Recovering Business Rules as Cascade Lookups for Enterprise Field Prediction

Changxu Duan

cs.LGOct 25, 2026

Recovering Business Rules as Cascade Lookups for Enterprise Field Prediction

Authors: Changxu Duan

Organizations: Technische Universität Darmstadt, Darmstadt, Germany

Abstract

Enterprise sales-order systems rely on manually maintained configuration rules to determine fields such as payment terms, shipping points, and sales ownership. When these rules are incomplete or stale, operators must enter values by hand, and downstream errors spread into billing, logistics, and reporting. This paper studies whether missing or outdated rules can be recovered from historical transactions instead of being relearned as opaque statistical patterns. We introduce a knowledge-graph-guided workflow in which an LLM agent proposes candidate business keys, meaning column combinations such as customer and document type, validates them through SQL profiling, and compiles the resulting evidence into deterministic cascade lookups. The LLM is used only during offline rule discovery. At inference time, prediction is a transparent key lookup with no model-serving requirement. Experiments on enterprise sales-order prediction and public relational benchmarks show that cascades are most effective when fields follow stable business rules, while learned graph models remain preferable when prediction depends on temporal or neighborhood patterns. Drift analysis further turns residual errors into diagnostics for configuration maintenance. Code and data: https://github.com/Fireblossom/CascadER.

Figures & tables

Field	Level	Downstream consequence of error
CustomerPaymentTerms	Header	Incorrect invoice due dates and early/late payment disputes
SalesGroup	Header	Misattributed revenue reporting and incorrect commission assignment
SalesOffice	Header	Distorted regional sales analytics
HeaderIncoterms	Header	Wrong allocation of shipping risk and cost between buyer and seller
ShippingCondition	Header	Suboptimal carrier selection and delivery delays
ItemIncoterms	Item	Per-line risk allocation inconsistent with contract

Table 1: Target field classification. Header-level fields apply to the entire order. Item-level fields are set per product line. Downstream consequence describes the operational impact of an incorrect value.

Figure 1: System overview. The top panel shows offline rule discovery producing JSON cascade files. The middle panel shows runtime key lookup providing editable suggestions during order entry. The bottom panel shows drift monitoring comparing current transactions against cascade mappings and surfacing alerts for governance review.

Method	PayTrm	Sales Grp.	Hdr Inc.	Item Inc.	ShipCond	ShipPt	Plant	Sales Off.	Avg
Majority Class	0.23	0.05	0.62	0.62	0.41	0.54	0.59	0.99	0.51
XGBoost [ 1 ]	0.57	0.51	0.70	0.70	0.68	0.95	0.99	0.99	0.76
LightGBM [ 11 ]	0.10	0.02	0.73	0.73	0.51	0.28	0.61	0.99	0.50
CatBoost [ 18 ]	0.44	0.16	0.81	0.80	0.71	0.80	0.99	0.99	0.71
CARTE [ 12 ]	0.62	0.46	0.77	0.75	0.74	0.97	0.99	0.99	0.79
AutoGluon [ 4 ]	0.52	0.34	0.78	0.78	0.74	0.98	0.99	0.99	0.77

Table 2: MRR on rel-salt. Top rows report baselines from SALT [ 13 ] without KG. Best (w/ KG) is the per-field maximum after adding KG-derived text embeddings [ 15 ] , computed from published base MRR +

\Delta

MRR at 2-decimal precision. Bottom row reports our cascade method.

Target Field	Ours (Acc)	GNN (Acc)
CustomerPaymentTerms	82.86%	37.47%
SalesGroup	70.00%	15.76%
HeaderIncoterms	77.22%	62.23%
ItemIncoterms	77.15%	69.36%
ShippingCondition	69.56%	56.85%
ShippingPoint	98.65%	98.39%

Table 3: Accuracy comparison with RelBench v2 baseline. GNN accuracy values are taken directly from the RelBench v2 leaderboard [ 6 ] using its default heterogeneous GraphSAGE [ 7 ] architecture, hyperparameters, and training protocol. We did not re-train or tune the GNN.

Figure 2: (a) Semantic key-selection ablation before full error-driven refinement. (b) Final cascade vs. GNN using heterogeneous GraphSAGE from RelBench v2 [ 6 ] , with per-field test accuracy on SALT-KG. Numbers above bars indicate absolute accuracy difference.

Field	Schema-only	Restricted KG	$\Delta$
Plant	76.6%	99.5%	+22.9
ShippingPoint	67.0%	78.9%	+11.9
HeaderIncoterms	69.1%	73.0%	+3.8
ItemIncoterms	69.1%	72.9%	+3.8
ShippingCondition	56.6%	59.3%	+2.6
SalesGroup	68.9%	71.0%	+2.2

Table 4: Ablation of semantic key selection on SALT-KG test set (accuracy). Restricted KG denotes a KG-guided cascade before the full error-driven refinement and fallback optimization used in the main results.

Dataset	Task	Classes	Ours (Val)	GNN (Val)	Ours (Test)	GNN (Test)
rel-arxiv	author-cat.	53	52.8%	52.6%	50.3%	50.7%
rel-stack	badges-cls.	3	77.6%	80.0%	79.3%	82.8%

Table 5: Generalization study (accuracy on RelBench v2 val and test sets).

Figure 3: Drift analysis. (a) Cross-drift conditional correlation matrix, where asymmetric cells show $P(\text{column drifts}\mid\text{row drifts})$ . (b) Drift rate vs. test accuracy, with bubble size proportional to field cardinality.

Tabular Prediction

Tabular prediction uses structured rows and columns to forecast labels, values, risks, and outcomes from data.

See all 112 papers

Continual Enterprise World Model Discovery in Dynamic Systems
Sep 17, 2026Shambhavi Mishra, David Vazquez, Perouz Taslakian +3World Model Learning World Model Evaluation
ServiceNow Research · LIVIA, ILLS, ÉTS Montréal · Université Polytechnique Montréal +3
Do Enterprise Systems Need Learned World Models? The Importance of Context to Infer Dynamics
May 12, 2026Jishnu Sethumadhavan Nair, Patrice Bechard, Rishabh Maheshwary +14Distribution Shift World Model-Based Planning
1ServiceNow · 2Mila
Business as Rulesual: A Benchmark and Framework for Business Rule Flow Modeling with LLMs
May 24, 2025Chen Yang, Ruping Xu, Ruizhe Li +2LLM Grounding LLM Prompting
Zhejiang University of Technology, China1 · Zhejiang Key Laboratory of Visual Information Intelligent Processing, China 2 · University of Aberdeen, UK +1

Sep 17, 2026J/K move · Enter open · S save

Continual Enterprise World Model Discovery in Dynamic Systems

Shambhavi Mishra, David Vazquez, Perouz Taslakian, Marco Pedersoli+2

ServiceNow Research · LIVIA, ILLS, ÉTS Montréal · Université Polytechnique Montréal+3

World Model Learning World Model Evaluation Interactive World Models

Open paper AskPDF arXiv

In an enterprise system, updating one field can set another, create a record, or start an approval. These effects are produced by business rules that are not built into the platform but written by each organization and revised over time. An agent working in such a system cannot predict the result of its own actions without knowing these rules. We study continual enterprise world model discovery, where an agent starts without knowledge of these business rules and discovers them by interacting with records and observing the outcomes. From those observations it builds a world model, which it revises as the rules change. To evaluate this, we introduce EnterpriseWorldShift, built on a live ServiceNow environment with nine tables, 25 hidden rules and 600 evaluation actions. It presents four versions of the same enterprise world, with the tables and records held fixed while a rule is modified, then added, then removed, so that discovery, revision, extension and retirement are each tested in turn. Our Continual Discovery Agent (CDA) builds such a model and carries it from one world to the next. It predicts the effects of the hidden rules more accurately than looking them up for each question, the approach taken by prior work, by up to 8.98 IoU points, and it answers from its own model without querying the running system.

Abstract

Tabular Prediction

Explore similar work

Continual Enterprise World Model Discovery in Dynamic Systems

Do Enterprise Systems Need Learned World Models? The Importance of Context to Infer Dynamics

Business as Rulesual: A Benchmark and Framework for Business Rule Flow Modeling with LLMs