Cutting announcement publish time from 12 minutes to 5

Cutting announcement publish time from 12 minutes to 5

Role

Product Designer

Team

2 developers,

1 product owner

Tools

Cursor, Codex, Github, Figma

Timeline

2 weeks

Overview

A tool for publishing announcements to insurance advisors and brokers

InsurFact's internal ops team publishes announcements, product news, sales tips, webinars, that appear on the front-facing platform for insurance brokers and advisors. This tool is how they create and manage those announcements.

Before the redesign, publishing one announcement took about 12 minutes

The form was disorganized, content and settings were mixed together with no clear structure, and admins had no way to preview what an announcement would look like to brokers without navigating away from the tool entirely.

The Problem

A form built around the data model, not the admin user's workflow

The original tool surfaced every field in a single undifferentiated form. Content fields and configuration settings sat next to each other, multiple content types (HTML, PDF, video, URL) shared the same form layout, and there was no preview capability. Admins had to open the advisor-facing platform in a separate tab to verify their work.

Painpoint 01

No structure between content and settings

Everything lived in one form. Admins had to scan the whole page to find what they needed each time.

Painpoint 02

No preview in context

To check how an announcement looked to brokers, admins had to leave the tool and navigate to the front-facing platform.

Outcome

From 12 minutes to 5 without touching the backend

↓ 58%

Task time

2 weeks

Design to ship

Constraints

What I had worked within

01

No backend changes

I could only redesign the interface. Data structure and APIs had to stay exactly the same.
No new endpoints, no schema changes.

02

Admin side only

The advisor-facing announcement display was untouched. Any preview had to work within the admin context.

03

Existing tech stack

The product is built on PrimeVue with Tailwind CSS. Solutions had to work within what was already in place.

User Research

I observed how admins used the tool and organized three insights

I sat with an internal ops admin while they published a typical announcement.
Rather than asking what felt difficult, I watched how they actually moved through the task where they paused, what they re-read, and what pulled them out of the tool entirely.

Insights

01/
Scrolling constantly to re-check what was already filled

02/
Filling fields out of order, jumping back and forth

03/
Logging into advisor profiles to check how it looked

Design Goal

HMW help admins publish announcements faster by making the form easier to navigate ?

Design solution 01

From a modal to a dedicated page

The original add/edit experience lived in a modal, which severely limited how much information could be surfaced clearly.
I moved it to a dedicated full page. This gave enough space to separate content from configuration, and made room for a proper preview experience.

This was the single biggest structural change. Everything else depended on having enough canvas.

Design solution 02

Organize by what admins decide, not by data schema

The original form was structured around the database: fields appeared in the order they existed in the schema.
I reorganized it around the admin's decision-making process: first decide what you're publishing (content), then who sees it and when (settings).

Content fields moved to the left.
Configuration, audience targeting, scheduling, distribution, moved to a right sidebar.
Admins no longer had to scan the entire form to find what they needed.

Zoomable image

Design solution 03

Add preview as a tab

Admins were spending time navigating to the broker-facing platform just to check how an announcement looked. This was the biggest single driver of the 12-minute task time.

I looked at how other tools handled this. Most SaaS content tools offer a real-time preview that renders as you type.
That would have required building a new renderer in the admin context which is a significant engineering lift.
Instead, I proposed a tab-based preview that reuses the existing broker-side render.
Less engineering cost, same outcome for the admin.

Competitive analysis

How other tools handle preview

Before proposing the preview solution, I reviewed four preview approaches used across content management tools by evaluating each against the constraints of this project.

Zoomable image


Real-time preview was the ideal. But given the no-backend-changes constraint and the two-week timeline, tab-based preview using the existing render was the right call. Same result for the admin, significantly lower cost to build.


Real-time preview was the ideal. But given the no-backend-changes constraint and the two-week timeline, tab-based preview using the existing render was the right call. Same result for the admin, significantly lower cost to build.

Design tradeoff

Submit-first preview

This isn't real-time preview. Admins need to submit before the preview tab appears.

Building a live renderer would have required backend changes outside the project scope.
Instead, the preview tab reuses the existing broker-side render, so what admins see is exactly what brokers see.

The real win is eliminating the account-switching detour.
Five steps across two sessions became one submit inside the same tool.

Known limitation: if something looks wrong in preview, they'll need to edit and resubmit.

Final Design

Before vs. After

Zoomable image

Impact

"I don't have to switch back and forth anymore. I can just check it right there." — Ops team admin,

↓ 58%

Task time

2 weeks

Design to ship

The biggest time savings came from the preview tab. Before, admins would finish filling out the form, submit, then navigate to the broker portal to check how the announcement looked and often come back to make edits. That round-trip alone accounted for most of the extra time.

By adding preview directly in the tool, that loop disappeared. Combined with the clearer form structure, the average publish task went from about 12 minutes to 5.