Building a PIMS from scratch
Wrangling the complexities of a Product Information Management System with a lean product approach, and saving $300M.
Background
In May 2019, Wayfair was at a crossroads with a critical piece of our vendor-facing platform - Catalog and Merchandising tools. These applications represent a significant part of the vendor experience, and are at the core of Wayfair’s eCommerce business, empowering our vendors to provide and manage data about the millions of products on Wayfair.com.
I was asked to lead a research project to inform a build vs buy decision. As part of this project, I proposed a timeline and research plan consisting of three main parts:
Competitor
research
April
Qualitative interviews
May/June
Quantitative analysis
April
I worked with the product manager (PM) and a junior researcher (JR) who was assigned to the project to identify a list of goals for the research we planned to conduct. We quickly jotted down ideas on a whiteboard and decided on a dual track approach:
Track 1: competitive research spearheaded by the PM
Track 2: recruitment for vendor qualitative interviews
Research process
Over 5 weeks, we conducted 11 intensive onsite and virtual interviews with our vendors, including US and EU vendors. These vendors varied in size from small mom-and-pop shops to large, multinational corporations with sophisticated supply chains; some of these vendors had Product information management systems (PIMS, both custom-built and off-the-shelf) and others did not. In order to efficiently use our time, we split the interviews so that the researcher conducted some in partnership with the product team, and I conducted others independently.
Synthesis
After all of the interviews were finished, I led two synthesis sessions with the PM, stakeholders, and research/design. To make the research more broadly consumable, I decided to store each observation in a centralized Trello board with a timestamp from the relevant interview. As we reviewed the observations and notes, we copied these into “theme” lists (ex. Product development is a split process between teams, tools, and sources for info). Is allowed us to recycle the same observation for multiple themes. Moreover, it has allowed other teams to leverage our raw data to conduct their own synthesis with their own lens.
After we had these affinity mapped groups, we worked as a group to refine some of these statements to be more precise. These became our 12 core user needs.
After we agreed on the basic user needs, I led a laddering exercise focused on abstracting up from these observational themes and user needs, to higher level insights. Based on this we arrived at three core insights: 1) Generating, Organizing, and Curating content is a complex and global process for our vendors, who are getting information from multiple sources and throughout the product lifecycle2) As Wayfair as developed new tools and expanded our schema, vendors have had to develop ad-hoc solutions to manage new data and deal with new needs as they arise. This had led to a patchwork solution that is brittle and inflexible3) The shift from a brick and mortar business model which generally accounts for the majority of vendor sales, to a click-driven (eCommerce) model has created a sudden need for far more content to market to digital consumers. Vendors are still “catching up” with the demands of this new model
Quantitative research
Afterwards, the Product Manager and Researcher developed a quantitative survey to measure the business impact of the initiative. While I was not directly involved in the synthesis, I provided guidance and support to the team as they designed the survey. We heard from >1600 vendors and found that there was a correlation between the perceived ease of catalog updates and catalog size and revenue
The lead product manager and I took the synthesized research (qualitative and quantitative) and presented it to senior leadership (Product Directors, VP of Catalog Product) during a recap meeting. We were asked to mock up some potential ways we could approach the problem.
Ideation
In response to the request from senior leadership to mock up and evaluate potential MVPs, I worked with the product managers and researcher to identify four key hypotheses we believed we could test. In our discussion, I drew whiteboard diagrams and visuals to help guide the team. Afterwards, I converted these diagrams into low fidelity mockups using Sketch. The purpose of these mockups was to drive decision making amongst the larger senior stakeholder team.
Prototyping
To accelerate our learning process, the Product Manager and I agreed to leverage available engineering resources build a data-driven working prototype that we could make available to a small select group of test users. This was because we were more interested in understanding how this MVP could solve our users’ problems than in understanding what problems we should solve. The product manager and I groomed a list of potential vendors that could be invited to be development partners; we wanted to find a group of three that were willing to participate and were a good representational cross-section of our user base.
Over the next 6 months, I worked closely alongside our developers to mock up screens, review what they built, and test this prototype with our three selected development partners. When I left Wayfair in 2020, the project was slated for an MVP launch sometime that year.
