OceanSofar Ocean2024

Making non-adherence explainable at sea

Captains deviate from routing guidance for good reasons. I designed the layer that let shoreside operators see why, instead of just seeing a lower score.

My role: Product design lead — research, interaction design, and data visualization
Focus: Product design, UX research, Data viz, User testing
Outcome: Beta with one customer, then released to all customers

Sofar's Wayfinder adherence dashboard shown over a photograph of open ocean water

Background

Sofar’s Wayfinder platform evaluates millions of routes each day and alerts captains and operators — in-app and by email — when a better path to port takes advantage of favorable forecasts. It also sends captains a daily RPM recommendation, because in changing conditions variable guidance always outperforms a constant power strategy.

The problem

Captains do not always follow the guidance, and they are often right not to. Conditions shift, and extenuating circumstances at sea don’t appear in a forecast model. But every deviation lowered the vessel’s adherence score, and shoreside operators were left looking at a number that told them that something went wrong without telling them what.

How can we allow shoreside operators to reasonably justify when captains do not adhere to our recommendations?

Framing it as justification rather than compliance was the pivotal decision. The obvious product response to a low adherence score is to push harder on adherence. That would have been the wrong tool: it treats the captain’s judgment as the defect, when the captain is the person with the most information about the actual sea state.

Finding the critical content

I asked customers what situations most commonly required them to depart from our guidance. Those reasons became the vocabulary of the feature — they gave operators a way to contextualize a captain’s decision-making rather than second-guess it.

Two versions of the deviation reason dropdown side by side. The shorter list has options including weather or sea conditions, marine traffic, mechanical issues, emergency, and ETA adjustment. The longer list adds RPM too low, RPM too high, and RPM sets off blowers. — The taxonomy came from customers, not from us. The second version splits out RPM-specific reasons after operators told us 'mechanical issues' was hiding three different situations.

Adding context to the data visualization

I introduced a column of notifications alongside the daily adherence graph, each one explaining what was occurring on a given day. Interactivity ties the two together: selecting an accordion item highlights the corresponding colored column on the graph, so the issue and its reasoning are visibly linked rather than sitting in two separate mental models.

Guidance adherence dashboard showing a 30-day line chart of RPM adherence with shaded bands, a scrollable column of dated notification cards to the left, and summary metrics for RPM adherence distribution and route suggestion acceptance below — Notifications sit beside the chart, not behind a tab.

The same dashboard with one notification card expanded, revealing the master remarks field and a reason selector for that day's deviation

Close-up of a master remarks entry in which a captain records that the vessel proceeded per Sofar route guidance, with the instructed RPM of 61 and expected speed of 11 knots — The captain's own words, surfaced verbatim. Paraphrasing them into a status would have defeated the purpose.

The design decision underneath this is small but load-bearing. The notifications could have lived in a separate tab, which would have been cleaner to build. Put next to the graph, they force the operator to read cause and effect together — which is the entire point of the feature.

Adherence dashboard for a vessel at 100 percent RPM adherence and 100 percent route suggestion acceptance, with the notification column showing routine entries — The dashboard also has to read well when nothing went wrong — otherwise operators only open it to find fault.

Outcomes

Released initially as a beta to one target customer. After several weeks of positive feedback, it shipped to all customers.

“One of the best tools you have prepared!” — Star Bulk