Guest intelligence is something Target use to drive marketing and merchandising decisions. Target, of course, is a large US retailer that does business in 49 states and has a major web presence. Target has guest data from panel surveys, store surveys and operational data and regards these all as valid sources of guest data, without attempting to resolve them 100% into a single source. The team focuses on “optimizing marketing and merchandising decisions with guest data and analytics”. The group has analysts, marketing campaign execution folks, data stewards and a team focused on rolling out and operationalizing new analytic capabilities. Within marketing the team focuses on guest insights reporting, web analytics, database and receipt marketing, SEO and media mix. Merchandising also gets guest insights and web analytics plus assortment optimization, adjacency optimization and planogram optimization. Three of these got discussed:
Guest reporting is very flexible but still just a report, though the plan is to bring it into the context of the decisions being made by the consumers of the report. Important to remember that the data has some key characteristics – it is not always additive (can’t add number of guests for this product to number of guests for that product as there could be overlaps), must baseline data as short time periods and unusual products tend to skew to better guests because only regular shoppers show up and can’t always compare year to year. Classic examples of need for some statistical understanding imposed on report consumers who might lack the necessary analytic skills. They also wait some time, a year for instance, before using data from a new store for example so that the data has time to mature.
Color receipt printing, in partnership with Catalina marketing, of marketing offers based on current basket and historical purchases. For instance some guests were purchasing artificial Christmas trees but not buying much trim. In this case the basket was used to print a discount coupon for trim. In another case they used historical patterns to grow categories in newly remodeled stores – general category growth if you had little history, very specific offer if you had purchase history. These campaigns had double digit (e.g. excellent) response rates.
Planogram optimization is the most complex, using optimization and predictive models to manage space optimization and product layouts within each store. Drilling down from stores to category (electronics) to departments (X-box) to specific products (games, accessories) – how much space goes to what and how to layout the store. Change the placement, size and location of each category and department for each store, because each store is a different size and shape. This whole process uses optimization and business rules (because there are lots of contractual rules).
Not only are there many guest markets and many floor plans, Target has 1,700+ stores, 40+ categories, 400+ departments and huge numbers of products. Very complex decisioning problem if you want to make this decision specific to stores not generic or average. Also needed to bring the “art” of merchandising to bear so that the results are “pleasing” while still boosting sales etc. Multiple data sources, rules from contracts and from buyers about presentation, objectives in terms of units/sales/presentation so can manage the tradeoffs. Interestingly this lets the buyer control the planograms without making them look at each one – they specify the rules they apply (helped by the team asking them “why did you place that there, why did you change/override that?”) and the business rules engine applies the rules for them. Classic decision management – stop experts making each decision and empower them to specify how to make the decision instead while breaking down into micro-decisions (one per store in this case). Target even go as far as using supply chain data to constrain how much shelf space a product gets so that they can guarantee to keep it stocked at that level.
Learnings:
- Embed analytics and reports into processes so they get used – find the decisions and focus on them
- Invest in data quality/data stewards
- Invest in training people to use the stuff you build
- Compete for the analytical talent you need
- Prototype, learn, develop. Then test, learn, scale
- Balance art and science of retail
Comments on this entry are closed.