Returns Management

Reverse logistics · Disposition decisions

Handle returned online and BORIS (buy-online, return-in-store) merchandise — sort, route, and decide for each unit whether to restock, refurbish, liquidate, recycle, or dispose. The unit-level decision is trivial; the operational lever is enforcing capacity caps on each downstream channel and feeding the result back into forward inventory. Foundational papers: Stock & Mulki (2009); Ramanathan (2011); Anderson, Hansen & Simester (2009) on free returns; Ofek, Katona & Sarvary (2011) on returns and online-retail competition.

Why it matters

Scale of e-commerce returns · the cost of asymmetric reverse logistics

~$816B
Estimated US retail returns in 2022 (~16.5% of total retail sales). E-commerce returns run ~20–30% of online sales vs. ~9% in-store.
Source: NRF / Appriss Retail (2022) Consumer Returns report.
15–20%
Reverse logistics processing cost as a fraction of original sales price — receiving, sorting, inspection, repackaging, shipping eat into margin.
Source: Stock & Mulki (2009), Journal of Business Logistics 30(1).
~$50B
Annual US merchandise routed to liquidators because restocking is uneconomic — a structural channel, not a failure case.
Industry estimates (Optoro, B-Stock); NRF reverse-logistics briefings.
policy paradox
Liberal return policies lift purchase intent and return rates (Anderson-Hansen-Simester 2009). Optimal policy is product-category-specific.
Anderson, Hansen & Simester (2009), Marketing Science 28(3).

Where the decision sits

Reverse-logistics symmetry to forward fulfilment

A returned unit arrives at a sort centre with a quality score \(q \in [0,1]\) reflecting condition (1 = like-new, 0 = damaged). The disposition decision routes each unit to one of five channels — main inventory restock, outlet/secondary-channel restock, bulk liquidation, materials recycling, or disposal. Capacity at the liquidator and outlet is contracted in advance; sending more units than the negotiated cap collapses the price recovered, so capacity caps bind. This problem mirrors omnichannel fulfilment in reverse: where forward fulfilment routes from inventory to customers, returns route from customers back to disposition channels.

Initiatecustomer return
Receivesort centre
Disposition decision5-channel routing
Executerestock / liquidate / recycle / dispose

Problem & formulation

Per-unit assignment with channel capacity caps

OR family
Capacitated Assignment
Complexity
\(\mathcal{O}(N \log N \cdot |D|)\) greedy
Solver realism
★★ Greedy + caps
Reference
Stock & Mulki (2009)

Sets and indices

SymbolMeaningDomain
\(i \in \{1, \ldots, N\}\)Returned unit at the sort centre over the planning horizondiscrete
\(d \in D\)Disposition channel: restock-main, restock-outlet, liquidate, recycle, dispose5 options

Parameters

SymbolMeaningUnit
\(q_i \in [0,1]\)Quality score (1 = like-new, 0 = damaged)scalar
\(p\)Original sales price$ / unit
\(q_{\text{thr}}\)Quality threshold to be eligible for main restockscalar
\(c_d\)Per-unit processing cost in channel \(d\)$ / unit
\(K_d\)Capacity cap for channel \(d\) (units)units

Per-unit value of each disposition

Recovery value depends on quality only for the two restock channels; liquidation, recycling, and disposal are quality-independent (the buyer pays a flat per-unit price, or a recycler pays scrap value).

$$\begin{aligned} v_{\text{restock-main}}(q) &= p \cdot q - c_{\text{main}} \quad \text{if } q \geq q_{\text{thr}}, \text{ else } -\infty \\ v_{\text{restock-outlet}}(q) &= p_{\text{out}} \cdot q - c_{\text{out}} - s_{\text{out}} \\ v_{\text{liquidate}} &= p_{\text{liq}} - c_{\text{liq}} \\ v_{\text{recycle}} &= p_{\text{scrap}} - c_{\text{rec}} \\ v_{\text{dispose}} &= -c_{\text{dis}} \end{aligned}$$

Disposal is the only strictly negative option — the safety net when nothing else is feasible or profitable.

Decision variables and objective

$$x_{i,d} \in \{0,1\}, \;\; \sum_{d \in D} x_{i,d} = 1 \;\;\forall i \qquad\quad \max \;\; \sum_{i=1}^{N} \sum_{d \in D} v_d(q_i) \cdot x_{i,d}$$

Capacity constraints (the only thing that couples units)

$$\sum_{i=1}^{N} x_{i,d} \;\leq\; K_d \qquad \forall d \in \{\text{liquidate}, \text{outlet}\}$$

Without capacity constraints the problem decomposes per unit and is trivial: pick the max-value disposition. The interesting OR enters because (i) the liquidator quotes a price valid for at most \(K_{\text{liq}}\) units per period, and (ii) outlet shelf space is finite. Restocked units feed forward inventory — link to multi-echelon inventory. Return-rate feedback shapes store location and ship-from-store routing.

Real-world → OR mapping

How returns-desk vocabulary translates to the assignment

At the returns deskIn the model
Daily returned units to be processed\(N\)
Inspector grade (A / B / C / damaged)\(q_i\)
Like-new threshold for main-channel resale\(q_{\text{thr}}\)
Liquidator contract volume (per week)\(K_{\text{liq}}\)
Outlet store shelf throughput\(K_{\text{out}}\)
Refurbish & repackage labour cost\(c_{\text{main}}\)
Disposition routing decision\(x_{i,d}\)

Interactive solver

Greedy max-value assignment with channel capacity caps

Disposition assignment solver
Beta-distributed quality · max-value with cap fallback
★★ Greedy
Units per day
Avg condition score
Spread (beta)
Units accepted at quoted price
Outlet shelf throughput
Reproducibility
—
Total recovery ($)
—
Avg per-unit recovery ($)
—
% restocked (any)
—
% liquidated
—
Total processing cost ($)
—
Net loss vs gross sale ($)
Restock-main Restock-outlet Liquidate Recycle Dispose

Under the hood

We sample \(N\) quality scores from a Beta distribution with mean \(\mu_q\) and standard deviation \(\sigma_q\) (parameters back-solved by moment matching). For each unit we compute the per-unit value of every disposition channel and pick the highest-value option. Liquidation and outlet have hard capacity caps: when a cap is hit, the algorithm falls back to the next-best option for that unit. Without capacity constraints the problem decomposes and the greedy is exact; with caps it is a heuristic, but for this LP-relaxable structure (sort by opportunity cost of the binding cap) it returns the LP-optimal integer solution. Runs in \(\mathcal{O}(N \log N \cdot |D|)\) — sub-millisecond for \(N = 500\) in the browser.

Reading the solution

What a returns-operations manager actually does with the assignment

Three patterns to watch for

  • Quality threshold dominates the mix. Drop \(q_{\text{thr}}\) from 0.70 to 0.55 and the main-restock share jumps; recovery rises but customer-experience risk rises too (returned-as-new units that fail).
  • Capacity caps create a fallback cascade. When the liquidator cap binds, marginal units cascade into recycling or disposal — the recovery ladder steps down sharply. This is why contract negotiation with the liquidator (raising \(K_{\text{liq}}\)) is often a higher-leverage move than internal process tuning.
  • The disposal floor is a red flag. If more than 5–10% of units land in disposal, the upstream return policy is too liberal for the category — revisit the policy (Anderson-Hansen-Simester 2009) rather than the back-end.

Sensitivity questions the model answers instantly

  • Negotiate liquidator cap up by 50% — how much recovery is unlocked?
  • Refurbish process improvement: drop \(c_{\text{main}}\) from $6 to $3 — does main-restock share grow at the expense of outlet?
  • Quality distribution shifts (better packaging, more in-store BORIS returns) — raise \(\mu_q\) and watch restock share climb.

Model extensions

From single-period assignment to richer reverse-logistics OR

SKU-level returns forecasting

Probabilistic forecast of return volume by SKU and condition class — feeds capacity contracts, sort-centre staffing, and inventory pooling decisions.

Return-aware ship-from-store

Route forward fulfilment away from zip codes with high return rates. Cross-link to ship-from-store.

Liberal returns policy design

Anderson-Hansen-Simester (2009) framework: jointly optimise the policy (window length, fee, condition rules) and the disposition pipeline.

Free-returns subscription

Membership programme bundling free returns — alters the demand and return-rate joint distribution; net economics depend on cross-purchase lift.

Fashion-specific returns

Size-fit issue dominates — size profile of the bulk inventory shapes return rates more than condition. Cross-link to fashion buying.

BORIS in store

Buy online, return in store: returned units enter the store inventory as forward stock candidates. Cross-link to omnichannel fulfilment.

Circular retail (refurbish + resell)

Refurbishment as a first-class disposition channel with its own labour, parts, and warranty cost — the basis of brand-owned secondary marketplaces.

Returns + carbon footprint

Add a CO\(_2\) cost coefficient per disposition channel; trade off recovery against environmental impact (disposal >> recycling > restocking).

Key references

Foundational reverse-logistics and returns-policy literature

Ramanathan, R. (2011).
An empirical analysis on the influence of risk on relationships between handling of product returns and customer loyalty in e-commerce.
International Journal of Production Economics 130(2): 255–261. doi:10.1016/j.ijpe.2010.12.013
Anderson, E. T., Hansen, K. & Simester, D. (2009).
The option value of returns: Theory and empirical evidence.
Marketing Science 28(3): 405–423. doi:10.1287/mksc.1080.0430
Ofek, E., Katona, Z. & Sarvary, M. (2011).
"Bricks and clicks": The impact of product returns on the strategies of multichannel retailers.
Marketing Science 30(1): 42–60. doi:10.1287/mksc.1100.0588
Stock, J. R. & Mulki, J. P. (2009).
Product returns processing: An examination of practices of manufacturers, wholesalers/distributors, and retailers.
Journal of Business Logistics 30(1): 33–62. doi:10.1002/j.2158-1592.2009.tb00098.x
Petersen, J. A. & Kumar, V. (2009).
Are product returns a necessary evil? Antecedents and consequences.
Journal of Marketing 73(3): 35–51. doi:10.1509/jmkg.73.3.035
Mollenkopf, D. A., Rabinovich, E., Laseter, T. M. & Boyer, K. K. (2007).
Managing internet product returns: A focus on effective service operations.
Decision Sciences 38(2): 215–250. doi:10.1111/j.1540-5915.2007.00159.x
Guide, V. D. R., Souza, G. C., Van Wassenhove, L. N. & Blackburn, J. D. (2006).
Time value of commercial product returns.
Management Science 52(8): 1200–1214. doi:10.1287/mnsc.1060.0628
NRF / Appriss Retail (2022).
Consumer Returns in the Retail Industry — annual report.
National Retail Federation, Washington DC.

Back to the retail domain

Returns management is the reverse-logistics counterpart to forward fulfilment — the operational lever that decides how much of a product’s revenue survives once a customer changes their mind.

Open Retail Landing
Educational solver · greedy assignment with hard capacity caps · validate channel recovery prices and quality-score distribution against your own returns-desk data.