McKesson Personalization & Optimization

Continuous improvement through data-informed design

Summary

McKesson has hundreds of products spanning a variety of audiences and care settings. With different business units constantly vying for space on the home page, I wanted to create a program that judiciously balanced business goals with user needs and displayed the most relevant content across a variety of visitor segments.

I designed and implemented a complex personalization and optimization program that used various visitor identification methods, including login cookie detection, Demandbase’s company identification service, and Adobe Target’s behavioral targeting functionality. Through continued optimization, the program more than doubled McKesson’s primary KPI’s including logins, revenue, and lead generation.

McKesson Homepage 2016

McKesson Site Relaunch, 2016

I joined McKesson in late 2015, partially to oversee a website redesign project already underway. A few elements of the site had been approved by the marketing committee (such as the homepage design) and the site eventually launched in March 2016. Overall, the new site was a major success, however, we knew there were few items that needed improvement.

In particular, the homepage hero banner contained a branded video that, although beautiful, contained no actionable purpose. While fine for branding the initial launch, it was difficult to argue any measurable business or user value from the video, and I quickly set out to make better use of this most valuable piece of real estate.

User Research

I began by examining the following data sources to better understand McKesson’s site visitors:

  • Web analytics
  • Search analytics
  • Voice of customer (site feedback)
  • User interviews

A next-page path analysis in Google Analytics showed that users were primarily navigating to the following types of pages from the home page:

  • Login
  • Product pages
  • About Us pages
  • Contact Us pages
Personalization Next Page Path Analysis
Voice of Customer Report

Voice of Customer Data

I launched a voice of customer program soon after joining McKesson. The first month of VoC data added complementary information to the web analytics, specifically, that visitors were primarily customers and prospects trying to perform tasks of high business value, such as:

  • Research products
  • Place orders
  • Receive customer support
  • Research McKesson company information

Aligning on Business Goals

Next, I conducted a study with company stakeholders to determine the top business goals for the website. The study was adapted from The Stranger’s Long Neck methodology by Gerry McGovern.

The top goals of the website, from business stakeholder’s perspectives, were to:

  • Enable users to find information about a McKesson B2B product
  • Enable users to place orders through a buying portal
  • Enable users to contact sales or customer support
Stranger's Long Neck Data

Research Summary

Based on both the user and business inputs, I established the following as the top user tasks on the home page:

  1. Login (customers)
  2. Research Products (everyone)
  3. Learn About McKesson (prospects, job-seekers, etc…)
  4. Contact Us (customers, prospects)

Using these four foundational user tasks, I mapped content that would display to the different site audiences.

Decision Tree of the Initial Personalized Experiences

  • I used login cookies to identify customers of our largest ecommerce portals (“Connect” and “SupplyManager”) and displayed the appropriate login functionality and product promos. Where appropriate, I targeted content to sub-segments of those particular portals (ie. family doctors vs. oncology centers vs. home care centers, etc…) With about $90B in annual revenue, these customers were McKesson’s most important segment.
  • Using Demandbase IP identification services, I customized experiences targeted to high value prospects, competitors, and internal employees
  • I ran A/B tests on some of the segments to find the optimal content and visual presentation
  • By the time I left McKesson, I had expanded the program to include additional McKesson customer portals, individual  companies (via our Account Based Marketing campaigns), and site visitors who had shown an interest in specific product types (behavioral targeting), for a total of 21 different personalized experiences.
Personalized Experiences Tree

Samples of the Final Experiences

McKesson Personalization SupplyManager1

SupplyManager Customer – Oncology and Surgical Centers Segments

McKesson Personalization SupplyManager1

SupplyManager Customer – Home Health Segment

McKesson Personalization Health Systems

Health Systems Segment

McKesson Personalization Anonymous Visitor

Anonymous Visitor

Results

This project transformed the homepage hero area from a branded video with no measurable user engagement to the most highly-engaged area of the site. Click-through rates, logins, transactions, and leads from the personalized experiences all grew steadily over the my time managing the program. Transactions and revenue for SupplyManager (the most popular McKesson ecommerce portal) through the personalized login experience more than tripled since the program’s inception. This showed that the personalized experiences were becoming more useful and habitual to customers over time.

By January 2018, this area had achieved the following engagement:

  • Click through rate for “SupplyManager” customers (logins): 61%
  • Click through rate for “Connect” customers (logins): 30%
  • Monthly product page views: 9,325
  • Monthly leads: 91
McKesson Personalization SupplyManager Growth
McKesson Supply Manager A/B Test Results

The Value of Personalization: Personalized vs. Non-Personalized A/B Test

I ran an A/B test on the largest customer segment (“SupplyManager” customers) to determine the effects of a personalized vs. non-personalized experience.

For the primary success metric (clicks to product promotions), the personalized experience showed a 111% improvement over the non-personalized experience (more than double the engagement.) Confidence intervals reached 99.96%.