Open Source Analytics for Smarter Pricing Decisions in Retail

• A bit about pricing in retail
  
• Key considerations for strategic and tactical retail pricing
• The role of (open source) analytics in pricing
• Case study
  
• http://rpubs.com/KakasA09/FARCon082015
  

• Smart pricing tactics > cost reduction efforts
  Source: Managing Price, Gaining Profit
  

• 2014 Income Statement components for top retailers
  
  • (Revenue - Cost of Goods Sold) = Gross Profit
    
  • (Gross Profit - Selling, General & Administrative Expenses) = Operating Income
  • (COGS + SG&A) = Operating Expenses
    
  • (Revenue - Operating Expenses) = Operating Income

• EDLP vs. High-Low vs. Hybrid
  • EDLP: Walmart, Winn-Dixie, Home Depot, Aldi, Costco
    
  • High / Low: Khol's, Meijer, many mid-size grocers
    
  • Hybrid: Publix, Giant, Fred Meyer
• National vs localized
• Brick & mortar vs. online channel parity
  
  • Or strategic differences to exist
    
• Manual vs. automated price setting
  

• Delivering on financial results for key stakeholders (duh!)
• Price competitiveness
  • Where do I need to be competitive?
    
  • Where do I need to beat competition?
    
  • Where is it okay to be higher priced than competition?
• Price perception
  
  • Can I raise price and still improve consumer perception?

• Pricing strategy needs to support product / category / business unit goals
  • Improve market share, revenue or margin?
• Price wars usually do not end well
  
• Departure from core pricing strategy needs to be done carefully
  
• Price match guarantees are great PR and effective strategy to self-segment price conscious customers
  

www.dilbert.com

• Price increase is not just about raising prices
  
  • Product mix shift can be powerful
  • New product innovation at a higher price per unit (per ounce, per lbs, etc.)
• Price implementation needs to be systematic, fast and flexible
  
  • Controls must be in place
    
  • Errors flagged before being implemented
    

• Pricing information is public
  
  • Retailer APIs (look at Terms & Conditions first!)
    
  • Company ecommerce channels (website and mobile)
    
  • Price aggregator / comparison shopping sites
    
  • High level price figures from information providers (Nielsen, IRI)
    
• Retailers are aware of competitors' prices
  
  • Weekly, daily, sometimes hourly
    
• Web scraping is taboo, but everyone is doing it
  
  • Explosion of specialist firms in recent years
    

• 40-60% of retailers engaged in analytical initiatives to improve pricing (and promotions) analytics
  
• ~50% of retailers remain at opportunistic and ad-hoc stages of maturity
  
  • On-demand analytics in support of projects and initiatives, not real-time analytics in support of ongoing decision-making
    
  • Data integration takes time
    
  • Lots of manual effort
    Source: The Big Data and Analytics Opportunity in Retail: Where's Your Business?
    

www.dilbert.com

• Pricing analytical flexibility and creativity
  
• Attract top talent
  
• Ensure latest statistical and machine learning methods
  
• Cheap(er)
  
• It is becoming more and more popular
  
• Open source and commercial software can co-exist
  

• Data-driven pricing decisions > {heuristics, managerial folklore, gut-feel}
  
• For most physical retailers, simple analytics can make a big difference
  • High-school math analysis: gains/losses, ratios, weighted figures, relative differences, etc.
    
  • Essential statistics: linear and logistic regressions, clustering, decision trees
• Real-time descriptive price analytics is most critical
  
  • What happened in the past until now?

• Point-of-sale (POS) data
  • Basis for real-time descriptive analytics
    
  • Evaluate price sensitivities (elasticities) to determine where to right-size pricing
    
  • Cluster products and stores based on revenues, profitability, market share, econometric data, etc.
    
  • Determine effectiveness of promotional strategies
• Competitive pricing data
  • Ensure price competitiveness where it matters…(RE: price sensitivities and price perception)
    
  • Monitor competitors' compliance to industry-specific pricing regulations

• Online / clickstream
  • Customer comments for online ratings and reviews to determine which products drive price perception
  • Product pageviews and conversion rates for descriptive analytics and clustering
    
• Social media (twitter, blogs, etc.)
  
  • Social media feeds to evaluate how consumer sentiment changes with certain pricing strategies
    
• The goal is to positively impact market share, revenues, margin and price perception

• Simulated data to mimic real aggregated POS data you would obtain internally
  
  • Geo and demographic data obtained publicly
    
• Primary analytical tool is R, with some Python
  
• Data sets and full code for presentation is publicly available at https://github.com/KakasA09
  
• What can we do with our data to drive decisions (and ideally, $$$)?
  

• Price elasticity definition:

…it gives the percentage change in quantity demanded in response to a one percent change in price (ceteris paribus, i.e. holding constant all the other determinants of demand, such as income). (source: wikipedia)

• Several methods for estimating price sensitivity (none is perfect)
  
• Log-log linear regression standard in retail and CPG
  
  • Coefficient for log (price) is price elasticity
    

• Visual analytics can easily be 5 dimensional
  
  • …and reproducible
    
  • …and value-added!
• Visual analytics dimensions are more than just axes
  
  • Use scatterplot3d package in R to help you out
    
• Let's look at the raw data again:
  

• Let's take a look at our data again
  
• maps and ggplot packages enable you to visualize results and performance trends at the zip code and above levels
  

• Social media, blogs, and more importantly consumer ratings and reviews can and should be analyzed for insights
  
• In pricing, we care about evaluating which products drive price perception
  • To what extent do they drive it?
    
  • How does price influence price perception?
• Most consumer reviews for retail products are fairly explicit
  • Again, simple analytics can drive powerful insights
  • No need for complex text mining (unlike for blogs, articles, etc. where context is important)

• Original data set of 12,000 tweets about Amazon Prime
  
  • Obtained via R's twitteR package
    
• Rudimentary lexicon based sentiment scoring + popular word and tag (ngrams) visualizations created with:
  • tm package in R for essential text mining
    
  • dplyr and stringr for data munging, string manipulations and sentiment scoring
• Hourly stock prices through a simple Python script
  
• Variables and dimensions of revised (pure) tweets
  

[1] “text” “created” “date” “hour”
[1] 8615 4

• Pricing is the most powerful business lever (and it can be fun!)
  
• Be price competitive where you need to be
  
  • Don't give up $$$ unnecessarily
    
• Simple things will make a big difference in business results
  
• Develop your price analytics core first
  
  • Near real-time, descriptive and exploratory analysis
    
• Listen to what your customers are saying (commenting, tweeting, blogging)
  
  • Analyze them and adjust strategies accordingly