' '

SOLUTIONS

PLATFORM    |     SOLUTIONS

RETAIL

Sales Forecasting

The business goal was to forecast the sales for individual items (both value and quantity) on a daily basis at the store level. Forecasts were made at the week level and allocated to derive a daily forecast, using historical day of week probabilities derived from historical sales data.

The model used eliminates outliers and factors in known promotion days and festive periods. The effect of week of the month was modelled and factored into the forecasts. An ensemble approach was used that uses a library of algorithms including time series models such as ARIMA based models, Holt Winter's Model (taking into account level, trend and seasonality), and decision tree based approaches. Business rules were also incorporated into the analysis to further fine tune the forecasts. The forecasts were adjusted to remove the stock out bias

The forecasts helped in revenue budgeting and operational planning and minimized the risk of stock outs.

BACK




Product Affinity Analysis

The business goal was to analyse product affinity to recommend optimal bundling strategies. The solution uses transaction (POS) data to perform market-basket analysis to identify items that sell together, based on the support and confidence of the association rules. Association rules could be derived based on a subset of relevant transactions such as transactions during a specific period, transaction for certain weekdays during a specific duration, transactions for a specific customer group and so on. The model also highlighted the number of lead bricks sold per dependent brick and this was a critical input into the bundling decisions. The model thus recommended optimal bundling strategies to increase revenue and when used in conjunction with CRM data, provided insights for identifying cross-sell/upsell opportunities.

BACK




Optimal Pricing

The solution focuses on identifying the right price and discount point that maximizes sales by performing what-if scenario iterations on various price points that have been observed historically. The solution takes into account the historic price-demand behavior at an Store-SKU combination, while superimposing the effect of inflationary shift in price levels over the history duration. The functional form of a price demand curve follows linear (or) power model as specified in the theories of Mathematical Economics, and the appropriate model is selected (at a Store-SKU level) on the basis of the log-likelihood parameter AIC . The solution factors in various parameters relevant to business, like the effect of Day of Week, Week of Month, Special Events, Offers and Promotions, etc. Some of the key benefits and value adds of the solution are - Increased sales and gross margins by optimizing the right mix of price, promotion and markdown strategies; Improved accuracy of balanced inventory levels and mitigate stock-outs; model future price and promotions.

BACK




Promotion Management

The solution uses a regression based approach to model the demand price relationship for a particular item at an outlet or at a group of outlets based on historical sales data. Additional regressors such as seasonality, day of the week and promotions are also factored in. Solution helps in driving pricing strategies to achieve desired revenue. When used in conjunction with known unit cost, the expected RGM (rupee gross margin) can be derived at a specific price point; this helps in crafting better promotional pricing.

BACK




Performance Analytics

The solution provides actionable sales performance indicators for executive management along with detailed view of the key indicators. The objective is to provide Sales and Marketing heads with region and product-line-wise operational figures such as Actual Sales, Margin (Contribution), Net Sales Volume, alongside projection figures like Planned Sales, Market Potential etc., and compare trends across time. Drill-down is enabled to track these metrics at granular levels (specific to a product or a location).

BACK




Inventory Optimization

Solution helps in identifying the high inventory concerns based on defined guidelines of optimum days cover and suggest inter-store transfer quantities based on certain business rules. The inter-store transfer rules recommended takes into consideration operational constraints like the geographic proximity between stores, transportation costs, time taken for transportation (for perishable goods), etc.

BACK




Sales Concern Identification

The solution uses a decision tree based approach to identify the driving KPIs contributing to the sales drop based on a YTD or MTD comparison. This includes analyzing KPIs such as sales volume, number of bills, average bill value, average selling price. Average footfall, etc. The solution also iteratively drills down the product hierarchy to identify the set of items that contribute significantly to the sales drop. Features include:

  • What-if scenario analysis capabilities to set various thresholds for determining significance
  • Facilitation of an interactive approach that guides the user towards fact discovery
  • Incorporation of business rules to exclude certain items from the analysis
  • Establishments of correlations with market intelligence data, when available
BACK




Merchandising Workflow Analytics

The solution identifies exceptions like bulky/non bulky articles, high days cover, high MSF flagging, highlight abnormalities in manual inputs shared by category for visibility and MSFs to be maintained, flag ARS/Not on ARS, flag zero MSFs, highlight inventory bucket impacted by floor min., and non selling articles, show whether MSFs are driven by category head/merchandising head input or system driven input, and identifies high days cover and low days cover due to high MSFs and low MSFs respectively.

BACK




Demand Forecasting

Solution provides daily demand forecast at the store-SKU level; Forecasts were made at the week level and allocated to derive daily demand forecast using the historic day of week probability. The model used eliminates outliers and factors in known promotion days and festive periods. The effect of week of the month was modeled and factored into the forecasts. An ensemble approach was used that uses a library of algorithms including time series models such as ARIMA based models, Holt Winter's Model (taking into account level, trend and seasonality), and decision tree based approaches. Business rules were also incorporated into the analysis to further fine tune the forecasts. The forecasts were adjusted to remove the stock out bias.

BACK




Personalized Target Marketing

The solution leverages the transaction data in conjunction with the CRM data to recommend personalized offers based on customer product purchasing propensity model that enables the business users to identify effective cross sell/up sell opportunities.

BACK