CASE STUDY
The Future of Steel
ROLE
Lead Product Designer
Team
Principle Product Manager, Director of Product,
VP of Engineering
TIMEFRAME
2023
ROLE
Lead Product Designer
TEAM
Principal Product Manager, Director of Product,
VP of Engineering
TIMEFRAME
2022-23
ABOUT THE PROJECT
When Tesla has a problem, you know everyone has a problem. The way people buy steel today is mostly through emails and over the phone. Buyers send out requests for quotes (RFQ) for raw steel materials to steel suppliers. They then have to manage all the responses in their various formats to track and analyze opportunities. Buyers are usually under a lot of pressure for time and so the potential for the best deal is often lost. The Felux Buyer Procurement Platform (BPP) organized and automatically analyzed all quotes received in one place.
I built this for Tesla’s procurement team and the other Felux customers that need very exacting, pre-processed material with razor-thin time windows.
On top of this, I was also challenged to refresh the UI and design system. Most of these screens are a reflection of that work in progress.
THE PROBLEM
Steel buyers have to organize various quotes from suppliers in various formats across several email threads and phone calls. This creates several challenges to quickly normalizing the data for comparison:
- Inconsistent quote formats: Suppliers use various formats, making it hard to quickly locate relevant information.
- Diverse pricing structures: Quotes may use different units, requiring conversion to a common format.
- Communication barriers: Language, naming, and currency differences can cause misunderstandings or misinterpretations.
- Steel grade variations: Different grades or specifications need checking for compliance with required standards.
- Additional costs: Quotes may exclude extra costs, like taxes or shipping.
- Time-sensitive quotes: Prices can change daily. Buyers must work quickly to compare quotes before they expire.
- Human error potential: Manual normalization and short time frames increases the risk of mistakes affecting decision-making.
- Automation difficulty: Varied formats make creating accurate, efficient tools for data extraction and normalization challenging.
PROCESS
Being in a startup means priorities can change at a moment’s notice, there’s almost always a lack of resources and time to do a lot of user testing. It’s nice when you have the time and space to do the design thinking process by the book, but often that’s not the case and you have to learn what research will be the most efficient, and what steps can be skipped.
We had other customers that have similar pain points, and our own customer success and sales teams have talked to way more potential customers that matched our persona (we’re not building just for Tesla). There’s always some valuable place to start like that. I quickly interviewed anyone I could.
I created several low-resolution mocks with different approaches and information displays. I love these light mocks. Asking questions is great but, even if I only had 5 minutes with someone, the mocks help get some really insightful reactions really quickly, or if some data point was missing or not needed.
This set us up to be in a great place for when we finally did get to start with the Tesla team. We were able to show them better educated guesses that they could rip into instead of starting from scratch. They did not have time for that.
EARLY IDEATION
I like to design for scale. More than just simply fitting in lots of content, I think about the next and prior logical steps in potential workflows and complementary features that might support any design. I also like to think of all the major features and platforms within a company as a suite, even if not every customer uses all of them. Equally critical, I’m thinking about how I can use existing components and patterns as much as possible. This builds upon familiar and validated designs while also reducing dev work.
SOLUTION
We had done enough rounds of testing the early ideas to get some clear signal on what was needed. It wasn’t going to be enough to have a single view or to rely on filters. And trying to fit everything into a single page would create something enormously complicated and difficult to use. There were a couple solid universal needs that surfaced. I took inspiration from pivot tables with the idea being to make the data effortlessly viewable from different angles. I designed several of these views but we first focused on a single one to build first, get feedback, iterate and inform the other views before we built those. You never want to design and build all the things all at once, instead, always start small, test, iterate and repeat. These are the three views:
- By Supplier: Displays the main quote points of total cost and total lead time from each supplier in a clear and organized manner.
- By Item: Allows the buyer to quickly compare prices and lead times for each item in the request across all suppliers.
- By Scenario: Utilizes an algorithm that considers various factors, such as price, lead time, and desired quality, to construct complete orders out of the best combinations from cherry picking items across all of the quotes received.
RFQ Analysis - By Scenario
Finally, an “impossible” view. I asked every buyer if they could make a wish, what tool would they invent to make their jobs easier. Almost every buyer repeated the same desire to take all the best individual items from any supplier’s quote to see if there were any unseen optimal opportunities, and also to use as the ultimate comparison to know the very best possible deals. We heard this request constantly from all our customers.
By Scenario: Collapsed View
Same as the By Supplier view but with the additional supply scenarios the platform constructs from taking pieces from all quotes. There’s not always going to be scenarios that might work so this tab only appears when there is value without cluttering the By Supplier view.
By Scenario: Collapsed View
Same as the By Supplier view but with the additional supply scenarios the platform constructs from taking pieces from all quotes. There’s not always going to be scenarios that might work so this tab only appears when there is value without cluttering the By Supplier view.
Award Flow
Making Selections
After analyzing all quotes, the buyer can select entire quote or items from separate quotes. A summary of selections replaces the instructional banner with a CTA to Review and Award.
Review and Award
The buyer now sees a summary of material they are about to commit to purchasing by awarding the selected quotes. Awarding will generate purchase orders and notify the sellers.
RFQ Dashboard
We needed a place for buyers to manage and to provide some insights into all their quote requests. We also wanted these insights to be useful for a brand new member from their first quote. We started working on a higher-level dashboard for insights that are useful once time and many cycles have occurred that would live on the “Home” tab.
OUTCOME
Tesla was able to utilize our BPP platform and reduce their typical quotation time period from 2-3 weeks down to 1 week with 10-15% overall cost savings on raw materials. A remarkable improvement, especially considering how urgent and specific Tesla’s procurement requirements are.
Another early adopter, Playpower, in their first month of true adoption, saved over $6500 by utilizing the efficiencies of the BPP platform that they were unable to do on their own prior. This was through leveraging the full diversity of their existing supplier relationships, because they actually had the time to evaluate multiple quotes. A good sign that additional gains would be possible through the expanding of their supplier base, which without the BPP, would not have been possible to do.