Do You Have Your Black Friday Product Recommendations Ready, or Still Hunting? Personalizing the Shopper Experience This Holiday Season

• Digital Analytics
• eCommerce

Both Thanksgiving and Black Friday are rapidly approaching, and we have already started getting the sneak preview of 2016 deals. Last year, shoppers spent close to $4.45 billion online during this two-day period. Out of which, a whopping $2.72 billion was generated on Black Friday itself, which was 14% higher than 2014. A significant percentage of these two-day sales were influenced by the Electronics & Toys category, and on an average, a discount of over 24% was offered.

According to Adobe, the five best-selling electronics on Black Friday were Samsung 4K TVs, Apple iPad Air 2, Microsoft Xbox One, Apple iPad Mini, and Sony PS4. The five most popular toys were Lego Dimensions, Shopkin dolls, Lego Star Wars, Barbie Dream House, and Lego Friends. Consumers sought out the promotions on Thursday and Friday, with sales from limited online deals accounting for 40% of all online sales.

At Blueocean Market Intelligence, this is an especially busy time of the year for us because of our work with retailers and brands, that includes but is not limited to:

• Setting-up promotions and forecasting their impact on revenue and gross margins using our proprietary promotion simulator
• Identifying and promoting recommendations (featured products, best sellers, top rated, cross-sells and up-sells) across multiple product categories with an objective to significantly impact the UPT and AOV.
• Gauging the effectiveness of the promotions and recommendations against our forecasts and provide a daily and weekly effectiveness reports
• Last but not the least, facilitating a comprehensive Thanksgiving and Black Friday Merchandising scorecard with hourly refresh

Our preparation for the holiday season starts in August and continues until the end of September. During these two months, we conduct a series of deep-dive analysis on the last three seasons of holiday data, and factor in the ongoing behavioral and transaction trends, as well as, competitive data and revenue forecasts for the fourth quarter.

Our intent is to answer the following key questions from merchandising teams:

• What should be our baseline and control limits for the topline measures like revenue, UPT, AOV?
• Across multiple categories, how much flexibility do we have with our gross margins considering the competitive prices, promotions, rising freight costs and our own revenue targets for the fourth quarter?
• What kind of promotional deals would have a significant impact on the revenue, gross margins and profitability?
• What should be our promotional calendar for this season, keeping in mind the response to our last year’s holiday efforts?
• What product categories and brands have generated the most engagement and revenue from shopper’s perspective? What were the top five products under each category?
• What kind of merchandise assortment would be extremely relevant for our audience and generate better click-through and cart additions?
• What should be the acceptable cart abandonment rate for this season and what factors have influenced it over the last two seasons?

Our set of product recommendations from different sections of the retailer’s website is crafted leveraging spatial choice modeling techniques. These techniques are based on the principle that the customer preference similarity stemming from prior response behavior is a key element in predicting current product purchase.

The model is dependent upon two complimentary methodologies: joint space mapping (placing customers and products on the same psychological map) and spatial choice modeling (allowing observed choices to be correlated across customers). Using a joint space map based upon past purchase behavior, a predictive model is calibrated in which the probability of product purchase depends on the customer’s relative distance to other customers on the map.There are multiple variants of Spatial Models, including:

1. Market Basket Analysis – A wonderful, old-school technique, which allows the retailers to gauge the customer affinity towards different products and categories, and help them identify hidden associations that are not apparent.
2. Item-to-Item Collaborative Filtering – It’s another most widely accepted technique for product recommendations made popular by Amazon and Netflix. The basic principle based on item-based techniques which first analyze the user-item matrix to identify relationships between different items, and then use these relationships to indirectly compute recommendations for users.

Both above techniques rely heavily on the transactional data, and it needs to be cleaned and prepared before leveraging for the model building exercise. Here are some of the most popular applications from these results:

• Raw product recommendations in CSV or TAB format which can be leveraged to manually merchandise the products as featured products, cross-sells and up-sells on the product and shopping cart page etc.
• Customized product recommendation dashboard in Tableau or Qlikview to enable the merchandisers
• Customized API on top of the recommendation engine to connect with the client’s catalog management system
• Feeding the model output to the enterprise data management platform and building segments for re-marketing
• A tag based recommendation engine which can dynamically serve the recommendations at different sections of the website
• Personalizing the landing pages of the new and returning shoppers

To learn more about Blueocean Market Intelligence’s product recommendation approach and how it can help you to personalize the shopper experience this holiday season, please contact us. Happy Thanksgiving!