Hyper-local sustainable assortment planning Nupur Aggarwal, - - PowerPoint PPT Presentation

Hyper-local sustainable assortment planning

Nupur Aggarwal, Abhishek Bansal, Kushagra Manglik, Kedar Kulkarni, Vikas Raykar

IBM Research

Sustainability

The fashion industry is considered to be the world’s second largest polluter, after oil and gas. Reducing unsold inventory is one step into this direction.

Assortment planning

Assortment planning, an important seasonal activity for any retailer, involves choosing the right subset of products to stock in each store.

current practices ü Heavily spreadsheet driven and relies on the expertise and intuition of the merchandisers. ü Not scalable when a merchandiser has to do planning for hundreds of stores. ü Typically stores are grouped into store clusters and an assortment is planned for each cluster rather than store. ü A sub-optimal assortment results in excess leftover inventory for the unpopular items, increasing the inventory costs, and stock outs of popular items, resulting in lost demand and unsatisfied customers. ü Existing approaches only maximize the expected revenue under certain store and budget constraints. Along with the revenue the choice of the final assortment has also an environmental cost associated with it.

store-wise hyper-local sustainable

assortment planning

What set of products to stock at each store ? Determine the optimal subset of products to be stocked in each store so that the assortment is ü localized to the preferences of the customers shopping in that store ü diverse ü sustainable The concept generalizes from store to ü distribution centre/warehouse ü region/state/country ü physical/online/omni channel

Literature Survey

Broadly there are three aspects to assortment planning,

• the choice of the demand model
• estimating the parameters of the chosen demand model
• using the demand estimates in an assortment optimization setup.
• Demand Models
• Independent Demand Models - The simplest approach is to assume product demand

to be independent of the offer set or the assortment, that is, the demand for a product does not depend on other available products.

• Choice models – Models demand for product j at store s when the assortment
• ffered at the store was q. In practice the demand for a product is heavily influenced

by the assortment that is under offer mainly due to product substitution (cannibalization) and product complementarity (halo effect)

Estimate sales using matrix factorization

(re-train model every season)

Given a a spar arse Product ct X Store mat atrix :

• The objective is to estimate the unobserved entries in it
• These entries can be 4th week / 8th week STR, sales per month etc.

1) 1) Product ct bias as:

• E.g. special promotions being carried

for Blue denim shirt will lead to higher sales 2) 2) St Stor

• re bias

as:

• E.g. Stores more accessible to the

society will show high overall sales 3) Var ariab able predict ction interval als:

• Observed values in the sparse matrix

might have different confidence levels i.e. some might have tighter UB / LB than others:

• Case 1: Observed 4th week STR for an

entry might be 0.4 but the upper and lower bounds can be 0.48 and 0.36, respectively.

• Case 2 : Observed 4th week STR for an

entry might be 0.6 but the upper and lower bounds can be 0.73 and 0.4, respectively.

Clearly, we need to have differential treatment for such values. 10 20 5 50 34 31 5 6 4 1 23 43 16 12 11 2 3 45 46 5 3 61 42

St Stor

• re 1

St Stor

• re 2

St Stor

• re 3

St Stor

• re 4

St Stor

• re 5

St Stor

• re 6

Product 1 0.24 0.32 0.17 0.64 0.59 0.66 Product 2 0.14 0.43 0.72 0.11 0.76 0.23 Product 3 0.87 0.83 0.86 0.51 0.41 0.88 Product 4 0.91 0.34 0.67 0.54 0.77 0.12 Product 5 0.33 0.11 0.43 0.91 0.44 0.86

No Note e : In red = Estimated

values for missing entries

Method deployed for estimating the missed

• pportunities :

Lat at 1 Lat at 2 Lat at 3 Lat at 4

Prod 1 0.2 0.3 0.1 0.6 Prod 2 0.1 0.4 0.3 0.2 Prod 3 0.6 0.8 0.7 0.3 Prod 4 0.9 0.6 0.1 0.5 Prod 5 0.3 0.3 0.4 0.8 Lat at 1 Lat at 2 Lat at 3 Lat at 4 Store 1 0.2 0.3 0.1 0.6 Store 2 0.1 0.4 0.3 0.2 Store 3 0.6 0.8 0.7 0.3 Store 4 0.9 0.6 0.1 0.5 Store 5 0.3 0.3 0.4 0.8 Store 6 0.2 0.7 0.6 0.1

Latent matrix for products Latent matrix for stores

1) 1) Al

Alternat ating Leas ast Squar ares

• ptimizat

ation

2) 2) Lear

arning vect ctors

3) 3) Differential

al treat atment Coupled with Coupled with

• The underlying hidden

structure (latent feature matrix) influencing the sales of products are learnt

• Product / store bias

features are also learnt

• To overcome the problem
• f variable prediction

intervals the estimated errors are penalized according to width of the confidence band

• Finally, the completed

matrix is fed to the Assortment Planner

Sustainability Scores

ü Sustainability in a product can be related to

social, economical or environmental. Even within environmental impact, it can be further measured using its affect on climate change, eutrophication, resource depletion, water scarcity, chemistry. ü Higg Index is one such index that gives sustainability scores by taking input information about various stages that were involved in fabric manufacturing. ü Higg Index is available for 1 Kg of the fabric. ü Since many garments are blended fabrics, we take a weighted average of the Higg index of fabrics present in the blended fabric and normalize it by the weight of the garment. ü To calculate sustainability score for an assortment, we take a weighted average of the Higg Index of the products in the assortment.

Diversity Ensure that the selected assortment is diverse. Without the diversity constraints the assortment tends to prefer products that are similar to each other. Similar products might cannibalize each others’ sales. Sustainability We would like to include those products that are sustainable from an environmental perspective. A particular jeans might be trendy and have a high demand but may not be environment friendly. Cardinality The number of products to be included in an assortment Complementarity A good assortment has products that are frequently bought together

Multi-objective

• ptimization

formulation

Assortment optimization is solved as a stochastic subset selection

• ptimization problem.

𝑦!" : binary variable indicating presence of product j from the assortment at the store 𝑡. 𝜌!" : revenue generated of product j from the assortment at the store 𝑡. 𝑒𝑘𝑡: the demand for product 𝑘 at store 𝑡 ℎ𝑘: the weighted Higg MSI score for that product 𝜇: a parameter through which the user can specify the relative importance of each objective.

Multi-objective optimization

Sustainability Score

Quality Score Feasible region Infeasible region Convergence to the globally optimal Pareto front Pareto-front

x

Solution on Pareto-front

x x x x x x x x

• An optimal assortment must that have:
• A high sustainability-score (or a low Higg MSI score)
• A high quality-score (or high sales / revenue)
• To optimize these two conflicting objectives a multi-
• bjective optimization (MOO) formulation is considered
• MOO problems have been solved using classical methods

and meta-heuristics in literature

• We choose the weighted sum method here due to its

simplicity in configuration and use.

• Coefficients of different objective functions are continually

changed; and for each coefficient-realization of these, a single-objective optimization problem is solved yielding an

• ptimal assortment.
• Solving these single-objective optimization problems for

multiple coefficient realizations yields a family of Pareto-

• ptimal assortments that are non-dominated with respect

to each other

ü Solve the optimization problem for different values of 𝜇 and 𝛽

ü At each iteration, select the product with highest marginal gain. ü Update the marginal gains of products similar and complementary to the selected product. ü Proven approximation guarantee if the objective function is submodular

Our implementation: ü Heap-based Approach ü Time Complexity O(klog(n)) instead

• f O(kn)

ü Inference time 7secs for 50000 products on a CPU ü Empirically approx. ratio in range (1.01 to 1.03).

Solve multiple single-objective optimization problems

Product Quality and Higg MSI score distribution

• 3 peaks corresponding to
• 100% cotton, 100% viscose, 100% polyester
• Cotton has highest Higg MSI (least sustainable)
• Polyester has least Higg MSI (most sustainable)
• Quality distribution evenly

spread across products

Pareto Optimal Fronts

Pareto frontier and the assortment cluster shrinks relative to the

• frontier. This is because as we aggregate the scores of more

products, the consolidated scores move closer to their mean. The 3 horizontal clusters correspond to the 3 peaks of 100% cotton, 100% viscose and 100% polyester respectively.

Pareto Optimal Fabric composition for different lambda

• Looking at 𝜇 = 0.0 and 𝜇 = 0.5 plots

we see that viscose fabric is dominant since its Higg MSI is lower than cotton and it has the best quality score in terms of quality scores as well.

• For 𝜇 = 1 (maximum importance to

sustainability), the fabric composition in the assortment products comprises mostly of polyester.

Conclusion and Future Work

• We have proposed a method of assortment planning that jointly
• ptimizes the environmental impact of an assortment and the
• revenue. We formulated the problem as a multi-objective
• ptimization problem whose optimal solutions lie on the Pareto

Optimal front. The proposed approach would allow retailers to meet their sustainability targets with minimal impact on the revenue.

• Future work
• cannibalization and halo effects in demand modelling
• diversity and complementarity of products in the assortment in the
• ptimization formulation.
• Use a cradle-to-grave sustainability metric for assortment planning.

THANK YOU Nupur Aggarwal (nupaggar@in.ibm.com)