LexisNexis Risk Solutions · brought in to lead
Forty risk datasets. An analyst could see all of them and read none of them.
Map View is the tool a property insurer uses to price risk on a building. Flood, subsidence, coastal erosion, crime, fire, historic claims, satellite imagery, ground movement, each one accurate, each one a separate layer.
Which meant the decision came down to whichever layer an analyst trusted most that morning. The product had a data problem on paper and an interpretation problem in practice.
What the research actually found
The bottleneck was never data availability.
The brief arriving from the business was about coverage: more datasets, newer sources, better models. That is a reasonable thing to assume when you are the company that owns the data.
Moderated sessions with fifteen insurance analysts said something else. Everything they needed was already on the screen. What took the time was working out what it meant together, and that time was being spent on every single quote.
Support tickets and survey responses said the same thing from a different angle, so I measured it directly: the same analysts, the same tasks, timed before and after. That number became the one I designed against.
Interpretation time, not lookup time. Analysts were not searching for data. They were reconciling it.
The call
Organise the map around what an analyst has to decide.
Every version of this product had been organised around what the data is: a catalogue of layers, one toggle each, because that is how the datasets are licensed and that is how the teams are structured internally.
An analyst is not answering "what is the flood layer". They are answering "can I write this policy, and at what price". So the interface was rebuilt around that chain of questions, and the layers became evidence underneath it rather than the thing you navigate.
The consequence is the part that mattered to the business: everything the analyst had been deriving in their head, or in a spreadsheet beside the screen, is now surfaced as an inference on the map. Show the conclusion, keep the layer one click underneath it.
Who gets the front of the screen
Rank the use cases by how often they happen, then spend the screen accordingly.
Alongside the analyst sessions I ran internal research with the product experts: underwriting specialists, the data team, support, and the people who had answered every awkward question about this product for years. They know the long tail better than any user does, because they are the ones it lands on.
Together we listed every job the tool is asked to do and ranked them by real frequency rather than by who was asking loudest. That ranking became the navigation directly:
- The handful of things done every day are on the primary surface, visible without a click.
- The weekly and monthly jobs sit one level down, reachable but not competing.
- The rare ones are two or three levels in. Configuration, entitlements, historic audit. Still complete, still findable, never in the way.
The argument that took the longest was not about layout. It was accepting that a feature somebody had fought hard for could be genuinely important and still belong three clicks away, because importance and frequency are not the same measurement.
The interaction that had to work both ways
Some analysts think in a map. Some think in a list.
The same job gets approached from two directions. One analyst starts from a location and works out what is there. Another starts from a row in a portfolio and wants to see where it sits. Neither is wrong, and a tool that only supports one of them makes half its users translate before they can start.
So the map and the panel are two views of a single selection rather than two separate features. Select a property on the map and the panel scrolls to it. Select the row and the map pans and highlights. One state, two ways in.
The part that took the arguing was the actions. It would have been simpler to put them in one place, and every action is reachable from both. Whichever surface you were already looking at is the one you can act from, because being made to cross the screen to confirm something you have already decided is exactly the friction that adds up across a hundred quotes a week.
What became assessable
New sources only count if somebody can act on them.
Alongside the redesign, new data came in: satellite imagery, and patterns inferred by models across the wider dataset rather than measured directly. Both are genuinely useful and both are harder to present, because an inference is a claim, and a claim needs to say how much it should be trusted.
Each one had to arrive on the map with its own provenance, its own confidence, and a route back to what it was derived from. An analyst signing a policy cannot put their name on something they are unable to interrogate.
Done that way, the count of risk factors an analyst could actually assess rose by twenty four per cent, at building outline precision rather than postcode.
What changed
Measured with the same analysts, on the same tasks.
Not a survey sent after launch. The same moderated protocol as the research, run again against the rebuilt product, so the numbers compare like with like.
Measured in moderated testing with insurance analysts, against the pre-redesign baseline.
More factors assessable per property, from new sources presented so they could be acted on.
Time given back on every quote, by surfacing what analysts had been deriving themselves.
The part that is usually left out
This ran inside a company that did not particularly value design.
There was no research budget to speak of, no design team to fall back on, and the redesign had to be argued for against ingrained practice rather than commissioned. Halfway through, the pandemic moved the whole team remote.
Which is why the measurement mattered more than the design did. An opinion about visual hierarchy loses that argument. Twelve per cent of an analyst's time, measured on the same tasks, does not.
It is also why I still run research inside the delivery cycle rather than as a phase before it. When nobody is obliged to listen to you, the only durable authority is having already tested the thing.
"Luis had a challenge ahead of him to not only understand the complex business logic, but also to guide a development team through the implementation of the future vision. He challenged preconceived notions and clearly articulated the benefits of different, UX friendly approaches. Luis achieved this, despite the challenges of a company that does not fully value UX, and of course the pandemic that forced us into remote working."
Michael ReidProduct Owner, LexisNexis