Observation Catalog

From WFM Labs

Observation Catalog is a series of pages, each of which generalizes one recurring field observation from service operations into a standalone reference article. The observations were gathered in multi-heritage service estates — organizations formed by successive acquisitions, running several routing and workforce platforms and several inherited sourcing doctrines at once — but each page is written in generic form, with no estate data, so that the mechanism it describes can be recognized in any service operation. An observation qualified for the catalog when the mechanism behind it recurred across more than one delivery node or heritage, could be stated without estate data, and could be attributed to an instrument, a structure or an inherited doctrine rather than to the people who produced the number. The series is organized by the three arguments the observations support: that a consolidated estate must work through its definitions before it can compare anything; that the placement of work across delivery nodes is a work-type question rather than a quality-score question; and that value-based planning is first a measurement program.

The catalog behind the series holds thirty observations. Eighteen of them had no home on the wiki and received a page of their own; the remaining twelve landed as diagrams and new sections in pages that already existed. Both sets are listed below.

Act A — definitions before comparison

A service estate assembled from several predecessor companies carries several correct answers to every operational question — what a headcount is, what "designated" means, what a quality score measures — and none of them reconciles with the others. The pages in this act describe the mechanisms by which unreconciled definitions produce contradictory numbers, and the sequence in which a consolidated estate can settle them. These observations bite hardest in estates at Level 2 and Level 3 of the WFM Labs Maturity Model™, where measurement is structured enough to produce a number for each heritage but not yet reconciled enough to compare them.

# Page The observation
1 Inherited Sourcing Doctrine in Merged Service Estates Several sourcing rules, each correct for the business that wrote it, freeze into place after a merger; a schema-first reconciliation dissolves them
2 Designated, Shared and Dedicated: What the Words Implement One commercial word carries five operational readings, and the one that is actually configured is the smallest
3 Headcount Reconciliation in Service Operations The census, the finance view and the human-resources system each count a different perimeter; the figure everyone quotes is the triple overlap that no single system produces
5 Instrument Effects During Measurement Rollout When every population's score moves at once, the instrument moved; the untreated population is the diagnostic
6 Baselines Before Targets A target set before its baseline is unfounded rather than merely wrong, and it is withdrawn, not revised
8 Platform Migration as a Definitional Forcing Function A workforce-platform migration turns shared definitions of gates, skills, pools, states and intents from hygiene into critical path
9 Intent Normalization Before Work Placement One taxonomy of what a contact is for, across channels and brands, must precede any supply-side placement design
11 Mix Effects in Blended Quality Targets When delivery nodes carry different targets, the headline falls by arithmetic alone while every node holds
12 Platform-Imposed Commercial Constraints: When the Platform Decides the Product A platform default, never consciously thrown, rewrites what the commercial organization has promised

Act B — placement as a work-type question

Where a piece of work should be done — which location, which delivery arrangement, which supplier, which pool — is habitually argued from quality scores. The pages in this act show that the answer depends on the physics of the work type: its intent mix, the state of the customer who brings it, its stakes and complexity, whether it can wait, and how much idle cushion the gate that receives it necessarily carries. The mechanisms sit at Level 3 and Level 4: a Level 3 estate has the routing constructs to act on them, and a Level 4 estate plans placement in distributions and so can price them.

# Page The observation
14 After-Hours Service as a Distinct Work Type After-hours contact is not the same work at an inconvenient time; it is a different work type with its own mix, customer state and staffing physics
15 Customer State and Attribution in Quality Measurement The customer's state — planning, mid-journey, or in a disruption — arrives before the agent does, and quality cannot be compared until it is adjusted for
16 Stakes and Complexity: Two Axes of Case Mix Stakes and complexity are independent; the highest-stakes contact in an estate can be one of the simplest
17 Deferrable Work and the Idle Cushion Small gates carry a large proportional idle cushion into which deferrable work fits almost for free, so where such work belongs has no standing answer
18 Gate Expansion: What It Buys and What It Spends Merging gates buys a pooling saving and spends three things: the idle cushion, the likelihood of designation, and the width of pool knowledge

Act C — value-based planning as a measurement program

A planning model optimizes only the objectives its inputs can measure. The pages in this act describe what happens when an objective is unmeasured, when a saving is quoted on rate rather than on cost, when a forecast table cannot say how good the forecast was, and when an assist tool is judged on the wrong metric. Each is a Level 3 to Level 4 transition problem: the instrument exists, and the question is whether the planning model is built on what it measures.

# Page The observation
22 Instrumenting the Objective Before Building the Model A multi-objective placement model collapses to the one term its nodes can measure; the condition is at least one measured attribute per objective family per pool
23 Counterfactual Savings and the Productivity Denominator A rate fifty percent lower, taking a quarter more hours, saves thirty-seven and a half; the saving is on rate times hours, never on rate alone
26 Forecast Vintages A forecast table that overwrites each forecast cannot be audited; the missing column is the date each forecast was made
28 Evaluating Agent-Assist Tools on Ramp Compression The business case for an assist tool is months of proficiency bought, not seconds of handle time saved

Act D — general observations about method

Six observations concern how an organization reasons rather than what it found. They recur in any estate that holds more data than answers, and each is stated so that the remedy is a practice, not a person.

  • Data Synthesis Before Decision — three gaps (definition, register, inference) sit between a stack of reports and one usable answer; the register of graded claims is the step that closes them.
  • AI Reads Everything and Thinks Nothing — a language model over unreconciled data returns fluent wrong answers faster than before; the practice that catches them is the thing at risk.
  • Not Assessable Is Not Performing Badly — "cannot be measured here" and "performing badly" are different findings with different owners and remedies.
  • Translation, Not Negation — a strong position becomes a constraint, a weight or a parameter in the model rather than an argument to be won or lost.
  • Build the Rule, Not the Destination — a strategy that is a list of moves ages from the day it is approved; the durable object is a rule that keeps producing a destination.

Three further method observations were folded into pages that already existed: agreeing what words mean (Platform Migration as a Definitional Forcing Function), reading an instrument change before reading the operation (Instrument Effects During Measurement Rollout), and treating inconsistency as inherited rather than as anyone's fault (Inherited Sourcing Doctrine in Merged Service Estates).

Observations that extended existing pages

Twelve of the thirty observations described a mechanism the wiki already covered. Rather than create a parallel page, each was added to the existing page as a diagram and, where the page lacked the mechanism, a new section. The table gives the catalog number, the page, and where on it the observation landed.

# Page Where it landed
4 Sample Size and Detectable Difference in Quality Measurement The same score on a different floor: why small samples cannot separate nodes
7 Occupancy Section Agent state discipline: occupancy measures whether the agent signalled, not whether they worked
10, 20 Sourcing Design Axes: Node and Client Ownership The organization as a view rather than a schema; section Where ownership is set on offers that reach into fulfillment
13 Comparing Delivery Arrangements The same supplier producing opposite results on different work types
19 Service Chain Decomposition and Node Sourcing Shrinking the specialist's scope rather than relocating the specialist
21 The Escalation Tax Section Completion depth: containment is not a capacity input
24, 29 Pooling Architecture in Service Workforces The dedication price curve and how automation steepens it; coverage floors that bind before workload does
25 Doubly Stochastic Arrivals and Demand Variance Decomposition Only one of the three components of demand variance is forecastable
27 Speed to proficiency curve Section The skill gate: time to proficiency as a continuity requirement rather than a cost lever
30 AI Agent Evaluation and Benchmarking Domain ratio and reliability under repetition: why an agent's benchmark performance compounds down in specialized domains

How the series relates to existing pages

The Observation Catalog is the empirical layer beneath the wiki's conceptual and practice pages. Where the Adaptive Concepts series presents the argument for adaptive workforce management from first principles, each Observation Catalog page starts from something seen in an operation and works back to the mechanism. The two series meet at the practice pages. The definitional pages of Act A land in Comparing Delivery Arrangements, Sourcing Design Axes: Node and Client Ownership and Service Chain Decomposition and Node Sourcing, because a definition that has not been reconciled shows up first as an invalid comparison between nodes. The placement pages of Act B land in Pooling Architecture in Service Workforces and Comparing Delivery Arrangements, where the physics of gates and the comparison key are worked out. The measurement pages of Act C land in Level 4: Planning in Distributions, Capacity Planning Cycle, Supply Elasticity in Workforce Planning and Placement Engine Architecture, the page that specifies the measured-attribute condition the objective page enforces.

Each page carries a diagram drawn to a common register and a See Also section that hands off to the page holding the adjacent mechanism, so that a reader entering at any observation can reach the rest of the argument without returning to this hub.

See Also

Retrieved from "https://wiki.wfmlabs.org/w/index.php?title=Observation_Catalog&oldid=4530"