Lumi Pulse

4 min read • 2026

Fixing a Fragmented Quoting Tool

At Lumi, I brought broker quotes into our internal lending platform, and cut send-backs from 55% to 37%. Brokers now submit with all the information Credit needs to decide.

Context

Scaling on broken foundations

Lumi grew faster than its internal software could keep up with. For an SME lender with a $500M+ loan book, everything you needed to quote a business loan was scattered across Pulse (our internal SaaS platform), HubSpot notes, emails and screenshots. Processing times were bloated, and 55% of deals got sent back between Sales and Credit.

Lumi kept expanding anyway. When I joined, the broker quoting tool (BQT) was in development, letting brokers submit a business loan quote directly to Lumi through their own platform. Those submissions didn’t connect to Pulse, our internal SaaS platform, at all, slowing decisions and causing double handling.

My role

The team was lean: a PM, a lead dev and me, starting from a list of field requirements. I ran the project end to end and sequenced the work with the PM. I owned the final designs and pushed for research time to ensure we were solving the right problem.

End-to-end,

Prototypes

User research

Dev Handover

Challenge

How might we connect broker quotes to Pulse so everyone gets to a decision faster and more reliably?

BQT integration isn’t just for brokers. It helps BDMs build broker trust, RMs move faster, and Credit decide sooner with all the complete information upfront. We needed to make this work for all of them. (Also, this was my first project at Lumi, which provided its own challenges.)

Research

Figuring out how to maximise impact

I was new to lending, so I front-loaded the research. With limited time, I picked three questions that would tell me the most.

1. How does broker lending actually work?

I started with desktop research, reading sales guides and credit training manuals to get across the industry. One key thing I learnt -

Everyone wanted as much information upfront as possible. As the sales guides put it, “just add it to the note, you never know when it’ll be vital.”

2. Are we asking for the right information?

The requirements were a 'guide' so I tested them. Sales and Credit both assess deals using the five Cs of credit (character, capacity, capital, collateral, conditions), so I grouped every field the same way. Then I audited them against real usage. Unused fields, like website, went.

3. What slows a quote down?

The data showed where quotes stalled, but not why. So I interviewed 10 people quoting, assessing and selling loans every day. See the affinity diagram below.

Findings

Everyone had to hunt for the right context

Quote changes risk your reputation

Brokers stake their name on the quote they give a client. If it changes without a reason, they look bad and so does Lumi. Without that reason, Sales just blamed Credit.

Brokers never want to hear the computer came up with it.

Nobody trusted the data upstream

8 out of 10 interviewees said broker quotes arrive missing information, so Sales re-checked every one. Then Credit re-checked Sales. Every edit overwrote the last.

Broker data is never really correct.

Workarounds became the real system

Pulse had plenty of information, just not the right kind. The deal summary everyone needed didn’t exist. So teams built their own re-pasting the same details at every handover.

5 mins on a lot of things, over and over again, really adds up.

Experienced brokers barely need us

Regulars send near-complete applications. New brokers need far more hand-holding, and they’re exactly who BDMs are trying to win.

If brokers know what Lumi wants, they bring better details and keep coming back.

Iterating and testing

Testing what Sales actually looks at

I wireframed in Figma and used Claude to explore multiple concepts quickly. The screen is information-dense, so I used AI heatmaps to check the hierarchy held up, then ran four testing sessions with Sales, which led to the side by side variation you see today.

Quick Win

Letting sales see credit decisioning

Every Sales interviewee wanted to see why Credit made a call, but it sat behind admin permissions and wasn’t in the requirements.

It was already built and the devs had capacity, so I pushed it up the roadmap. It shipped in a week, which bought us time for the extra research. And Sales could finally explain a changed quote.

Saving and comparing quotes

Sales can see exactly what changed between the broker's quote and their own, without opening a second screen.

Research

Once Sales overwrote a broker's entry, the original was gone. So to explain a change, they'd dig through the application line by line.

Solution

Pulse keeps a locked copy of the broker's original. The latest version shows by default, with badges on anything that changed. The rest is a click away, so nobody wades through what they don't need.

All details in one place

Quote details puts everything Sales and Credit need to honour a quote on one screen.

Research

Quote context was split across Pulse and HubSpot, so Sales and Credit worked across two screens, and changes to key fields were easy to miss. Testing also showed that when a quote went wrong, both teams went straight back to banking.

Solution

I centralised the key information Sales and Credit use to honour a quote, with badges on any field that’s been overwritten. Key banking figures sit up front, with version history and the Prospect Hub a click away for the individual transactions.

Adding customisation to the quote calculator

One toggle shows a quote with or without property, so Sales can give brokers both options without reaching for a spreadsheet.

Research

90% of deals under $150K settle without property, but Pulse assumed property by default. So Sales pulled the numbers into a spreadsheet. And when unsure, they'd quote both ways anyway: "We can do $X, or $Y with a caveat."

Solution

A toggle flips between unsecured and property-backed, so both quotes are one click apart. It matches how BDMs already think about a deal: with property it's one number, without it's another.

Knowing who you're quoting for

Research

A broker’s experience shapes almost everything: application quality, process knowledge and how much help they need. Regulars barely need us, while new brokers need a great first impression. Some just want a ballpark, but Pulse had no way to log one, so they became $0 deals.

Solution

Each quote shows the brokerage, when the broker's last deal was, and flags anyone new. Workshops get their own state, so they're tracked without posing as real deals.

Behind the designs

Mapping the flow

Before designing anything, I mapped a quote’s journey from broker to Sales to Credit. It showed exactly where context dropped off at each handover, and every screen above fills one of those gaps.

Trade Offs and Reflection

What we didn't build, and why

Reflecting upon the process, if I did it again, I’d talk to brokers directly. Everything I learnt about them came through Sales. Overall, two trade offs were made.

Two quotes, not every version

Sales wanted every version saved, but that was out of scope for a stretched dev team. In practice they only ever compared two quotes, across two screens. So I designed for that first, with room to grow.

Summary first, detail later

Midway, a new PM took over and Risk added requirements that weren’t needed for launch. I used the research to show what Sales and Credit actually relied on, and we agreed to ship the summary first and roadmap the detail.

Outcomes within two months of launch

I came in thinking this was about connecting some fields. It turned out to be about trust: whether anyone could rely on a quote without checking it themselves.