Vision for retail data quality How data quality powers effective decision making in consumer goods retail
Introduction
Overview Aspects of Data Quality Why measure and improve DQ? Master Data in retail Bricks vs Clicks Why does it matter? The art of the possible Case studies: Issues Case studies: Successes Summary @guycuthbert @atheonanalytics #visualdataquality
Who am I? Neither a Data Quality professional nor a retailer! Computer scientist by background, visual analyst by experience Formed Atheon in 2005 Visual analytics since 2006, largely applied to consumer goods retail Our work sparkles from high quality data or exposes poor quality data
Aspects of Data Quality
Aspects of Data Quality Data are of high quality if they are fit for their intended uses in operations, decision making and planning. (J.M.Juran) Poor Data Quality ( DQ ) impacts 1. Automated systems and processes 2. Informed decision making 3. Effective planning This presentation focuses on points 2 and 3; decision making and planning
Aspects of Data Quality Transactional record Sales: EPOS Orders: purchase order, acknowledgement, delivery advice, invoice DQ focus on accuracy of record: Is this what we did? Master Data Person: customer, prospect, employee Product: name, colour, size, flavour, weight, volume, brand Location: store, depot, delivery location DQ focus on richness and completeness: Is this who/what/where?
Why measure & improve DQ?
Why measure & improve DQ? Ideally, we would all work with perfect data A Data Quality Index of 100% is highly desirable but almost certainly unaffordable Imperfection is a constraint The scale and impact of the constraint is worth knowing so that we can work within its limits Ignorance of data quality results in poor decisions and plans Measurement leads to management Movement from unconscious to conscious incompetence
Master Data in retail
Master Data in retail Most retailers have an accurate transactional record From electronic tills, through to sophisticated e-commerce carts Master data is less well managed In retail, there are three master data sets of primary interest: People Products Locations
Master Data: People Simple Name, date of birth, sex Intermediate Socio-economic group, addresses, interests Advanced Behavioural segment(s), frequency, recency Challenges: identity, data protection, multi-channel activity Opportunities: single customer view, the market of one
Master Data: Products Simple Name, category, supplier Intermediate Size (unit weight, unit volume, units per pack), brand, images Advanced Ingredients, source history, competitor pricing Challenges: new lines process, attribution & hierarchy Opportunities: flexible range, own label, mass customisation
Master Data: Locations Simple Name, postcode, type (store/depot) Intermediate Size (sq m), format, footfall Advanced Floor plan, classified space (linear space), customer flow Challenges: store compliance, data acquisition Opportunities: flexible format range, return on space
Bricks vs Clicks
Attitudes to Master Data Pure e-commerce demands and enables great data quality Absolute need for rich, consistent, comparable product information Self-service potential for customer s personal data Delivery location data essential to offer Physical retail has tended to lag behind but e-commerce and loyalty programmes are helping Data shared with trading partners often ignores data quality Grocery retail EPOS exchange, for example
Why does it matter?
Decision making and planning depend on data Drive to analyse everything; the world of Big Data but many organisations struggle to deal with Small Data We need to be confident in our transactional record We need to consider trends and patterns across transactions We need to understand detail in the context of the big picture Analysis generates a return on investment for your data assets How do customers shop my store? What is my optimal product range? Where should I open new stores?
The nature of decision making and planning There are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns; there are things we do not know we don't know.
The art of the possible
Interactive visual analysis of complex data
Case studies: Issues
Is all the data present?
Are we sure it s all present?
Do we have any duplicates?
Any other anomalies?
Do we have any outliers, and do we know why?
Do we have all stores?
Have we classified products correctly?
Northern region... or is it?
Year on year, sales are down or are they?
In fact, sales were up!
Case studies: Successes
Local product listing limits multi-channel offer International fast food retailer Multiple sites and brands at each location Keen to present clear and consistent offer to consumer Similar products e.g. coffee available in each branded outlet Variety of price points, offers 429 unique coffee SKUs across branded outlets Is our offer clear and consistent?
Local classification of product prevents comparison Each brand (and even, each unit!) recorded its own product data Differences in naming approach, brand name, spelling, sizing Analysis all but impossible; over 400 SKUs for coffee
Standardised product attributes aid understanding Grouping like products by type enables comparison of just 38 key groups Numeric size data enables comparison by litre, kg etc. Distinct Good / Better / Best offers identified and price points adopted
Sizing stock for optimal distribution National high street and online fashion retailer Stock allocation challenges impact profitability Rapid out-of-stock in some stores whilst excess markdowns in others Traditional approach to size allocation Fixed size ratio pack for each store Limited variation by format Is there a better way of distributing stock?
Normalising all size types shows marked large / small distinction Enriched product master data with normalised sizing XS S M L XL Distinct insightful patterns emerge from sales data; clear regional size bias New stock allocation rules defined and implemented
Own label opportunity from brand and size Emerging market general retailer Rapid store and range expansion Aiming to establish strong own-label offer Looking at best-practice from US / European markets Limited experience of master data management Simple spreadsheet product lists What should we develop?
Full range analysis to determine brand & category performance Data enriched with brand and size, allowing comparison across pack sizes and between brands Price, promotion and popularity used to define Good / Better / Best brand profile Analysis by brand and category identified gaps in existing range, enabling improved margin
Own label strategy teased from visual data analysis Own-label G/B/B opportunities identified, and prioritised according to category/margin targets New own-label brands introduced, with products introduced according to identified gaps 50% own-label participation growth inside first twelve months
Conclusions
Data Quality is the foundation for understanding Data Quality often suffers from invisibility Visualising your data brings it to life Exposing the good and the bad Making you aware of the issues and opportunities Collecting, managing and enriching master data adds flexibility New understanding of how your customers shop from you Identify opportunities for greater, or tighter, product range Examine the case for store openings and closures Improve assessment of competitor offers
Recommendations Invest in effective Master Data Management People Processes Technology Treat data as an asset Look at your data Explore it visually Measure its quality
Thank you Guy Cuthbert Managing Director Guy T. +44 0 8444 145 501 M. +44 0 7973 550 528 E. guy.cuthbert@atheon.co.uk