Firefly’s website now brings in ~15⁠–⁠17% of revenue. I rebuilt its catalog around how buyers choose.

Firefly’s website now brings in ~15⁠–⁠17% of revenue. I rebuilt its catalog around how buyers choose.

Firefly’s website now brings in ~15⁠–⁠17% of revenue. I rebuilt its catalog around how buyers choose.

Role

UX Designer, with Anchal Nagdev, then solo

Team

2 designers, Rohit Mali (Director of Sales and Marketing), MID (development partner)

Timeline

Jan 2021 to Jul 2024

Tools

Figma, FigJam, Miro, Photoshop

At a glance

The short version, in thirty seconds

The move

The catalog follows the buying decision

100+ products regrouped by scenario, power and flow, the order a buyer answers them in.

The system

One catalogue, two products

One catalogue sits behind a sales tool for ~100 users and the public website.

The honest part

No analytics baseline was kept

The revenue share is the sales team’s number for website enquiries that convert, not a tracked conversion rate.

01 · Context

Fire pumps are specified, not shopped. Buyers match a duty point, pull the datasheet and open a formal enquiry.

Fire pumps are specified, not shopped. Buyers match a duty point, pull the datasheet and open a formal enquiry.

Fire pumps are specified, not shopped. Buyers match a duty point, pull the datasheet and open a formal enquiry.

Firefly Fire Pumps has built fire-fighting pumps in Kolhapur since 1963 and sells to state fire services, the armed forces, refineries, port trusts and export distributors. Nobody adds a fire pump to a cart.

01
Match a duty point
a flow rate and a pressure class, from a tender or a scenario
02
Download the datasheet
the spec sheet is the product page
03
Open a formal enquiry
no cart, no price, a conversation

The three steps every buyer takes on fireflypumps.com. There is no cart and no price, so the site’s job is to start the right enquiry.

02 · Problem

The catalog followed the factory. Buyers arrive with a situation, not a series code.

The catalog followed the factory. Buyers arrive with a situation, not a series code.

The catalog followed the factory. Buyers arrive with a situation, not a series code.

Over 100 pumps sat in one flat list, ordered by mounting, material and series. A brigade, a refinery tender and a truck builder each had to learn the factory’s codes before they could find a pump.

How buyers choose
First question
The situation: a fire, a tender, a chassis
Then
Flow rate
Then
Pressure class
How the factory lists them
First question
Mounting type
Then
Material
Then
Series code

Buyer 01

Fire brigades

Which pump fits our fleet and the fires we fight?

Buyer 02

Procurement

Which model meets the duty point in this tender?

Buyer 03

Distributors

What is the full range, and how do we become a dealer?

Buyer 04

OEM body builders

Which pump mounts on this chassis and its PTO?

Buyer 05

Service and spares

Where are the parts and manuals for the pump we own?

Buyer 06

Forestry crews

What can a crew carry up a slope?

Six buyers, six different first questions. Before the rebuild, all six went through one flat list and one generic enquiry form.

03 · The sales tool, 2021–22

One record chain for ~100 people. Customer, inquiry, quote and order, with the catalogue beside it.

One record chain for ~100 people. Customer, inquiry, quote and order, with the catalogue beside it.

One record chain for ~100 people. Customer, inquiry, quote and order, with the catalogue beside it.

With Anchal Nagdev, I structured the internal sales tool that replaced each rep’s spreadsheets and PDF catalogs, and MID built it from our Figma specs. Reps found products faster and quotes went out faster. No analytics were set up, so there is no number.

~100

sales and operations users on one set of records

8

modules on one flow, from inquiry to handoff

4

catalogue layers, from product family to documents

MID

development partner, built from our Figma specs

D1Make the customer the root record and hang everything else off it
Excel had trained everyone to treat the quote as the document, so I made it the third link in a chain. A customer holds contacts, inquiry history, quotes, and orders. An inquiry holds requirements and its quotes, so a second inquiry from the same plant started with the site conditions filled.
D2Layer the catalogue so a shortlist never needs a datasheet
The PDF catalogue gave every attribute equal weight. I split it into four layers so each screen read only as deep as it needed. L1 is family and application fit. L2 is key operating specs: flow, head, pressure, power. L3 is the full technical set, including options and compatibility rules. L4 is documents, drawings, and certificates. Selection filters on L1 and L2, comparison shows L2 and opens L3, and the quote copies L2 and links L4.
D3Keep the model fixed and fit every new module to it
Every new request came in asking for its own screen, so we set one rule. A new module attached to an existing record type or extended the catalogue layers, and reused guide components first. Each got a walkthrough with Rohit and sales, then a build review with MID. The guide got a changelog and a recurring pass against what MID had built.
Make the customer the root record and hang everything else off it
Figure 2. The four record types and the handoff summary reading across them. Recreated in Figma in 2026 from memory, since the original files weren’t retained.

04 · The website IA

Scenario, then power, then flow. The catalog asks the buyer’s questions in the buyer’s order.

Scenario, then power, then flow. The catalog asks the buyer’s questions in the buyer’s order.

Scenario, then power, then flow. The catalog asks the buyer’s questions in the buyer’s order.

Scenario is portable, trailer mounted or vehicle mounted. Power is the engine: petrol, diesel or PTO. A segmented grid does the narrowing on arrival, so there is no filter to learn, and every product page leads with the spec.

D1Segmented the grid by pressure class, with no filter
MFV-LP above the rule, MFV-MP below. The narrowing happens on arrival, with no control to touch and no vocabulary to learn.
Flat 100+ item list→Segmented grid, three levels
D2Made the datasheet the product page
Model number, duty points and EN 1028 / NFPA 1901 lead the page. Downloads is a primary tab, not a footer link.
Pitch-first page→Spec-first page
D3Built the buying decision into the category page
Applications, plus a fourteen-criterion guide to choosing. It sits below the grid, so it serves a first-time buyer without slowing a specifier.
Segmented the grid by pressure class, with no filter
Vehicle-mounted category page. Normal-pressure models above the rule, multi-pressure below, so the grouping does the filtering.

Screens from fireflypumps.com as it runs today. The structure and enquiry routes are still live; a later vendor changed the visual layer.

05 · Enquiry routes

One form became three routes. Each enquiry now arrives with the model number.

One form became three routes. Each enquiry now arrives with the model number.

One form became three routes. Each enquiry now arrives with the model number.

The old site sent every buyer to one generic form. I split it into three routes by intent, and a product enquiry carries the model number of the page it came from.

Before · one form
Enquiry
01
Product
Sent from a product page, with the model number filled in.
Model MFT-1800-D
02
Dealership
For firms applying to sell Firefly pumps.
Route Dealer
03
Spares & Service
For parts and service on a pump already in use.
Route Service

Routes as they run on fireflypumps.com today: Product, Dealership, and Spares & Service.

06 · Solo to launch, 2023–24

Nineteen months as the only designer. Through the 2023 launch and the releases after it.

Nineteen months as the only designer. Through the 2023 launch and the releases after it.

Nineteen months as the only designer. Through the 2023 launch and the releases after it.

From January 2023 I carried the website alone: ten language editions, one of them Arabic and right-to-left, a 40+ component design system, and developer handoff and design QA with the external build team.

19

months as the only designer, through the 2023 launch

10

language editions, one of them right-to-left

40+

components in the website design system

50+

pages designed across the site

07 · Outcome

What it is worth today. The sales team’s number, not an analytics dashboard.

What it is worth today. The sales team’s number, not an analytics dashboard.

What it is worth today. The sales team’s number, not an analytics dashboard.

Rohit Mali, Director of Sales and Marketing, puts website enquiries that convert at about 15–17% of Firefly’s annual revenue, roughly $1.9M. It measures the whole channel, not the redesign alone.

~15–17%

~15–17%

~15–17%

of Firefly’s annual revenue now comes from website enquiries.

of Firefly’s annual revenue now comes from website enquiries.

Per Rohit Mali, Director of Sales and Marketing

In his words

“Atharva rebuilt our catalog around how buyers actually choose a pump, not how we build them. Enquiries went up, the wrong ones went down, and he stayed with it from the first interview through launch.”

“Atharva rebuilt our catalog around how buyers actually choose a pump, not how we build them. Enquiries went up, the wrong ones went down, and he stayed with it from the first interview through launch.”

“Atharva rebuilt our catalog around how buyers actually choose a pump, not how we build them. Enquiries went up, the wrong ones went down, and he stayed with it from the first interview through launch.”

Rohit Mali, Director of Sales and Marketing, Firefly Fire Pumps

08 · Reflection

Structure moved the number, not the visuals. Next time I would measure from the first day.

Structure moved the number, not the visuals. Next time I would measure from the first day.

Structure moved the number, not the visuals. Next time I would measure from the first day.

Two lessons I carry forward, and the limits on what this page can claim.

Lesson 01

Model the records before the screens

A screen can be redrawn in a day. A record chain that ~100 people have filled cannot.

Lesson 02

Keep a baseline from day one

No analytics baseline was kept, so the redesign’s own effect cannot be split out from the channel’s.

Limit

The figure is the whole channel

~15–17% covers every website enquiry that converts, not only the share the redesign created.

Limit

The visuals have changed

A later vendor replaced the visual layer. The structure and the enquiry routes are still live.

The revenue share comes from the sales team, per Rohit Mali, Director of Sales and Marketing. It is not a tracked conversion rate.

Next case study

All work →

NurseShift analytics on a laptop with the handoffs view on a tablet

NurseShift

NurseShift

NurseShift

Structuring shift handoffs to cut nurses’ cognitive load

Read the case study →