In May 2017 artnet Auctions celebrated, what was at the time, a new world record price achieved for a work by the infamous street artist Banksy. After a fierce bidding war which spanned across 3 continents, 9 countries and 15 bidders, the final bid was placed by a collector in the Philippines and the final bid count was above 70.
While the business side was overwhelmed with joy and excitement about setting a new record and lift the market for works by Banksy to new heights, we on the product side were equally excited, but mostly about how our just released and updated bidding module was able to contribute to this success!
Artnet Auctions has become the dominant player in the online auctions art market, with 1MM unique visitors per year from over 180+ countries, placing bids in over 100 sales per year. All of this meant that the stakes were now higher, risk on the product side increased and critical parts of the system needed to be closely monitored and improved upon in order to support the growth trajectory.
The bidding module was such a high risk/ high reward project but its ultimate success was rooted in countless hours of hard work from our team which started at the beginning of the previous quarter.
The challenge
The bidding experience represents the core of every auction business, online as well as offline, and determines every outcome of every auction. The system itself focuses on the following jobs
- Placing bids on behalf of the buyer
- Track bidders and bid activity
- Alert users on their bid status i.e. highest bidder or being outbid
All of these jobs had the same goal: Maximize the value of the auction and ensure that the auction ends successfully with a winning bid at the end, while creating detailed logs for our internal reporting and data warehouse, with key metrics around bids placed and won, tracked by volume and value respectively.
The previous experience was closely modeled after traditional auction houses and their auctions, with a strong focus on placing individual bids with a specific amount at a specific time. As a user I had to type the amount into an input field Users who were outbid, would receive email notification when a higher bid was placed and then needed to place a new one in the system, making this a time consuming and frustrating process.
Automated bidding was available, but hidden in advanced settings and in most cases done by a specialist who would be in touch with their client throughout the process and had previously signed an agreement detailing the maximum amount they would be willing to spend, between 50k to 250k on a single artwork identified in our analysis.
Our research consisted of qualitative and quantitative methods, that I’ll go into more detail further below, but I want to share some highlights. We observed and identified key trends which posed significant challenges for our current bidding experience and the module itself:
- We saw significant increases in the average bid amounts that were placed
- A high percentage of new buyers were also new to purchasing art online and haven’t participated at traditional auctions before
- Our bidding logs indicated that a significant number of buyers would place their starting bids but not further participate in the auctions
- Many of our higher value winning bids were placed by our in-house specialists on behalf of their clients
These insights led the team to prioritize the bidding module over other projects on the roadmap, due to its criticality to the overall system but also due to its high impact, high reward potential.
We kicked off the project with an initial hypothesis that needed to be refined and tested:
Hypothesis: If we make it easier for our users to place automatic bids, then we would increase the average number of bids per artwork and ultimately GMV for the platform, due to less time consuming and more efficient workflows.
Audience
Artnet Auctions serves a wide range of users, from collectors who are just about to start their art collecting journey, to dealers and art advisors who professionally buy and sell art.
The bid module is at the core of the overall auction experience and had to address all our users needs and goals which meant we had to carefully conduct our research and gain a deep understanding of our audience to feel confident in our decision making.
My role
I led the project as a product designer on the team as well as lead UX researcher in close collaboration with our data science team. I also worked alongside the lead PM, lead engineer as well as a TPM who provided assistance with the system documentation to mitigate operational risk.
Design process
Research
The marketing team had done a fair amount of research previously in the form of an extensive audience analysis (i.e. demographics, age groups, professions, spending etc) so this served as our starting point and informed the basis of our research phase.
We conducted qualitative and quantitative research in the form of structured interviews. We made sure to gather users that were representative of our key audiences by observing Google Analytics behavioral/traffic data and bid activity logs from our reporting system to be able to get a clear overview of the full journey.
After 4 weeks of interviews and data analysis we were able to gather enough data to feel confident about our path forward and why we should prioritize and invest in this project.
Research finding highlights that influenced my design process later on were the following:
- Lower starting bids attracted more buyers and would make them emotionally invested in their purchase
- The majority of buyers have a budget in mind when they decide to bid on an artwork, influenced by previous auction results of similar artworks presented on the artwork page
- Artworks with large volume of bids, that were significantly higher than the average, were placed by our automated bid system
- Users were frustrated with having to manually enter and place bids
- A significant amount of users increasingly placed bids on mobile, which at the time wasn’t optimized
Design and prototyping
I started the design phase by translating our research findings into user journey maps and updating the existing bid schema flows that were provided by the TPM on the auctions engineering team. In a second step, I then sketched out initial versions first on paper and then as concepts in Sketch.
[insert bid sketches]
After design critiques, feedback and several iterations, which included key members of the auctions team, I felt confident enough to start testing our designs with real users for which we created a functional HTML prototype in a controlled testing environment with real data.
I wrote a test plan and scheduled several sessions with select users representative of our audiences as well as members of other artnet team who weren’t as familiar with the auctions platform to get feedback and validate our design decisions. The sessions allowed us to pay close attention to all scenarios as well as edge cases identified during our research phase and lead to refinements of my initial concepts.
For the overall design of the experience I focused on three key elements:
- Bid amount selection and confirmation
- Help dialogues and guides
- Mobile experience
Changing the way users would input their bid in the system was the first important step to improving the bid experience. Instead of hiding automated bidding in advanced settings we made it the primary way to input a bid into the system.
The result was a simple dropdown menu, where the user would select their maximum bid amount from a list that contains the next respective bid increment i.e. 50,000 USD, 52,500 USD, 55,000 USD, fetched directly from the bid increment table itself.
Another important step was to ensure that users would have all the information they needed while placing a bid. Help text and FAQ’s were previously available, but the majority of users in our research were unaware of their existence, or found it hard to find which led them to call or email customer service directly. We therefore made key information i.e. the most frequently asked questions in respect to the bid experience, more accessible by adding a CTA “How bidding works” that would explain the bidding process in a modal dialogue and help text was also added around the conditions of the sale.
Throughout the process, we spent a significant amount of time to ensure that users would be able to have a good experience while placing bids on their mobile devices. To minimize scrolling and have the bid button always accessible we fixed it in a bar at the bottom of the viewport which also included the dropdown menu selection.
Constraints
The main limitations that we had to account for, especially on the product and design side were in two key areas: Lack of system documentation regarding schemas and system logic, and limitations around testing and roll-out strategy.
We had to spend a significant amount of time upfront to gather documentation of the bidding system in order to feel confident updating the experience. Although improvements had been done over the years, functionality and processes have not been documented well and if documentation was available it was not compiled in one place, which we needed to address in order to mitigate risk. Understanding its inner workings was instrumental to not only design the best possible experience, but also to take into account any edge cases.
An additional constraint was around testing and rolling out the updated experience to the entire user base. The decision to focus on the max bid functionality represented a significant shift and also business risk if anything would go wrong. Each transaction was important which meant that we could not A/B test nor choose lower value auctions as a test bed to gather data and insights around its performance. After discussions with our counterparts on the business side we aligned on hosting additional “test” sales with real users who when selected would be enabled to register for the auction and participate in the sale, able to win an auction but would not impact the core business. We tested this way for several weeks.
Design validation
- Increased avg bid amount
- Increased overall bid
- Increased avg bids per lot
- Increased sell-through-rates per sale
- Banksy lot which sold to a collector from the Phillipines for 7x the estimate
Future optimizations
- Chandelier bidding
- Timed max bids which allow to be triggered at the end of an auction
- Pre bids to select group of clients
- Advanced notifications