Supply Promise Anchor Audit

1. Executive summary

The three findings that matter most:

1. The promise dates are not being maintained, and that — not vendor performance — is the dominant source of anchor error. Four vendors (ACOM Business Solutions, Micro Shop, Cable Plus Distributors, The Apple Store) show an identical, mechanically increasing lateness pattern: their PO lines carry a frozen expected receipt date of 2021-08-05 while POs continued to be cut weekly, so measured "lateness" climbs in exact 7-day steps from 0 to 114 days. Actual receiving cadence was consistently 0–7 days after order. The anchors were wrong because nobody updated them, not because the vendors slipped. Harris Technology shows the mirror image — a padded 2021-11-30 promise while goods arrived within a week, producing up to 388 days of "earliness."
2. Every open purchase order line in the account is catastrophically overdue against its recorded promise. All 36 open PO lines ($76.2K) promise receipt in Aug/Nov 2021 — 4.7 to 5 years ago. Any commitment or availability calculation anchored to these dates is fiction. They should be re-dated or closed.
3. The largest vendor by received value is entirely unmeasured. Bobbi Brown accounts for $1,544,717 of received value across 151 lines — 41% of all receipt value — and none of those lines carried an expected receipt date. That vendor is not "fine"; it is invisible. It is graded Unknown and must be treated as unmeasured risk, not as reliable.

2. Anchor scorecard MEASURED

Click a column header to sort. Variance = actual receipt date − promised date (negative = early). On-time = arrival on or before the promise. Confidence basis = receipt lines with a promise date. Fill rate is quantity received ÷ ordered on receipt-linked lines. Tolerance bands: strict (≤0 days), ≤+3 days, and |±3| days.

Grading rule applied: Reliable = |mean| ≤ 3d and σ ≤ 5d; Biased-but-predictable = σ ≤ 10d regardless of mean; Volatile = σ > 10d; Unknown = <10 lines with a promise date. No vendor met the Reliable or Biased thresholds. Note that Harris Technology's 93.5% "on-time" rate is an artifact of promise padding — it delivered up to 388 days before the promise, which is a broken anchor, not good performance; the ±3-day band (9.7%) is the honest number. Fill-rate caveat: every receipt-linked line in history was received in full (no partials/over-receipts found), but that also reflects how this account's receipts were entered.

3. Volatility ranking

Ranked by standard deviation of signed variance — these are the vendors whose anchors destabilize commitments most. MAD shown as the outlier-robust check.

4. Lead time drift observed vs configured: none

There is no configured purchase lead time anywhere in this account: item.leadtime is null on all 553 items, and all 58 itemvendor records have empty predicteddays/schedule. A drift comparison is therefore impossible — which is itself the finding: replenishment and expected-availability logic is running with a zero/absent lead-time baseline, and observed reality is 3–10 days. Suggested initial values from observed medians:

The prompt's >20% understatement flag technically applies to every item (configured 0 vs observed 7), but presenting 553 rows of the same absent-baseline finding adds no information; the vendor-level view above is the actionable version.

5. Exposure register INFERRED item-level match

This account does not have the Supply Allocation feature (feature check: not enabled), and no order-allocation/pegging link rows exist between sales order lines and purchase orders. All supply-to-demand connections below are inferred by matching item (location was ignored because open SO lines mostly carry no location). Every figure in this section inherits that caveat.

All 36 open PO lines are 4.7–5 years past promise. A "risk-adjusted arrival" (promise + vendor P90 lateness of ~100 days) still lands in 2021 — meaning the risk adjustment is moot: the supply anchor for every one of these items is simply broken, and 100% of the open demand on these items ($48,698.90) ships — or was supposed to ship — before any credible arrival of the linked supply. $38,353.90 of it has a ship date already in the past.

At-risk sales orders (highest value first)

Exposure roll-ups INFERRED

Committed vs backordered: 6 of 9 at-risk lines are backordered (waiting on future supply); 3 lines show committed quantity, meaning on-hand stock is reserved for them — those are at lower physical risk but their ship dates are already 8–23 months past, so the commitment itself is stale. Note also the demand/supply mismatch: SO 277 alone needs 20 units of item 112 against only 3 open incoming units.

6. Overdue supply MEASURED

Every open PO line in the account is past its promise. 36 lines, $76,165.88, all dated 2021-11-27, promising receipt 2021-08-05 (1,818 days overdue — the promise even predates the order, itself a data-integrity flag) or 2021-11-30 (1,701 days overdue).

7. Recommended actions

A. Promise dates to renegotiate / re-date (vendor-facing)

1. All vendors, all 36 open PO lines (POs 195–206): the promises are 4.7–5 years stale. Confirm with each vendor whether the orders are still live; re-date to a real commitment or close the lines. Until then every availability view of these 18 items is anchored to fiction. Priority order by open value: Brocade ($33.1K), Harris ($18.5K), Micro Shop ($11.6K), Apple Store ($8.3K), ACOM ($4.6K), Cable Plus ($84).
2. Bobbi Brown: require an expected receipt date on every future PO line. $1.54M of historical volume with zero promise dates means your highest-volume supply stream is unmeasurable.
3. Harris Technology: stop padding promises. A vendor delivering ~350 days "early" against its recorded date is holding phantom future supply on the books that planning will defer against.

B. Item lead time settings to correct (system-facing)

1. Populate leadtime (or item-vendor schedules) for the 18 actively traded items using the observed medians in Section 4 (7 days for all measured vendors). Today the baseline is empty, so any reorder/expected-availability logic defaults to zero lead time — systematically optimistic.
2. Fix the buying-process defect that freezes expectedreceiptdate when POs are copied: the frozen 2021-08-05 date across four vendors' POs is the single largest distortion in this dataset.
3. Begin promise-date snapshotting (see Methodology) so future audits can measure drift, not just terminal variance.

C. Customer commitments to proactively re-communicate

1. Thomas Industry Supplies — SO 277 ($35,976 at risk, ship date 23 months past): 20 units of item 112 backordered against only 3 open incoming units from a vendor whose PO is 5 years stale. This order cannot be fulfilled as written; re-scope or re-source before the customer escalates.
2. Karma Sports and Outdoors — SO 342 ($9,490, ships 2026-08-15 — 18 days out): the nearest live commitment. Both items depend on inferred supply from stale POs (ACOM, Micro Shop); confirm stock or re-date now, before the ship date passes.
3. Thomas Industry Supplies — SOs 313, 296, 289, 279 ($2,378): all ship dates already past; either fulfill from committed stock (296/289/279 show commitments) and close, or re-date.
4. Vaibhav A — SO 357 ($855, ships 2026-12-28): lowest urgency; resolve the Apple Store supply question with the above.

8. Methodology and limitations