Application of Decision Sciences
to Solve Business Problems
Analytics
for Retail
Merchandising
Store-layout Planning
Effective macro space planning is critical for consumers to have a favourable perception about the store’s environment and to increase sales throughput. Store layout planning shows the size and location of each department, permanent structures, fixture locations, merchandizing and overall aesthetics in order to maximise revenue, increase consumer footfalls and conversion rates.
Store layout includes category space allocations and adjacencies in order to optimize overall sales and footfalls.
It helps in answering business questions like:
Which departments will produce the highest traffic? Which departments should be placed adjacent? Where different departments should be spaced to facilitate the deepest penetration and widest
dispersion of consumer flow throughout the store?
Where different categories should be placed so as to have maximum impact on consumers? Where the seasonal items should be placed?
House Hold Laundry
Party Needs
Toiletries
Crisps & Snacks
Confectionery Pet Foods Beverages Cookies Stationary Magazines Cereals Milk Juices Cosmetics
Cakes Soft Drinks Bakery Items
Fish & Meat Grocery
Sauce Pickle Canned-Soup Canned-Vegetables
Baskets
B
as
kets Beer Stacks
Chilled Beers & Wines
Red & White Wines Beers & Cigars
D ai ry P ro duc ts Fr o zen Fo o d RTE Food Vegetables Fruits Entr anc e Stairs C o u n te r 1 C o u n te r 2 C o u n te r 3
Merchandising
Assortment Optimization and Planogramming
Different products and SKUs within a category are assorted based on their profit and revenue contribution and local consumer preferences. It can be assorted for a store cluster or localized at an individual store level. Dividing the products into—core/destination (drive the store sales), complementary/accessory (add-on items for core products), secondary items (not core, but have the potential to develop over time) and impulse items—also serves as the basis for assortment.
Once the right assortment has been decided for each category, the next logical step is placing them in the most effective manner on the shelf. Planogramming is a widely used technique for the same. It enables the retailers to stock the product, at the right place, at the right time, with the right facing to attract the consumers and prompt them to buy.
Category1 Category2 Category3 Category4 Category5 Category6 Category7
Shelf space allocation for categories based on incremental revenue per unit space
1 2 3 5 6 4 X Y
MINIMUM Marginal Space Allocated (%) MAXIMUM
M ar gi n al S al e s (% t o th e A vg . S al e s
Supply Chain
Sales Forecasting
A good demand forecast helps improve sales volume, cash flow and hence the profitability, by optimizing inventory and by minimizing out-of-stock. Besides considering historical data, external factors like changing trends and consumer preferences, seasonal impact and promotion influences on demand, price changes, different store formats and channels are also considered for more accurate forecasts. Sales forecast in retail is very essential for:
Stock replenishment by categories and SKUs
Predicting excess and stock-outs at SKU level and hence minimizing costs
Designing store promotion activities and optimizing resource allocation for the same
Capitalizing on peak sales weeks: Accurate forecasting ensures right product mix to take full advantage of operational capacity and peak market demands
Statistical techniques (like Moving averages, Holt Winters, Regression, ARIMA, etc.) are employed to project future demand on a category and SKU level, based on historical data.
0.0 1.0 2.0 3.0 4.0
5.0 Actual Sales Forecasted Sales Base Line Sales
M ill ion ca ses so ld
Supply Chain
Lead time : It is the time lag between
when the order is placed and the point at which the stocks are available; A lead time of 4 days implies that there should always be stock for 4 days supply to avoid
stock-out scenario
Safety stock is the buffer quantity to
cover any unplanned excess requirement taking into account delivery delays
Reorder point is the minimum level of
stock at which procurement should be triggered and quantity of warehouse stock should never go below this point
If the quantity of warehouse stock is less than re-order point, there is shortfall
Stock Time Release date Safety Stock Reorder point Availability date Lot size Replenishment lead time
Inventory Management
Optimal inventory management is an indispensable function to ensure un-interrupted product supply to meet the consumer demand. Stock out analysis on a category and SKU level helps in:
Optimizing inventory and service levels by streamlining ordering processes Minimizing stock out which leads to loss of sales
Handling overstock which results in increased inventory handling costs and cost to liquidate the excess inventory
Maximizing warehouse space utilization
Designing store promotion activities and optimizing resource allocation
Concepts of lead-time and re-order point are utilized for inventory planning. Lead time is the time lag at which order is placed to the point at which stocks are available. The buffer quantity to cover any unplanned excess requirement, taking into account delivery delays, is referred to as safety stock. Providing for safety stock, on top of lead time demand, will give the re-order point, which is the minimal level of stock at which procurement should be triggered. Warehouse stock should never go below the re-order point. Re-order point will assist in deciding what would be the optimal order quantity and when to place an order.
Supply Chain
Vendor Management
For the efficient and smooth functioning of a retail store, various departments have to work in tandem. Mostly these day to day operations are outsourced to vendors. Constant monitoring and evaluation of vendors is necessary to maintain the smooth functioning of different departments. It enables to control costs, drive service excellence and mitigate risks to gain increased value from their vendor by:
Minimizing potential business disruption Avoiding deal and delivery failure
Ensuring more-sustainable multi-sourcing, while driving most value from the vendors Improving operational efficiencies, control costs and planning of workforce
Partner Strategic Fit Brand Equity Financial Health Ability to operationalize Final Score Status
Vendor 1 9 8 10 7.4 8.75 Pass Vendor 3 10 9 8 7.4 9.00 Pass Vendor 3 10 7 6 7.4 7.50 Pass Vendor 4 10 10 8 10.0 9.50 Underleveraged Vendor 5 9 7 8 7.4 7.75 Pass Vendor 6 2 7 6 8.2 5.50 Risky
Vendor Filtration Methodology & Process Flow
It includes vendor identification, recruitment, monitoring, tracking and evaluating the vendors on certain KPIs like: Pricing: Competitive pricing
(comparable to other vendors), stability (low variance), advance notice of price changes
Quality: Compliance with purchase order, conformity to specifications, reliability (rate of product failures), durability, support, warranty
Delivery: Time, quantity, lead time, packaging, emergency delivery, technical support
Marketing
Loyalty Analytics
In today’s competitive business scenario with consumers having a multitude of options, their preferences & buying patterns have been constantly evolving. It is necessary for retailers to gain insights into changing consumer trends & accordingly tailor their offerings.
CRM analytics helps analyse consumer’s transactional and others behavioural patterns to facilitate optimal decisions regarding marketing strategies.
It helps the business to:
Identify consumer segments based on demographic, psychographic and purchase behaviour. Design
customized product offerings and marketing strategies relevant for each of these consumer segments.
Track these segments over time to study how the industry is evolving.
Closely track and maintain constant touch-point with your most profitable & loyal consumer segments. Identify any signs of attrition in advance and accordingly formulate the right retention strategy
Formulate cross-selling and up-selling strategies by analysing product affinities & associations. Identify the consumer segments which can be targeted for the same. This helps in increasing overall revenue contribution from the same customer base.
They are frequent visitors and prefer brewed coffee
They visit mostly during weekends, to sip their coffee over an
enjoyable experience of a football game or a
live concert
These consumer have no set routines, and
visit during lunch hours on weekdays, and prefer not to be disturbed over their discussion. Cappuccino is their preferred drink
They come generally for the desserts & smoothies, visit usually
during evening hours
“Coffee Junkies” “Entertainment Seekers” “Business over Coffee“ “The Sweet Tooth”
16,000 consumers (37.2%) - $178 12,000 consumers (28%) - $60 8,000 consumers (18.6%) - $125 7,000 consumers (16.2%) - $29 M T W T F S S M T W T F S S M T W T F S S M T W T F S S
Mor Aft Eve Mor Aft Eve Mor Aft Eve Mor Aft Eve
Marketing
Pricing Analysis
Pricing strategies are crafted to meet two key objectives: profit and revenue maximization. It helps in identifying the best pricing strategy for a retailer. Price optimization enables retailers evaluate cost, assortment, margin targets and promotions. It employs predictive modeling techniques for:
Evaluating price elasticity for their private labels and deciding the optimal price points
Identifying price gaps/thresholds to decide the optimal price points and associated discounts for different brands and SKUs while maximizing category sales
Determining base, promotion, markdown and discount prices
Identify price thresholds by brands Optimum price corridor for retailer’s own label
53.2 44.0 30.4 20.4 19.1 18.3 0 10 20 30 40 50 60 70 0 5 10 15 20 25 30 35 $0.90 to $0.98 $0.99 $1.00 to $1.08 $1.09 $1.10 to $1.18 $1.19 $1.20 to $1.28 $1.29 $1.30 to $1.38 $1.39 $1.40 to $1.48 $1.49
% ACV Brand A sales rate
0 10 20 30 40 50 60 70 80 90 100 110 120 W k-1 ( '0 9 ) W k-4 ( '0 9 ) W k-7 ( '0 9 ) W k-1 0 ( '0 9 ) W k-1 3 ( '0 9 ) W k-1 6 ( '0 9 ) W k-1 9 ( '0 9 ) W k-2 2 ( '0 9 ) W k-2 5 ( '0 9 ) W k-2 8 ( '0 9 ) W k-3 1 ( '0 9 ) W k-3 4 ( '0 9 ) W k-3 7 ( '0 9 ) W k-4 0 ( '0 9 ) W k-4 3 ( '0 9 ) W k-4 6 ( '0 9 ) W k-4 9 ( '0 9 ) W k-5 2 ( '0 9 ) W k-3 ( '1 0 ) W k-6 ( '1 0 ) W k-9 ( '1 0 ) W k-1 2 ( '1 0 ) W k-1 5 ( '1 0 ) W k-1 8 ( '1 0 ) W k-2 1 ( '1 0 ) W k-2 4 ( '1 0 ) W k-2 7 ( '1 0 ) W k-3 0 ( '1 0 ) W k-3 3 ( '1 0 ) W k-3 6 ( '1 0 ) W k-3 9 ( '1 0 ) W k-4 2 ( '1 0 ) W k-4 5 ( '1 0 ) W k-4 8 ( '1 0 ) W k-5 1 ( '1 0 ) W k-2 ( '1 1 ) W k-5 ( '1 1 ) W k-8 ( '1 1 ) W k-1 1 ( '1 1 ) W k-1 4 ( '1 1 ) W k-1 7 ( '1 1 ) W k-2 0 ( '1 1 ) W k-2 3 ( '1 1 ) W k-2 6 ( '1 1 ) W k-2 9 ( '1 1 ) W k-3 2 ( '1 1 ) W k-3 5 ( '1 1 ) W k-3 8 ( '1 1 ) W k-4 1 ( '1 1 ) W k-4 4 ( '1 1 ) W k-4 7 ( '1 1 ) W k-5 0 ( '1 1 )
Price index vs. competition Volume share
Marketing
Consumer & Trade Promotions
Trade promotions and consumer promotions refer to different marketing activities implemented in the store, to increase footfalls and to drive sales and profit. The most commonly implemented programs are features, in-store displays, TPRs (temporary price reductions), couponing and loyalty reward programs.
Advanced econometric modeling techniques are used to help stores refine their promotion strategies, to understand the lift generated by various promotional programs for different categories and the associated ROI. This information is then used by marketers to:
Optimally allocate budget among different promotion vehicles—features, displays, TPRs and couponing while increasing category sales and maximizing ROI
Optimally allocate budget for different brands as per their revenue and profit contribution Design programs specific to a category instead of following “one-size fits all” approach
0% 10% 20% 30% 40% 50% 60% 70% 80% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% TPR Only Feature Only Display Only Feature & Display
Im p act on V ol u m e S h ar e Level of Discount 0% 5% 10% 15% 20% 25% 30%
TPR Feature Display Feature & Display
R
O
I
Promotion program Spends
Streaming Sales Data fed weekly or monthly as is available
Promotion Calendar fed into the system periodically Marketelligent PRISM µ Display µ Feature µ Consumer µ TPR Decomposed Lift (µ)
Marketing
Real-time evaluation of promotions
Marketelligent has developed an in-house proprietary tool called PRISM, for continuous monitoring and evaluation of trade and consumer promotions on a real time basis, using the test-control approach.
Identifying the control samples for each of the test group takes most of the time and effort. PRISM minimizes the time required for the same and identifies the control samples on a real time basis, based on
historicalsales trends and outlet demographics.
PRISM uses sales in test and control outlets, to calculate the lift factor for each or combinations of trade promotion programs. Based on the lift factor, incremental sales and ROI are calculated for each activity. The effectiveness of promotions can be compared at different levels – channels, categories, brands and markets.
Marketing
Market Mix Modeling
Usually for marketing, retailers utilize radio, magazines, newspapers and outdoor for creating awareness. Market mix modeling helps managers develop an optimal media investment strategy that provides the required sales lift and also maximises the returns on investment by media vehicle.
The model aids in:
Establishing key relationships between sales and marketing driver inputs Quantifying impact of each marketing driver on sales
Optimizing allocation spends across media vehicles to maximise sales
Calculating saturation spends for each media vehicle based on diminishing returns Evaluating decay impact, if any, for each of the media vehicles (also called ad-stock)
Decompose sales into baseline and incremental Evaluate ROI from each media vehicle
R O I in cr em en tal v o lu me Sp en d s i n USD Jan ’0 9 Fe b ’0 9 M ar ’0 9 Ap r’ 0 9 M ay ’0 9 Ju n ’0 9 Ju l’0 9 Au g’ 0 9 Se p ’0 9 Oc t’ 0 9 N o v’ 0 9 D ec’ 0 9 Jan ’1 0 Fe b ’1 0 M ar ’1 0 Ap r’ 1 0 M ay ’1 0 Ju n ’1 0 Ju l’1 0 Au g’ 1 0 Se p ’1 0 Oc t’ 1 0 N o v’ 1 0 D ec’ 1 0 0 100 200 300 400 500 600 700 800 900 0 2 4 6 8 10 12 14 16 18 20 2% 4% 6% 8% 10% 12% 14%
Total Spends Radio spend Newspaper spend Outdoor spend Baseline sales Online incr. sales TV incr. sales Daily incr. sales
Online spend TV spend Dailies spend
R
O
Marketing
Market Basket Analysis
Market basket analysis is done to evaluate consumers’ purchasing behaviour and to identify the different items bought together in the same shopping session. It uses store’s transactional data and is leveraged for creating cross-selling opportunities for furthering sales.
It aids retailers in:
Product placements--Which products should be placed next to each other Customizing layouts, assortments and pricing, to the local demographic
Affinity promotion--Designing more profitable and effective consumer promotions like couponing based on associated products
• Increase the profit from sales of complementary products, which do not sell by themselves
• Stimulate trials and increase consumer awareness during launch of new products and variants
• Handling excess stock by designing offers among associated products
Support, Confidence and Lift are used to identify the combination of products consumers buy together most often.
CONFIDENCE Product 1 Product 2 Product 3 Product 4 Product 5 Product 6 Product 7 Product8 Product 1 100% 25% 9% 6% 18% 2% 28% 31% Product 2 42% 100% 7% 8% 22% 6% 29% 22% Product 3 31% 16% 100% 5% 10% 4% 18% 17% Product 4 35% 29% 8% 100% 28% 7% 26% 12% Product 5 47% 35% 8% 12% 100% 3% 37% 24% Product 6 37% 66% 18% 19% 21% 100% 25% 21% Product 7 45% 28% 8% 7% 23% 2% 100% 25% Product 8 57% 24% 9% 3% 17% 2% 29% 100%
Probability that Product 8 is purchased given that Product 1 is bought is 31%
Probability that Product 1 is purchased given that Product 8 is bought is 57%
Fraud detection & Loss prevention
Fraud and shrinkage is one of the most common challenges faced by retailers resulting in financial and consumer trust loss. It can originate with consumers, employees, or external sources. Different types of fraud include credit-card fault, fraudulent merchandise returns and shrinkage due to shoplifting, embezzlement and human error.
Predictive modeling helps in identifying unusual patterns of purchase and product movements that can help detect fraud and shrinkage. It also helps narrow down the categories and sale seasons that are most sensitive to fraudulent behaviour. The retailer can then take extra precautions to safeguard against loss among these sensitive categories and shopping periods.
Fraud Multiplier by Industry
Store
Operations
0 5 1.9 2.0 2.0 2.2 2.3 2.3 2.3 2.6 2.8 2.9 3.1 3.3 3.4 2.3 0 1 1 2 2 3 3 4 4 5 5 Housingwares/Home Furniture Automotive/Motor Vechile and Parts…Telecommunications or data service… Flowers/Gifts/Jewelry
Sporting Goods Computer/Electronics/Software Books/CDs/Videos/DVDs/Music Textiles/Apparel/Clothing Drug/Health & Beauty Office Supplies General Merchandise Stores Hardware/Home Improvement Toys/Hobbies Total
Category Sales Reporting & Analysis
Constant tracking of sales and regular reporting helps the sales force analyse category sales so that they can have an action plan before the next sales cycle starts. Also, it serves as the base for formulating sales strategies.
It helps in:
Identifying which categories, products and SKUs are selling the most in the store
Analysing consumer preferences and buying patterns in the store
Evaluating growth potential for product portfolio (categories, brands and SKUs)
Planning and managing store promotions
Evaluating the performance of the store by categories and SKUs on a regular basis
Enabling root-cause analysis in case of sales/profit decline: help identify the epicentre and rectify the
same -1% 2% 4% -5% 1% -1% 8% -8% 1% -1% 6% -5% 10% 18% 22% 50% Color cosmetics Skin Care Personal Care Hair Care Current Month Market Share YTD Market Share Change YOY Market Share Change 3 month MOM Market Share Change
55% 0.1% 0.3% 0.4% 1.0% 3.2%
50%
12% 32% 52%
YTD'11 Company1 Company2 Company3 Company4 Company5 YTD'12
Narrowing down on share loss within Hair Care category
52% 4.1% 1.1% 0.3% 8.7% 48% 30% 45% 60%
YTD'11 Brand1 Brand2 Brand3 Brand4 YTD'12
Further narrowing down on the brand(s) causing the share loss
Store
Operations
Workforce Analytics
Sales force, for a retailer is an equally important asset as the product that they sell. A good, experienced sales force yields higher consumer satisfaction and hence increased sales. It is therefore critical to optimize the employee recruitment, training and supervising process. Retailers can use analytics to increase productivity and can help enable an effective and sustainable retail workforce.
The advantages of work force analytics include:
Acquisition of talent- identifying the most effective employee attributes
Skill set mapping- placing employees in the ideal role based on their capabilities
Talent building- recognizing employee training needs in key skills and ensuring all employees meet store standards
Improve scheduling effectiveness- based on predictions of when and where consumers are most likely to shop, analytics can help schedule the most-productive employees appropriately
Retention- by understanding the key risk factors that drive attrition, employers can preemptively mitigate these risks.
Improve safety- detect the underlying causes to workplace accidents and rectify
Store
Operations
Opening of new stores
Site selection is crucial to a retailer and identifying the ideal location to open a new store has to be a strategic decision.
Integrating census data, which provides population and income data, along with survey data, providing demographic, psychographic and competitor store data, and financial data will give the retailer a better understanding on areas with the greatest potential. With this information, a strategic model can be built, which can help determine the best sites and best strategy for that area.
This process helps retailers identify: The ideal location for the stores
The type of store format that is needed in a specific instance Whether to remodel or not
What merchandising approach to adopt
Strategy &
Planning
Tracking Overall Performance
Retailers need to get a bird’s eye view on changing business conditions and emerging trends, and growth potential based on the sales and profits earned from their stores and categories. Accordingly they can adjust plans and forecasts to meet the new challenges and opportunities.
This requires a close monitoring and tracking process of the sales and financial measure of the overall market and then correlating it with the individual store performance.
It helps retailers:
Analyse market trends and buying patterns in the retail industry and identify the gaps and opportunities Evaluate and benchmark store performance on key metrics like traffic counts, conversion rate, sales per
square feet and sales per employee
Track sales activity for all outlets by region/sub-region/category Identify profitable categories in various regions
Monitoring & tracking performance across outlets
Strategy &
Planning
Store Clustering
Retailers need to customize their product and service offering to meet the taste and preferences to diverse cultural and demographic consumer segments. Implementing strategies at an outlet level will be operationally difficult to manage, while an overall promotional campaign and strategy for all outlets, despite being operationally more feasible will not be able to meet localized consumer needs. To counter this issue, retailers need to identify stores that exhibit similar demographics, locational proximity, personal income and shopping behaviours of local consumers and device a localized approach to run their marketing activities. Cluster analysis uses loyalty card transaction data and survey data to identify similar stores that form a cluster based on shopper demographic data and their shopping patterns. The retailer is then able to tailor specific promotional campaigns, assortment, planogramming, pricing and promotion strategies, store formats, layouts for servicing each of the identified clusters. This garners the retailer better returns on their strategies since it is more focused to shopper needs and increases consumer satisfaction due to the “personalized” approach.
Store clusters for a leading mass retailer in the US
Strategy &
Planning
Strategy &
Planning
Key Value Item Analysis
A few SKUs have a disproportionate impact on consumer price-value perception and can cause consumers to switch stores when those SKUs are not priced appropriately. These price sensitive items are known as Key Value Items, or KVIs. A retailer can use this knowledge to have a significant control over the items’ perceived price image and thus regulate the store’s image by carefully fixing the everyday pricing and the promotional pricing.
Key Value Item Analysis blends behavioral data (sales, household penetration, purchase frequency) and attitudinal data (consumer awareness of product, accurate price recall, price differential across similar retailers). The KVIs are identified across categories based on revenue coverage, price sensitivity, sales volume, and the role and prevalence of the item in the market basket.
By managing true KVIs through aggressive pricing, promotions, wide range availability and correct placement, retailers will be able to:
Influence consumers’ overall perception of the store Drive sales and footfalls
Gain market share
Factors determining a KVI Sales Volume Price Sensitivity Revenue coverage Sa le Price
Role & Presence in Market Basket
Price differential across retailers
Business Situation:
The client, a leading retail chain offering various products across categories, wanted to understand its customers to better plan customized campaigns and promotions with the objective of increasing customer engagement and overall revenues.
The Task:
Identify appropriate customer segments based on various factors such as purchase patterns, promotion response and demographics of the customers.
Framework:
Customer Personas:
Analytics in Action
Increasing Revenues by better Understanding Customers
Client: A Leading Retail Chain
Define & Build
customer segments Segment analysis Customer profile
Identified an appropriate customer base based on the #
of visits and days on books
Built customer segments using clustering algorithms after
treating the outliers
Analyzed the segments and identified the customer personas in each segment
Got a detailed profile of customer in a segment to
target for promotion
Who? What? When? 70% sales from FMCG & Staples Early morning Weekend Early Morning Weekend Shoppers Large family High visits 60% sales from FMCG & Staples. Multi-category shopping Afternoon to Evening High sales, large family shoppers Salaried staples shoppers 70% sales from Staples Shops in rice, oil, pulses and flour Morning to Afternoon 1st 10 days Salaried, Health conscious, staples shoppers Salaried Large family 50% sales-staples, 30%-FMCG. Multi-category shopping Morning to Afternoon Weekend 1st 10 days Salaried, large family,
weekend shoppers
Low visits Low sales and high margin 45% sales from Apparels Shops in Men’s casual and formal, ethnic wear Morning to Afternoon Weekend Weekend, apparel buying
shoppers Single family with kids Health conscious 70% sales-FMCG High proportion of baby care and health SKUs Morning to Afternoon single/small family shoppers Discount seekers 50% sales from Home needs. Shops in utensils, bed and luggage Afternoon to Evening Weekend Discount seekers 70% sales from Staples & FMCG Evening Evening shoppers The Result:
• Developed relevant Customer personas like discount oriented, large family, weekend specific category shoppers, impulsive buyers, high end buyers, etc
• Customer personas helped the business to appropriately target customers based on the day, time, affinity and category of purchase with appropriate promotional offers, leading to incremental revenues
Identify an appropriate Customer Base Small to Medium size families Large families shops mostly in FMCG and Staples
Business Situation:
The client, a B2B US-based retailer with presence in North America, Australia, Europe and Middle East noticed a significant increase in % Bad Debt for new prospect acquisitions; from 5.1% of total prospect sales in 2007 to 8.8% in 2011. Business wanted to manage this without compromising on lost sales.
The Task:
Design, develop and implement a robust predictive strategy that will help in quantifying the forward-looking risk at a Prospect-level. This will be a quantifiable and reliable benchmark for the business to leverage and decision whether to extend credit, go for credit card pre-payment, or completely avoid a particular prospect.
Analytical Framework:
Developed a Risk or Q-Score using firmographics and transaction information on prospects acquired between 2007 – 2011. Upon implementation, each Prospect had a risk score between 1 and 10; with 1 being the most risky prospects; and 10 being the least risky Prospects.
The Result:
• Based on validation results; the predictive model was able to significantly separate prospects who paid and who defaulted
• Prospects with Q-score = 1, 2, 3 are high risk prospects. These prospects on average have 2.2X default rate as prospects with score = 4 to 10
• Recommendations implemented by the business: Review of all orders > $100 for score < 4 and Prepayment for all orders > $200 for Q-score < 4
• Overall Bad debt decreased by 32% in the subsequent year
Analytics in Action
Prospect Acquisitions: Lowering Bad Debt by 32%
Client: A leading US-based Retailer to Small & Medium Enterprises
0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 14.0% 1 2 3 4 5 6 7 8 9 10 % B ad P ro sp e ct s Q-Score Model Random Model Results 0.0% 2.0% 4.0% 6.0% 8.0% 10.0% 12.0% 14.0% 1 2 3 4 5 6 7 8 9 10 % B ad P ro sp e ct s Q-Score Model Random Validation Results Defining Modelling universe Risk Tables Variable selection and Model development Build Model, assigning Q-scores
Prospects segments with limited data and low bad debts ignored (these segments
were treated separately) Risk factors calculated for each prospect variable such as location, transacted value, organization size by prospect segment, etc Variables (factors) with high information value (IV) to risk identified; inter-correlated
factors discarded to include most relevant factor in the model*
Q-score calculated based on model and assigned to each prospect
Val id ation an d M o d e l r e fi n e m e n t % P ros p e ct s w h o d e fa u lt e d % P ros p e ct s w h o d e fa u lt e d
Business Situation :
US-based manufacturer of personalized gift items; with presence across North America, Australia, Europe and Middle East. They have over 3,000 sku’s on offer; across 8 product platforms and 13 countries. SKU’s are supplied from Far East suppliers, with a 3-month lead time. And almost 60% of Customers purchase gift items for immediate consumption; with the remaining having a deferred shipment date through out the year. In addition, the manufacturer runs SKU-level promotions through-out the year, which result in SKU-specific demand. Critical to have accurate sku-level demand forecast so that all orders are met; while maintaining optimal inventories.
The Task :
- Develop a framework and relevant forecasting models for improving the forecast process and accuracy. - Obtain a robust and accurate SKU-level forecasting for each week over a year.
- Implement the predictive models so that forecasting is improved, inventory levels are optimal and customer satisfaction is improved.
Analytical Framework :
The solution was aimed at simplifying the process and improving the timeliness and accuracy of demand forecast: 1. Simplify and automate some of the current processes that were cumbersome and susceptible to human error.
2. Use statistical analysis to learn from historical trends, project future demand, and create an Early Warning System to predict weekly excess and stock-outs at SKU level.
3. Improved the existing process of predicting repeat business using cannibalization models and also provided shipment profiles with insights on patterns of how products shipped out to customers. This helped in placing timely and appropriate Purchase Orders with suppliers. 4. Provided dashboards for measuring forecast accuracy and also performance of shipments.
5. Adhoc analytics to support current forecasting, customer care and marketing decisions e.g. quantifying financial impact of late shipments
The Result :
• Better sku-level forecasts and ability to react faster to products becoming hits.
• Less obsolete inventory at the end of the year meant freeing up working capital and reducing waste. Lower stock-out rates also meant better customer satisfaction in addition to revenue. Since repeat customers are their main focus, this factor is critical in preventing unnecessary attrition.
• More scientific approach to forecasting, thereby eliminating any bias in subjective forecasting logic.
Analytics in Action
Improve Demand Forecasts. Sales up by $ 3MM, stock-outs down
Client : A US-based Manufacturer of Customized Gifting Products
Forecast Variance Sa le s gr o w th (P Y ) Over- forecasted Growing Under-forecasted Growing Under-forecasted Declining Over-forecasted Declining WC57001A, +, + TD72601B, , WC87901A, , -WC74846A, +, + WC57001B, , -WC59401A, +, + WC69501A, +, + WC62545A, , -WC93001A, +, + WA38001A, , WC62545B, , -WC58802A, +, + WC59004A, , -WC74846B, +, + WC30146A, , -WD25646A, +, + WC74903B, +, + WD06001A, , -WC74903A, +, + WC59401B, , WC59004B, , -WC83504A, +, + WD31503B, , WC80446A, , -WC28801A, +, + WC81001A, , -WC83945A, +, + WC28301A, , -WC87901B, +, + WA85401A, , -WC58802B, +, + WC88102A, , WD11202A, , -WC89803A, +, + WC83945B, , -WD01993A, +, + WC30146B, , -WD18006A, +, + -5000 0 5000 10000 15000 20000 25000 30000 -40000 -30000 -20000 -10000 0 10000 20000 30000 40000 1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 MAPE - 23.47% 2UCL 3UCL 2LCL 3LCL Variance 235386 359051 198157211658.2184196496.3737 0 50000 100000 150000 200000 250000 300000 350000 400000 1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031323334353637383940414243444546474849505152 Plan C FY_2009 YTD_2010 Regression 1 Regression 2 forecast
• Further, it was found that key style groups and colors remained consistent year-on-year
• And that Customer repurchases on deleted style groups and colors were not significantly impacted the
following year
• Recommendations made to rationalize marginal style groups and colors, leading to approximately 30% SKU
reduction and significant cost savings in business complexity and working capital.
5% 5% 5% 5% 11% 10% 8% 6% 11% 11% 11% 10% 10% 11% 11% 12% 26% 25% 26% 28% 30% 30% 31% 30% 0% 20% 40% 60% 80% 100% 2008-09 2009-10 2010-11 2011-12 13% 34% 64% 44% 15% 36% 57% 49% 0% 10% 20% 30% 40% 50% 60% 70%
Reactivation Continued SKUs Reactivation Discontinued SKUs
Customer Segment1 Customer Segment2 Customer Segment3 Customer Segment4
• Business had over 5000 SKU’s across a few unique Product Platforms. Each platform had SKU’s across many style groups and colors.
• Style groups and colors were investigated for marginal contributions to Revenues. • It was found that only a few
key style groups and colors accounted for over 90% of revenues
Client : A leading B2B Retailer of Personalized Gift Products
Blue Gunmetal Red Black Burgundy Green Rest
Analytics in Action
MANAGEMENT TEAM GLOBAL EXPERIENCE. PROVEN RESULTS.
Roy K. Cherian CEO
Roy has over 20 years of rich experience in marketing, advertising and media in organizations like Nestle India, United Breweries, FCB and Feedback Ventures. He holds an MBA from IIM Ahmedabad.
Anunay Gupta, PhD COO & Head of Analytics
Anunay has over 15 years of experience, with a significant portion focused on Analytics in Consumer Finance. In his last assignment at Citigroup, he was responsible for all Decision Management functions for the US Cards portfolio of Citigroup, covering approx $150B in assets. Anunay holds an MBA in Finance from NYU Stern School of Business.
Kakul Paul
Business Head, CPG & Retail
Kakul has over 8 years of experience within the CPG industry. She was previously part of the Analytics practice as WNS, leading analytic initiatives for top Fortune 50 clients globally. She has extensive experience in what drives Consumer purchase behavior, market mix modeling, pricing & promotion analytics, etc. Kakul has an MBA from IIM Ahmedabad.
ADVANCED ANALYTICAL SOLUTIONS
MARKETELLIGENT, INC.
80 Broad Street, 5th Floor, New York, NY 10004
1.212.837.7827 (o) 1.208.439.5551 (fax) info@marketelligent.com
CONTACT
www.marketelligent.comIndustry Business Focus Tools and Techniques
Consumer Finance Investment Optimization SAS, SPSS, R, VBA
Credit Cards Revenue Maximization Cluster analysis
Loans and Mortgages Cost and Process Efficiencies Factor analysis
Retail Banking & Insurance Forecasting Structural Equation Modeling
Wealth Management Predictive Modeling Conjoint analysis
Consumer Goods and Retail Risk Management Perceptual maps
CPG & Retail Pricing Optimization Neural Networks
Consumer Durables Customer Segmentation Chaid / CART
Manufacturing and Supply Chain Drivers Analysis Genetic Algorithms
High Tech OEM’s Supply Chain Management Support Vector Machines
Automotive Sentiment Analysis
Logistics & Distribution
Y O U R PA R T N E R F O R
D ATA A N A LY T I C S S E R V I C E S
Greg Ferdinand
EVP, Business Development
Greg has over 20 years of experience in global marketing, strategic planning, business development and analytics at Dell, Capital One and AT&T. He has successfully developed and embedded analytic-driven programs into a variety of go-to-market, customer and operational functions. Greg holds an MBA from NYU Stern School of Business