Constrain the inventory policy before training the agent

A demand forecast informs replenishment, but it does not decide an order. Current stock, lead times, supplier limits, substitution and the treatment of unsold goods all influence that decision. Improving the forecast is valuable only if the ordering policy turns the additional information into better outcomes.

Before introducing an inventory agent, agree which actions it may take and which costs it must balance. Specify the operating limits and fallback. This gives the learning system a defined objective and gives the business a policy it can review before orders affect stock and cash.

Prediction and action fail differently

The earlier article on demand-forecast overrides asks when a human should change a prediction. The question here is downstream: given a prediction, which replenishment action is permitted and how is its effect measured? A forecast error is a statement about information. An order error changes cash, stock availability and sometimes the next observation of demand. Stockouts hide sales the firm could have made, so future training data can also reflect yesterday’s policy rather than true demand.

The DeepStock paper reports reinforcement learning for inventory management on Alibaba’s Tmall platform. Its authors regularise learned policies using classical inventory ideas, including a base-stock policy, and report deployment on that platform. The useful design lesson is not that every retailer should reproduce their system. It is that a policy can learn within a structure informed by how inventory decisions work, rather than treating every possible order as equally sensible. The paper’s synthetic comparisons and one deployment do not establish a return for another assortment or supply chain.

Draw the policy boundary first

Define the action space in business terms. Which products may be reordered? How far ahead can the agent commit? What supplier minimums, shelf-life rules, storage limits or cash ceilings apply? Which items require approval because they are scarce, regulated or strategically important? These are not features to discover from sales history. They are choices the operator must make.

Give the system a stable incumbent policy to beat. A base-stock or reorder rule, even if imperfect, gives the buyer a comprehensible baseline. Compare the proposed policy on service level, stockouts, holding cost, waste, expedite cost and working capital, not a single reward that can conceal a harmful trade-off. Preserve the action recommended, action taken and reason for each override. That record is needed to evaluate the policy when staff intervene.

Separate simulation from release. A backtest using historic sales must account for missing demand when shelves were empty and for actions the old policy never tried. A simulation can expose a fragile policy but cannot prove that customers and suppliers will behave identically in production. Begin with a bounded assortment, approval thresholds and a manual route if data or supply conditions move outside the tested range.

Define the conditions that require review. Supplier delays, promotions, unfamiliar products or substantial cost changes may invalidate assumptions learned from earlier operations. Decide whether the response is to limit an order, seek approval or return to the incumbent rule. The model should not acquire additional authority simply because conditions are unfamiliar.

Consider an item with a strong demand forecast but a long, uncertain supplier lead time. A forecast alone suggests ordering enough to cover expected sales. The policy must also decide how much cash to commit, what stockout risk the firm accepts and whether an approved substitute changes the order. If the supplier misses one delivery, a learner trained on normal lead times may increase the next order precisely when the buyer wants a manual decision.

For the lead-time example, set an ordering band under ordinary conditions and send breaches to the inventory owner. Retain the forecast, recommended order, permitted band, actual order and reason for intervention. These records help distinguish the policy’s contribution from staff corrections and identify where its assumptions no longer hold.

The release comparison should include the incumbent rule on the same items and period, accounting for promotions and supplier shocks. If a simulator assumes every missed sale was lost demand, document that assumption and test how the result changes when customers substitute or return later. A policy that wins only under one unobservable demand estimate has not yet earned authority to place live orders.

The limit of the claim

DeepStock is evidence that a structured learned policy can be deployed in one large ecommerce setting. It is not evidence that reinforcement learning is necessary for a smaller inventory problem. Many firms should first improve stock records, lead-time estimates and an ordinary reorder rule. The extra complexity is justified only when a measured decision can improve after those basics are working.

The operations owner should approve the permitted policy before training or release. The model can then be evaluated on a specific task: improving replenishment within agreed service, cash and risk constraints. If a simpler rule meets those requirements, additional learning complexity needs a demonstrable benefit.