Excession Insight
2022 - 2024 · Chief Design Officer
As part of the product team, I prototyped, designed and tested an open source intelligence tool to explore massive datasets.
The problem
_____ ___ _ ______ ____ ______ __ ___ ____________ _________ ____ ________ ____ ___ ________ - ___ __ ___ ____ ___ _________. _____ _____ - ___________ - __ ___ ____ __ _____ ____ ______ _________ __ _____ _________ __ _____ ___ __ ___ ___ _______ ___ _______.
The product
We built a geospatial database based on novel computer science that could ingest and query datasets in excess of 200b records. In front of it we created a SaaS analytics tool based on a humane, sentence-like query syntax and a series of visualisations to let analysts ask questions and sift the results.
tl;dr
- Pitched the product idea to Leadership.
- Mentored Data Science as they built the prototype.
- Interviewed early users to build problem understanding.
- Facilitated design thinking workshops to help with ideation and ownership.
- Collaborated with the PM to build a value-based development roadmap.
- Iteratively designed the UI with Product, Engineering and customers.
- Proposed product hypotheses and faciliated discovery sessions.
- Fostered a collaborative relationship with Engineering, involving them early and encouraging their feedback.
Impact
122% ARR contract uplift agreed to by our "lighthouse customer".
Product was selected for a ________ _______ _________ and a 7-figure, multi-year contract was negotiated.
Used to track ___ ______ _____ in Africa, people smugglers in _________ and ________ __ _______. We received word that another customer had found a "smoking gun" using our product in one of their investigations, which is the ultimate goal in the intelligence world.
Key screens
- Insight: Analysts aren't particularly technical but the questions they want to ask of their data can be extremely complex.
- Solution: I proposed a human, natural language-like search syntax. The analyst was guided through the query building process one step at a time, with selector widgets to help with each data type.
- _______: _______ ____ _ ______ ___ ____ ______ ___ __ _________ ___________.
- ________: _______ __ _____ _____ ___ ___________, _ ____ _ ___-__-___ __ ________ ___ ______ __ ____ __ ___ _______. ____ ______ ________ __________ _______ _ ______ ___ ________ ____ ___ ____ __ ______ _____ ______ ___ ___, _____ ____ _______ ___ ___ _____.
- _______: ________ _____ _____ __ _____________ ____ _ _____ - _____ _________ ____ ______.
- Solution: I made locations a primary object in the UI. An analyst can draw any number of locations and use combinations of them in their queries.
- _______: ___________ _____________ _______ _______ __ ___ __ __ _____________.
- Solution: The Data team computed "dwells" for each device - places where they have stopped above a certain threshold. I used the visits data to flatten a device's movement into a simple timeline.
- Overlaps with other devices are highlighted, which was a recurring request in user testing. The Data team had been planning a server-side algorithm but it was proving difficult at scale. This purely front-end solution was almost as good and far simpler to implement.
- The day and night bands let us quickly explore another thorny data science problem - deducing "pattern of life". Analysts could visually scan for patterns using the banding for reference.
- _______: ______________ ____ ____ _____ __ _____ - _____ ________ ____ __ __ __________ ____.
- ________: ____ ______ ___, ___ ___ _____ _ _____________ _______ ___ ____ ______, ______ __ __________ _______ __ ____. __ ___ _______ _____ __, _ _____ __ ___ __________ _____ ____ __ ____ _____ ___ ________ ____ ______ ___ _____.
- _______: ___________ ___, __________ _______ _________ _____ _ ____ _____________ ____.
- ________: _______ __ ___ ____ ____'_ "______" _____, _ _____________ ______ ____ ____ _______ _____ __ ___ _______ ____ _____ ____ _____'_. ____ ____ ____ ___ ______ __ _________ _____ _____________ ___ __ _______ ___ _________. ___ "_______ _________" ____ ___ ___ ____ ________ __ ____ _______ __ ________ __ ___ ___.
Quotes
"You are doing things with Insight that are not currently possible." _______ _______
"This tool is really simple, really intuitive." ______
Process
Discovery
As part of a separate project I created a small prototype for a query interface to be used with geospatial data. The basic idea was sentence-style, composable queries that would allow non-technical analysts to ask extremely complex, operational questions.
With a data scientist, I did a feasibility study and pitched the idea to the leadership team. The data science team got the go-ahead to probe the space, and I mentored them as they prototyped ideas with Streamlit.
Our commercial team sourced some potential customers for feedback. Two of those saw enough value in it to pay 5-figure monthly licenses for access to the work-in-progress. We held regular in-person meetings with those customers, giving us access to detailed use cases and a forum to test ideas.
Building confidence
As confidence built in our understanding of the problem and our solution to it, a small product team was made available to me. I took the learnings from the prototyping phase and created a clickable prototype in Figma. We tested this both internally for strategic alignment and externally to assess problem fit.
I checked their feasibility with engineering to ensure I stayed realistic. I wrote test scenarios, along with the product manager, that we could facilitate with some users. I also mentored one of our customer success team members to increase our user-testing capacity.
I iterated on the mocks and the scenarios until feedback stabilised. At this point, leadership decided the signal was strong enough to pivot the whole company to focus on this opportunity. To create alignment and excitement, I presented the learnings we'd made through the prototype and final set of mocks to the whole company.
As a pre-cursor to full production development, I sketched and then prototyped visualisation options in Storybook with React and Mapbox, which were included in the v0.
Production
Before moving into the production phase, the PM and I chose one of our prototype customers as our "lighthouse customer" - essentially a development partner. We would use them to provide real-world feedback, while allowing us to deploy early and often without fear of broader repuational damage.
The MP and I took the mocks and broke the application down into value chunks. We used a combination of story mapping, card sorting and value dependencies. We refined these discrete pieces of value with engineering - with the goal of delivering each within 2 sprints.
As development progressed, I took each piece of upcoming work into a design phase.
Sometimes this was on my own, other times I'd run crazy-8s or similar group ideation sessions with engineering, data science and customer success.
I then hammered out the details, interactions and states.
For areas where my confidence was low, we conducted user testing sessions with our "lighthouse customer" - sometimes as static mocks, sometimes as Figma prototypes. I also reviewed designs with engineering for feasibility, iterating as required.
Once complete, I added the Figmas to Jira and fielded questions during refinement and planning, and remained available to engineers during implementation.