Skip to content
Atomic Bible
Current status
Research case · fully synthetic manual reconstruction

Demand, inventory, and advertising point in different directions.

A worked weekly decision brief that keeps unlike evidence bases, populations, interpretations, and conditional actions separate.

Fully synthetic reconstruction. No private report data.Every organization, person, product, identifier, date, and value in this case is invented for publication safety. Values were chosen for arithmetic coherence and were not scaled from a private report. No mapping key exists.
Marketplace Portfolio Alpha · weekly

Synthetic decision brief

Identity
AB-PUBLIC-CASE-001
Example period
Illustrative Sunday-to-Saturday week
Job
Monitor, explain, decide
Method
Manual synthetic reconstruction
Decision frame

Where should advertising, supply planning, and marketplace operations intervene next?

The synthetic scenario supports tighter ad allocation and two conditioned inventory moves. It does not support mixed-basis arithmetic or an inventory API-only release. Every proposed action remains subject to an explicit human approval gate.

Current frame

Four measures, four evidence bases

Candidate lenses: MEASURE-001, FRAME-001, VIS-COMPARE-001. Can explicit framing prevent false equivalence?

ERP net sales$62.4k

+30.0% vs $48.0k prior illustrative week

390 invoiced units · synthetic ERP basis · USD
Marketplace retail gross$53.2k

174 fulfilled + 126 merchant units

300 order units · synthetic Orders API · USD
Blended ACOS91.6%

$12.6k spend / $13.75k attributed sales

Synthetic weekly ad extract · advertised groups
Zero-stock candidates31 / 72

43.1% of inventory API records

Modeled recent demand · manifest check required

Integrity guard: Do not subtract, add, trend, or ratio the $53,200 marketplace retail-gross total against the $62,400 ERP net total. Event timing, recognition, adjustments, dollar basis, and unit populations differ. Ad-attributed sales overlap rather than forming another additive revenue base.

Metric and source map

Every denominator and comparison frame

Candidate lenses: PROVENANCE-001, VIS-INTEGRITY-001, NOTATION-001.

All values and populations are synthetic and illustrative.
MeasureValueDenominator and frameEvidence role
ERP net sales$62,400390 invoiced units; illustrative week; ERP basisObservation
Week-over-week net change+$14,400 / +30.0%Prior-week ERP net sales of $48,000Observation, descriptive only
Fulfilled-inventory share174 / 300 = 58.0%Marketplace order unitsObservation
Merchant-fulfilled share126 / 300 = 42.0%Same 300-unit denominatorObservation
Blended ACOS$12,600 / $13,750 = 91.6%Ad spend divided by attributed ad salesObservation
PX-101 spend concentration$5,400 / $12,600 = 42.9%Share of weekly ad spendObservation
Zero-stock API records31 / 72 = 43.1%Inventory records with modeled recent demandObservation with manifest caveat
Traffic match coverage148 / 180 = 82.2%Matched listings divided by portfolio listingsObservation
Reasoning chain

What is observed, inferred, and proposed

Candidate lenses: ARGUMENT-001, COMM-ANSWER-001, UNCERTAINTY-001.

  1. Observation

    Modeled demand remains concentrated on established listings, while stock is disproportionately present on newly introduced listings.

  2. Inference

    The mismatch supports testing a fulfillment-coverage constraint. It does not prove that stock placement caused the demand pattern.

  3. Recommendation

    Rebalance selected ad groups and consider conditioned inventory moves only after current receiving manifests are reconciled and a human approves release.

Signature comparison

Stock placement by listing cohort

Each collection's total listings are the denominator for both stocked counts.

Synthetic listing-status records joined to modeled trailing demand.
CollectionEstablished stocked / totalNew or bundle stocked / totalRead
Collection A0 / 4214 / 4233.3% stocked, all on new listings
Collection B11 / 365 / 3644.4% stocked across both layers
Collection C2 / 244 / 2425.0% stocked across both layers
Collection D0 / 186 / 1833.3% stocked, all on new listings
Collection E3 / 122 / 1241.7% stocked across both layers
The week

Descriptive movement, not a process signal

ERP net sales rose by $14,400 / $48,000 = 30.0% from the prior illustrative week. The three displayed values are $45,600, $48,000, and $62,400. Three weeks are insufficient to establish a stable variation signal.

Separately, the Orders API records 174 fulfilled-inventory units and 126 merchant-fulfilled units, totaling 300 units and $53,200 retail gross before fees and adjustments.

Advertising

Concentration and economics, not absent demand

PX-101 has 42.9% of spend and 95.6% ACOS. The synthetic scenario's modeled contribution ceiling is 55%.

Attributed sales overlap marketplace sales and are not ERP net sales.
Advertising groupACOSInterpretation
Empty Group OneNo salesSpend with zero attributed return
Empty Group Two320%Far above modeled ceiling
Group Three180%Far above modeled ceiling
Group Four105%Above attributed sales value
Group Five82%Above modeled ceiling
Group Six49%Below modeled ceiling

Inference: PX-101 also has the highest synthetic sessions and unit demand. Reducing its ad concentration follows from concentration and economics, not an absence of demand.

Separate populations

Five universes, not one clean funnel

Candidate lenses: COMM-INTEGRATE-001, VIS-TASK-001, VIS-INTEGRITY-001.

Counts use different units and inclusion rules. They are not additive.
UniverseCountDefinitionLimitation
Portfolio listings180Complete synthetic listing registryListing count, not a product count
Traffic matched148 / 180Listings matched to Traffic rows32 unmatched listings are unknown, not zero
ERP products165Product-master transaction universeProduct count, not listing count
Inventory API records72Products returned by the inventory APIOmits products without an API record
Zero-stock candidates31 / 72Modeled demand and zero visible API supplyRequires receiving-manifest reconciliation

Across 148 matched listings, 1,644 / 12,840 = 12.8% unit-session conversion. Thirty-two portfolio listings outside that universe are not assigned zero sessions, conversion, or demand.

Uncertainty and counterevidence

Receiving evidence can change the answer

Two synthetic manifests were checked. Manifest M-204 contains 360 units of PX-101 that are absent from the inventory API view, plus 90 units of PX-102 and 45 units of PX-103. The check removes a modeled 180-unit duplicate from the proposed restock.

A receiving manifest is point-in-time operational evidence. Every release must reopen current manifests because shipment state can change after report preparation.

Synthetic restock

A formula followed by a manual control

suggested send = round to 5 of ((30-day units / 30) × 60 − sellable − API inbound), then reconcile each row against current receiving manifests

Synthetic ID or groupProduct label30-day demandSuggested sendRelease condition
PX-204Collection A, variant 190120Recheck receiving manifests
PX-318Collection D, variant 33265Recheck receiving manifests
PX-411Collection C, variant 42755Recheck receiving manifests
PX-425Collection E, variant 22550Recheck receiving manifests
30 tail productsCollections A through EAbout 126310Row-level reconciliation required
TotalOne illustrative shipmentAbout 300600Exclude PX-101; recheck all manifests
Conditional actions

Owners, conditions, checks, and approval

Candidate lenses: PURPOSE-001, NARRATIVE-001, RECURRENCE-001.

Owner roleActionConditionSuccess checkApproval
Advertising ManagerPause three groups and reduce PX-101 concentrationPromoted listings remain emptyNo empty-listing spend; concentration below 35%Human approval required
Supply Planning LeadRelease one approximately 600-unit restockReconcile every current receiving manifest34 products covered without duplicate unitsHuman approval required
Marketplace Operations LeadSeed nine Collection A established listingsLabels and product data are readyNine listings become fulfillment eligibleHuman approval required
Provenance and publication boundary

Synthetic evidence, explicit limitations

This is a separate, fully synthetic reconstruction of a recurring-report pattern. It does not reproduce, sanitize in place, or authorize distribution of any private operating report. There is no mapping key back to a real organization, person, product, identifier, date, or value.

  • ERP net: 390 illustrative invoice units; not retail gross.
  • Channel mix: 300 order units; a different event basis from ERP.
  • Advertising: attributed sales overlap marketplace revenue.
  • Traffic: 148 / 180 listings matched; 32 remain unknown.
  • Inventory: 72 API records; receiving units can be absent.

Research status: manual test vehicle, not a generated report, validated template, or proof of superiority.

Lens status: candidate identifiers remain under evaluation. Publication does not make them accepted principles, standards, or schema requirements.

Reader evidence: no controlled reader comparison has been completed. Claims about accuracy, speed, calibration, or usefulness remain untested.

Research annotations

Candidate lenses under evaluation

Frames before comparison

Can readers identify basis, period, unit, and universe before comparing values?

MEASURE-001 · FRAME-001 · VIS-COMPARE-001
Visible reasoning state

Can readers distinguish observations, interpretations, uncertainties, and conditional recommendations?

ARGUMENT-001 · COMM-ANSWER-001 · UNCERTAINTY-001
Evidence at the decision

Does aligning demand, ads, stock, and receiving evidence reduce unsafe duplicate-shipment actions?

COMM-INTEGRATE-001 · VIS-TASK-001 · VIS-INTEGRITY-001
Recurring recognition

Can a stable sequence lower reader effort without hiding changed populations or incomplete inputs?

RECURRENCE-001 · ACCESS-001 · ACCESS-002

PURPOSE-001 · MEASURE-001 · PROVENANCE-001 · VARIATION-001 · VIS-TASK-001 · VIS-COMPARE-001 · VIS-INTEGRITY-001 · NOTATION-001 · ARGUMENT-001 · COMM-ANSWER-001 · COMM-INTEGRATE-001 · NARRATIVE-001 · UNCERTAINTY-001 · FRAME-001 · ACCESS-001 · ACCESS-002 · RECURRENCE-001

These are candidate lenses, not accepted principles. The page states hypotheses and planned tests, not a claim that this reconstruction is better than another report.