Analyzing Historical Sales Data From All Sources, Our Big Data Predictive Analytics Algorithms Generate Accurate Predictions Of Future Buys Of Customers. A Unified Retail View Across multiple Business of Customers and Products can be created.
Process of generating future buys of customer involves analysis of past purchase histories of customers and items. To derive meaningful analytical insights repeat sales history of same item is required.
While huge repeat sale history of any single item is present for top 1% like Amazon. For most of the retailers this is an uncommon luxury.
By the time enough sales history of an item is recorded. That item is no longer in the inventory.
An out of stock item can not be recommended to customers. Nor can it become part of other items "people who bought this also bought these" recommendations.
One of kind sellers like jewelers where every item is unique can be an extreme case of frequently changing catalog problem.
For a new item or a new customer there exists no previous data.
Hence no predicted buyers of the item or item to item recommendations can be made on day one.
A new product in the catalog, because of lack of its own prior sale history can not be recommended to customers. Nor can it become part of other items "People who bought this also bought these" recommendations".
A retail business is characterized by one or more types of products with different features.
Exact values of different features combined together is the primary buy decision making factor for customer.
Unlike the content based recommender system where every feature value like "color" is treated as independent decision factor. In our model the exact values of all features that has occurred in all historical sources of customer preference data forms the character of retail business.
Affordability of the same customer can be different for different types of products.
A person who likes to buy most expensive Diamond Solitaire ring may be willing to buy quartz watches only in the lower price ranges. For predictive analytics purposes a classification of buy within just a single defined price range is also not sufficient. Just above or below the defined price range the person still has interest in that specific product type having same decision making feature set value.
Some customers' have a tendency to be more sensitive to the amount of discount in making the buy decision compared to others. This tendency is not only restricted to the customer dimension but can be generalized to include the product dimension also i.e. some products have the higher likelihood of being bought with deep discounts while others are not so sensitive to discounting.
A discounted sale record should have lesser impact on process of generating predictions, compared to a sale record at full price.
Furthermore for marketing purposes customer segmentation on the basis of Frugality consideration can be very effective for different business objectives like clearance sale or premium product launch.
Buying behavior of customer changes over period of time. This is intelligently captured in our system.
Seasonal purchasing of different products is different. We note it down so that retailer can utilize this information.
So far we discussed five fundamental factors affecting the retail business. For deriving analytical outputs and predictions the data set available must have all the underlying attributes present in product data or customer preference data.
If the product data does not have some of those attributes explicitly defined, or even if some are present. Similarity between the products can be obtained using Natural Language Processing (NLP) on various text based product attributes. Cosine similarity and KMeans clustering provides additional dimension for analytics.
Unify the customer preference data from all historical sources including offline and online store sales history, shopping cart, browsing history, social media etc. Each data source can have different importance or weight.
Exact values or content based approach on product present in all historical data sources, of all different feature sets and corresponding affordability price ranges forms the structure of data used for collaborative predictive analytics.
Affect the strength of sales record contribution to predictive analytics data structure.
Applies Big Data technologies to handle scalability.
Sparsity in Unified abstracted predictive analytics data structure is handled by Dimensionality reduction techniques. Effect of frequent purchases of a small set of products or customers buying all products is handled by normalization techniques.
Along with predictions Seasonality of Products and Frugality of customers (as well as products) is noted down. These are powerful customer segmentation criteria for RetailReco campaigning system.
Portfolio Management, Investment Value Optimization, Personal Stock Advisor, Portfolio X-Ray
Prediction of future most likely diseases of a person & Personal Preventive Plan.