy and customer confidentiality: unknown, Sharon to check
• Chart likes and dislikes: none given. Sharon will react to a draft

9. Assumptions

10. Open questions

11. Change log

File: 2026.10.23 - Kestrel Medical - 01 Project Plan.md

Kestrel Medical: Project Plan

Test case, Sample data. Built from Business Requirements v1.1 (approved at the Stage 2 gate, 23-Oct-26). Dates and effort are Gary's proposals. The first pass covers phases 1 to 5; Phase 2, Source missing data, follows it.

1. Goal and scope

• Goal: a first trusted dashboard Sharon can take to global as evidence and use in rep meetings, covering gross margin, revenue growth, big tender wins and one agreed set of numbers
• In scope: the Top 5 measures and their six sub-KPIs in Business Requirements v1.1. First pass: built where the data supports it, placeholder tiles with labelled stand-ins for Gross margin per cent, Gross margin by product group, Monthly reports that match the agreed set of numbers, and Our price against the winning price on lost tenders. Phase 2: source the missing data and switch those four on
• Out of scope this time: CRM win rate, reps' reported competitor price cuts, service and field measures

2. Phases and sprints

The data adds a sprint to Build KPIs: cost of sales isn't named yet, the reports disagree, and the tender file stores values as text. Four KPIs can't be sourced in the first pass and move to Phase 2. The finance report owner and bid manager can start their parts (naming the cost source, agreeing definitions, back-filling winning prices) during the first pass, so Phase 2 starts ready.

Phase 2 tasks

3. MoSCoW

4. Effort estimate

5. RAID log

6. Milestones and approval gates

7. Review rhythm after go-live

• Monthly: sales leader reviews the Top 5, in line with today's monthly reports (Sharon's expectation, to confirm)
• After each big tender decision: bid manager updates the outcome, winning price and debrief
• Before each global review: pack produced from the dashboard (date and format open)
• Monthly, from Phase 2: the data engineer's automated check runs each report against the governed sales table; the finance report owner reviews the failures. Until then, quarterly: the finance report owner checks the monthly reports against the agreed set of numbers

File: 2026.10.23 - Kestrel Medical - 02 Kewpie Substitutes for Missing KPIs.md

Kestrel Medical: Substitutes for KPIs That Can't Be Built Yet

Sample data. Every number is calculated from the Sample data in 03 - Working Files/data/ by 03 - Working Files/kewpie_substitutes.py (results in substitutes_results.json). Data runs to 2026-09-30. Nothing here is real Kestrel data. Nothing in the Calculation Pack, calc_kpis.py, spec.json or the Business Requirements has been changed. No KPI is added, dropped or reprioritised: each real KPI stays in the Top 5 and gets a placeholder tile marked for Phase 2. The substitute sits beside the placeholder as a clearly labelled stand-in, never in place of the real KPI's name or target.

Author: Kewpie, KPI Creator Stage 2. Date: 23-Oct-26.

Summary

Everything else in the Top 5 and the sub-KPIs can be built now (section 3), so no placeholder is proposed for them.

1. What I propose

• Placeholder tile for each of the four KPIs above, named exactly as in the Business Requirements, with the text "Phase 2: waiting for data" and the owner's role. No status colour, no number in the headline slot.
• Substitute shown as a small, separate stand-in, titled with its own plain name, labelled "Sample data" and "Stand-in, not the KPI". It must never carry the real KPI's name, target, status colour or the sales leader's stated 41%.
• Honest limits up front. For Gross margin per cent (K1) and Gross margin by product group (K1.1) the substitute is a price-side signal only. For the other two it is a closer relative of the real KPI.
• Phase 2 gate. Each placeholder turns into the real KPI only when the data engineer's feed reconciles to the owner's figure (for K1, to finance's 41%).

2. The four KPIs, one by one

Gross margin per cent (K1)

• Why it can't be built now: no cost, COGS, margin or standard cost field in any of the 11 files. The stated 41% can't be reproduced.
• Substitute (plain name): Net price realisation.
• Calculation: SUM(net_value_aud) / SUM(list_price * qty), shipped lines, by ship date, January to September.
• Source fields: sales_order_lines.net_value_aud, .list_price, .qty, .actual_ship_date.
• Baseline (Sample data): 87.36% for January to September 2026 (net sales $149.36m). Same months of 2025: 87.67%. By full year: 2021 89.72%, 2022 89.28%, 2023 89.28%, 2024 87.61%, 2025 87.22%. Excluding warranty replacement lines (full list value, $1.39m in 2026 to date): 87.26% against 87.58%. Monthly 2026: Jan 89.70, Feb 85.52, Mar 87.37, Apr 85.62, May 86.14, Jun 87.37, Jul 89.35, Aug 86.02, Sep 86.89.
• Evidence label: Observed Fact for the calculation. Working Hypothesis that it moves with gross margin, because there is no cost data to test the link.
• What it tells Adam: how much of list price Kestrel keeps after discounting. The step down in 2024 (89.28% to 87.61%) lines up with the 3-point rise in tender discount in the Calculation Pack (challenge note C5). A fall in realisation will pull margin down unless cost falls too.
• What it doesn't tell Adam: gross margin. It ignores cost of sales entirely, so it can't say whether margin is 41%, 43% or anything else, and it can't be set against the 43% target or given a pace needed.
• How it could mislead: (1) List prices can rise or fall, so realisation can move with no change in margin. (2) Product mix: Service, Parts and consumables carry little or no discount, so a shift in mix moves the total (Service fell from 16.7% to 14.2% of net sales, which alone nudges realisation down) without any price decision. (3) Realisation can be flat while cost rises. (4) Reading a rise as a margin gain would be wrong.
• Recommendation: show it as a stand-in beside the placeholder. Never as a margin proxy with a target.

Gross margin by product group (K1.1)

• Why it can't be built now: same missing cost of sales.
• Substitute (plain name): Net price realisation by product group, with each group's share of net sales.
• Calculation: as above, grouped by material_group.
• Source fields: as above plus sales_order_lines.material_group.
• Baseline (Sample data, Jan to Sep 2026, change from the same months of 2025 in points, share of net sales in brackets): Beds 84.33% (-0.34, 21.8%); Defibrillators 84.17% (+0.12, 19.6%); Defibrillator consumables 95.00% (0.00, 8.7%); ED stretchers 84.45% (+0.30, 7.1%); Loading systems 82.91% (+0.06, 8.4%); Stair chairs 83.10% (-0.08, 2.8%); Manual stretchers 83.12% (-0.15, 3.8%); Powered stretchers 82.95% (+0.01, 9.1%); Parts 100.00% (0.00, 4.4%); Service 100.00% (0.00, 14.2%).
• Evidence label: Observed Fact for the calculation. Working Hypothesis that group margin follows group realisation.
• What it tells Adam: which groups carry deep discounts. The equipment groups all keep about 83% to 84% of list, so no single group stands out on price. Within-group realisation barely moved (all within 0.4 points), so the small fall in total realisation is mostly mix, not price.
• What it doesn't tell Adam: which groups make money. Service and Parts at 100% of list say nothing about their margin (service cost is labour, parts cost is purchase price), and equipment groups differ in cost.
• How it could mislead: the 100% for Service and Parts looks like "best" but is only "no discounting". Ranking groups by realisation would send attention to the wrong place.
• Recommendation: show as a stand-in table beside the placeholder, with the note on Service and Parts.

Monthly reports that match the agreed set of numbers (K5)

• Why it can't be built now: finance hasn't defined the agreed set, so "match" has no test. The index holds what each owner says the report uses, one snapshot, no history, so no run rate.
• Substitute (plain name): Reports on the draft agreed set, with the number of different combinations in use.
• Calculation: COUNT(sales_definition = ship date AND customer_grouping = parent where known) / COUNT(reports), plus COUNT(DISTINCT sales_definition, customer_grouping). The draft set is the rule used for Revenue growth on the year (K3), labelled "draft" on the tile.
• Source fields: monthly_reports_index.workbook, .sales_definition, .customer_grouping, .last_refreshed.
• Baseline (Sample data, index at 30-Sep-26): 3 of 30 reports (10%): 08 Work orders by tech, 12 Tender pipeline, 25 Rep targets. 17 different combinations in use. 12 reports have n/a in at least one field. 14 of 30 use ship date at all. 22 of 30 refreshed in the last 30 days.
• Evidence label: Working Hypothesis (the agreed set is open, and the index is self-reported).
• What it tells Adam: how far the reports are from one rule, if the draft rule is chosen. It shows the scale of the clean-up: 17 combinations is nowhere near one agreed set.
• What it doesn't tell Adam: whether any report's numbers actually agree. A report can say "ship date" and still be wrong.
• How it could mislead: (1) If finance picks a different agreed set (for example invoice date), the 3 changes. Each alternative candidate gives 1 of 30. (2) n/a reports might not carry sales numbers at all, so 3 of 30 may understate the real position. (3) Retiring reports lifts the percentage without fixing anything. (4) Owners can relabel the index.
• Recommendation: show as a stand-in labelled "against a draft set". Don't set a target until finance defines the set.

Our price against the winning price on lost tenders (K4.1)

• Why it can't be built now: winning price is recorded on 5 of 25 lost tenders (20%), and 2 of the 3 big losses.
• Substitute (plain name): Price score gap on lost tenders.
• Calculation: for each lost tender with a scoring table, (preferred supplier's price score - Kestrel's price score) in weighted points out of 30, across the three scoring layouts (same parser as the debrief KPI, Why we lost K4.2). Positive means Kestrel was behind on price.
• Source fields: tender_register.scoring_table, .outcome, .estimated_value, .contract_length.
• Baseline (Sample data, whole register): 20 of 25 lost tenders have a scoring table. Kestrel was behind on price in 11 of 20 and level or ahead in 9. Behind by 1 point or more in 5. Average gap -0.03 points, median +0.45, range -2.4 to +2.1. Big losses: Southern Metro Health Network +0.60, Western Plains LHD +1.20, Highlands Ambulance -0.90. On the same lost tenders the average gap across the other five areas is +1.32 points.
• Cross-check against the real KPI: the 3 tenders that have both a winning price and a scoring table agree in direction (our price +1.9% and +2.1% above the winner with price score behind by 0.9 and 0.6; our price 1.5% below with the price score ahead by 0.9). Three points is too few to call it a validation.
• Evidence label: Strong Inference (scores come from the buyer's debrief, parsed from text; the cross-check has only 3 tenders).
• What it tells Adam: whether price was where we lost ground in the buyer's own scoring. It supports the Calculation Pack note that the three big losses were not mainly on price (challenge note C10).
• What it doesn't tell Adam: the dollar price gap. A score is the buyer's view of price, shaped by its scoring formula, not our price against the winner's price.
• How it could mislead: a small score gap can hide a large dollar gap, and the reverse. Five lost tenders have no table. Buyers differ in how they score price.
• Recommendation: use as the first-pass substitute. It is a stronger stand-in than the 5 known winning prices, which would show an average of +0.9% from a sample too small to trust.

3. Checked and buildable now (no placeholder proposed)

4. What Phase 2 needs

5. Assumptions and open points

• I assumed "build properly" means calculated from a source Adam can trace and defend, so a stated figure (finance's 41%) or a self-reported index doesn't qualify.
• I treated the real KPI's name, target and status colour as untouchable. A substitute only fills the space beside a placeholder.
• Gary needs to draft the Business Requirements change for Adam's yes (placeholder display rule, Phase 2 mark, the stand-in text). I haven't touched them.
• If Adam prefers no stand-in for Gross margin per cent (K1) and by product group (K1.1), because the price-side signal risks being read as margin, the alternative is the placeholder only. I lean to show it with the labels above: it gives the sales leader something real about discount pressure, and it is the part of margin the team can influence.

File: 2026.10.23 - Kestrel Medical - 02 Gary Phase 2 Changes.md

Kestrel Medical: Gary's Phase 2 Changes

Test case, Sample data. Proposals only, for Adam's yes. Nothing in the Business Requirements v1.0, the Project Plan or spec.json has been changed, and Sharon hasn't been contacted.

Adam asked for placeholders for the KPIs that can't be built in this first pass, marked for Phase 2, where Kewpie works with a data engineer to source the data. Four KPIs are affected: Gross margin per cent (K1), Gross margin by product group (K1.1), Monthly reports that match the agreed set of numbers (K5) and Our price against the winning price on lost tenders (K4.1). None is dropped or reprioritised: each keeps its name, target and owner, and gets a placeholder tile now and the real calculation in Phase 2. Each placeholder sits beside a small stand-in that Kewpie calculated. A stand-in is labelled "Stand-in, not the KPI" and never takes the KPI's name, target, status colour or finance's stated 41 per cent. This file merges four of my earlier 11 rows (rows 1, 4, 5 and 9 in "02 Gary Response to Challenge Notes"), so Adam reads one consistent set. My other seven earlier rows (2, 3, 6, 7, 8, 10, 11) stand as written.

1. Business Requirements: Before and After (proposed v1.1)

"(none)" in Before means a new item. Quotes in Before are from Business Requirements v1.0, word for word.

How my earlier 11 rows fit:

2. Project Plan: Before and After

3. Decisions for Adam

1. Yes to the Business Requirements rows (1 to 29) and Project Plan rows (1 to 14) above, so I can write v1.1 and update spec.json?
2. Kewpie's option: show the margin stand-in (Net price realisation) or the placeholder only. I lean to showing it with the labels in row 8, because it's the part of margin the sales team can move. If the pack for global is the main use, the risk in R5 is real, so it's your call.

File: 2026.10.23 - Kestrel Medical - 02 Gary Response to Challenge Notes.md

Kestrel Medical: Gary's Response to Kewpie's Challenge Notes

Test case, Sample data. Gary's recommendations for Adam's SME sign-off on the 11 challenge notes in Calculation Pack v1.0. Nothing has been changed yet, and Sharon hasn't been contacted. Every Before and After below is a proposal, applied to Business Requirements v1.1 and spec.json only once Adam says yes.

Recommendation on approving Stage 2

Approve Stage 2 with conditions. Gross margin per cent (K1) and Gross margin by product group (K1.1) are documented but can't be run. The targets for Big tender win rate (K4) and Monthly reports that match the agreed set of numbers (K5) are still open. None of these stops Viktor building the story and the tiles. They do stop Jennifer's validation ("calculated as documented") and go-live. So:

• Close all three before Stage 4 is approved, not before Stage 3 starts
• Until then, the Gross margin per cent (K1) tile shows "Waiting for cost of sales from finance". The stated 41 per cent appears as a Working Hypothesis, with no status colour and never as a calculated number

One line per note

C10: the data doesn't support "lost on price"

• What it shows (Strong Inference, Sample data, 3 tenders): service and support was the biggest scoring gap on all three big losses. On Highlands Ambulance we were 1.5 per cent cheaper than the winner. Across 20 scored losses, price was the biggest gap in only 3
• Does it change the requirements? No. Sharon's confirmed words were that she wants "the evidence to prove or disprove it". The Problem Statement and the tender measures were built to answer exactly that, and they have. Changing the statement now would put words in her mouth
• What it does change:
  • Viktor's story should lead with it, because it's the answer Sharon asked for, even though it isn't the answer she expected
  • It backs the tighter discount target proposed for C5. More discount room may cost margin and still not win the tender
  • The Project Plan risk "The evidence may show losses weren't mainly price" (R2) has now happened. Sharon needs to hear it from Adam at the walk-through, not find it on a tile
  • New open question for Sharon: "The debriefs point to service and support. Do you want a service measure under Big tender win rate (K4) in the next release?" I wouldn't add one without her. She raised no service objective, and that's her call to make

Before and After (proposed Business Requirements v1.1)

New or updated open questions

1. For the bid manager (Deb Harlow): What count out of the last 5 big tenders is a realistic target? Can you add winning prices from the debrief letters, starting with Western Plains?
2. For finance (Tom Nguyen): Where does cost of sales come from? What is the agreed sales definition and customer grouping? Which reports carry a sales number? What tolerance counts as a match?
3. For Sharon: Is the tighter discount target right (back towards about 14 per cent)? Is the 4 per cent on 2026 actuals? How should reps be credited? Do you want a service measure, given what the debriefs show?

File: dataset_description.md

Dataset description (Sample data)

Sample data for Kestrel Medical, copied unchanged from the test bank. Data to 2026-09-30. Not real company data.

No file has a cost of sales field. Re-run: python calc_kpis.py then python build_pack.py (needs pandas).