Benchling Bioanalytical

01 / Problem

A complex, disjointed process

During drug development, scientists develop assays (or colloquially, lab tests) that can detect and measure things. They repeat the following process many times until they have everything understood, documented, and ready to execute.

After they're developed, these assays are run over thousands of samples. The workflow shifts towards importing, testing, and re-testing them using specific business logic (e.g. if result > x, re-test at dilution y).

02 / Approach

Designing plates where the work happens 

I partnered with my teammate, who'd recently picked up where another designer left off. They'd left us with some prior concepts which were a solid start, but hadn't considered how plates could plug into a completely new product area. It'd been planned as a legacy notebook feature, simplifying some of the build, but also constraining what we might realistically be able to ship.

Given that scientists use both tables and plates during experiment planning, my teammate and I explored inserting plate design as a third step in the flow.

It was well-received, but the integration would be costly. An alternative was what I called the "split path." The user partially fills out the table, navigates a few steps, and then designs the plates in the experiment's worksheets. It was inelegant, but it still gave scientists the functionality and flexibility to run their experiments. Here, the need for instant, clean data overrode UI polish.

Tests: stress-testing our task data model

Once an assay is finalized, it needs to be run in a tightly controlled sample testing workflow. During a clinical study, the system needs to automatically assign tests to samples and help scientists find, test, and re-test them until the final results look acceptable.

At first glance, it seemed like we had the existing building blocks in place, but this fell apart on closer examination.

Yikes. Our workflow engine was built to handle team-to-team requests, like requesting a sample from another lab. But test ordering needed things the workflow system didn't do: automatically assigning tasks to samples on receipt, tracking status through result evaluation, and encoding re-test conditions so a failed result could spawn a new task with the right parameters.

We circled around this problem for a while, scarred by a history of releasing features with unclear boundaries and overlapping use cases. Everyone wanted to avoid creating a duplicative, workflow-like thing for a single use case.

03 / Solution

From plate design to test results

For plates, remember that scientists need to define experimental conditions and then decide how they want to run them within individual plates. A detail I skipped before is that the assay procedure itself encodes definitions (materials, parameters, equipment) that can be reused across experiments, so we start here:

04 / Outcomes

Shipped, validated, and expanded

After months of work and weeks of rigorous user acceptance testing, we saw Merck go live with a public announcement and 400 users onboarded.

Other teams within Merck started asking about Benchling, with expansion opportunities following the successful partnership. With no sales effort, Bioanalytical picked up two more prospects purely through word-of-mouth. One deal got far enough along to validate the whole approach and was on track to close by the time I left Benchling.

Our dashboards showed more than just logins, too: scientists were running more experiments per user over time. Combined with what we were hearing qualitatively, it signaled real adoption: Merck scientists were working this into their day-to-day and getting the results they needed.

05 / Reflection

Leading through trade-offs

There are always trade-offs. We left friction in the flow that, with more time, we might've polished away. But the goal stayed in view: getting scientists end-to-end data in minutes instead of days.

The harder lesson was about supporting a design team through a project this size. Early on I got tough feedback: they didn't fully understand what was going on or why our approach made sense. Beneath that was a worry that in our push to reuse existing features, we'd ship something we weren't proud of.

I completely understood the feeling, but I didn't want our meetings to turn into therapy sessions. Instead, I pushed us to focus on what we could control and to celebrate the wins that did come up along the way. I set up new meetings and feedback channels, including directly with the core team and our product and engineering partners, so at least we could have more transparency into the decision-making.

What helped most was time with scientists. A giant table of test results might seem mundane to us, but getting it instantaneously really wowed our scientists. For them, having that capability felt like magic, even when we knew that with more time, we could've made it even better. I learned to live with the trade-offs and keep our focus on the outcome that mattered most to the customer.