The Data Debt of Stockouts: Partial Identification and Minimax-Regret Inventory Control
Authors: Jinbiao Chen, Shuang Jin, Hanzhang Qin
Organizations: Department of Industrial Systems Engineering and Management, National University of Singapore · Institute of Operations Research and Analytics, National University of Singapore
Problem definition: Inventory decisions select which demand enters sales logs. Unrecorded forecasts can leave demand frequencies unidentified even when stocking levels are recorded. We study which multistage ordering decisions these logs can justify and what finite samples can guarantee. Methodology/results: Nested stockout thresholds yield exact sets of compatible demand laws and finite representations for policy evaluation. We choose one deterministic inventory rule to minimize its largest expected cost gap from the informed optimum under the same law. This population value is data debt for the specified information and policy class. Under independent demand, carryover can create positive debt despite common optimal actions in every isolated period. Finite samples add learning error above debt. The required sample count grows linearly with the selection bound, with upper and lower orders matching up to logarithmic factors. This rate requires fixed demand and cost bounds, sufficient revelation, and a uniformly positive probability of clearing all feasible stock across the confidence sets. Controlled and semi-synthetic studies separate policy improvements from tighter guarantees. Managerial implications: More records can tighten guarantees while debt persists. Deeper revelation can establish common optimality without fully identifying demand. Using these guarantees requires validated logging, a credible selection range, and a regret tolerance in cost units. Information acquisition must distinguish tighter assurance from operating savings.
Figures & tables
Figure 1: The same stockout log supports different demand frequencies and opposite optimal actions. (a) Two compatible completions at Γ=11 . Percentages refer to the full population. (b) The selection bound varies while the log stays fixed. (c) Ordering and no-order costs. The actions tie at p0=0.5 , and the other action at either endpoint adds 0.25 in expected cost.
Figure 2: Rectangular data debt in the two-period instance of Proposition 7 . (a) Each point is a product demand law with demand-one probabilities (θ1,θ2) . Colors show the informed optimal policy, with indifference on θ1+θ2−θ1θ2=0.70 . L and H denote the endpoints 0.35 and 0.65, in period order. (b) Both undominated deterministic policies are evaluated at the four corners, where their worst regrets occur. The smaller row maximum is ΔrectM=0.1225 .
Figure 3: Controlled CP diagnostics at Γ=5.5 . (a) Means over 20 datasets. Shading is the interquartile range of U . (b) Population debt, selected-policy suboptimality, and assessment conservatism at N=1,000 . (c) Dispersion and the count meeting tolerance 0.15.
Metric
Lower
Equal
Higher
Range of mean differences
CP-regret minus plug-in-regret, both at Γ=10
Common certificate U
9 (7)
18
0 (0)
[−0.0522,0.0000]
Excess population risk
6 (4)
18
3 (3)
[−0.0089,+0.0778]
True-law regret
6 (4)
18
3 (3)
[−0.0314,+0.0523]
CP-regret: Γ=10 minus Γ=1
Common certificate U
14 (13)
13
0 (0)
[−0.4223,0.0000]
Table 2: Cost conditions separate changes in certificates from changes in risk
Study
Information restriction
Decision target
Qin et al. (2026)
Demand law fixed conditional on the full boundary array.
Inventory policy gaps and value estimation from scalar records.
Han et al. (2026)
History-based stocking; current demand law fixed given history.
Offline policy gaps; online cumulative regret.
Jin et al. (2026)
Marginal sensitivity model for unobserved treatment confounding.
Inventory control studies how to order, allocate, and position stock to balance service levels, operating costs, uncertainty, and capacity constraints.
We study data-driven multi-period lost-sales inventory control under censored demand, where a stockout reveals only that demand exceeded the stocking level. We develop a unified, model-based framework for policy learning from censored data, built on a new cost decomposition for base-stock policies and a biased sample-average approximation (SAA) approach. The cost decomposition allows us to propose a new coverage condition under which censored observations are informative enough for sample-efficient policy learning. Guided by this coverage condition, we design two biased SAA algorithms: an upper-biased one that achieves near-optimal sample complexity under the offline coverage condition, and a lower-biased one that actively generates the required coverage and achieves near-optimal regret online. More broadly, this biased SAA approach provides a general principle for implementing pessimism and optimism under censored feedback, which may be of independent interest.
We study the offline data-driven newsvendor problem when historical demand is censored by past inventory levels. We introduce a class of policies for which worst-case expected regret over all bounded demand distributions admits an exact finite-dimensional characterization, valid for any sample size and historical inventory design. The dimension depends on the number of distinct inventory levels, rather than the number of observations. This framework yields exact analyses of the sales-as-demand heuristic, the Kaplan-Meier policy, and recent policies from the censored-newsvendor literature. We leverage this reduction to derive sharp insights on the achievable performance of standard inventory policies under demand censoring. In particular, our analysis of the Kaplan-Meier policy shows that while demand censoring fundamentally limits what can be learned from passive sales data, just a small amount of targeted exploration at high inventory levels can substantially improve worst-case guarantees. Finally, we further leverage our reduction to design new policies that push the frontier and substantially improve upon the worst-case regret of known policies, reducing it by up to 40%.
We study inventory control with decision-dependent censoring, focusing on the censored or repeated newsvendor (R-NV), where each order quantity determines whether demand is fully observed or censored by sales. Existing approaches based on parametric Thompson sampling (TS) can be brittle under prior mismatch, while offline imputation methods need not transfer to online learning. Motivated by the predictive view of decision making, we combine these ideas by taking oracle actions on learned completions of latent demand. We propose in-context generative posterior sampling (ICGPS), which uses modern generative models that are meta-trained offline and deployed online by in-context autoregressive generation. Theoretically, we show that the Bayesian regret of ICGPS with a learned completion kernel is bounded by the Bayesian regret of a TS benchmark with the ideal completion kernel plus a deployment penalty scaling as T times the square root of the completion mismatch. This yields a plug-in template for operational problems with known TS regret bounds. For R-NV, we derive sublinear Bayesian regret by reducing censored feedback to bandit convex optimization feedback. We also show that, under reasonable coverage and stability assumptions, the online completion mismatch is controlled by the offline censored predictive mismatch, so offline predictive quality transfers to online performance. Practically, we instantiate ICGPS with ChronosFlow, which combines a frozen time-series transformer backbone with a trainable conditional normalizing-flow head for fast censoring-consistent sampling. In benchmark experiments, ChronosFlow-ICGPS matches correctly specified TS, outperforms myopic and UCB-style baselines, and is robust to prior mismatch and distribution shift. ChronosFlow-ICGPS also performs well for the real-world SuperStore dataset, especially under heavy censoring.
Sohom Mukherjee, Anh-Duy Pham, Richard Pibernik +1
Julius-Maximilians-Universität Würzburg · National University of Singapore Zaragoza Logistics Center