Websites crashed without fast support

When websites—like during holiday sales—crashed due to slow support from agents, it cost big money

Downtime costs

~$9,000

per minute

$540,000

an hour

For Amazon

$220,000

per minute

Without fast support,  
these crashes kept users waiting and hurt sales.

I worked at Splunk, a tool that watches websites to catch problems using Web Vitals.

Big picture load

Largest Contentful Paint
(LCP)

How fast the main stuff  (like pictures or text) shows up.
Good is under 2.5 seconds; poor is  over 4 seconds.

First click wait

First Input Delay
(FID)

How quick users can click buttons.
Good is under 100 ms; poor is  over 300 ms.

Jumpy pages

Cumulative Layout Shift
(CLS)

If the page moves around unexpectedly.
Good is under 0.1; poor is over 0.25.

Why it matter

If these are bad, users get annoyed, leave, and sales drop—so fixing them is a big deal!

Support agents needed to fix issues fast to stop these losses, but they were stuck with scattered data.

Slow fixes wasted time finding clues

Average time to fix

Currently, at Splunk, agents took 5 minutes just to find the right clues to fix website issues

Area taking the longest

4 minutes were spent digging through a messy list of stats

—like “Page Views,” “Page load,” “LCP,” “CLS,” and “FID”

—to find things like “how many visitors” or “how fast the page loads.

Users searched right metrics in a long optional menu and opened tabs to easily compare metrics.

These stats were like puzzle pieces with no picture, scattered across teams.

Impact of delay

Users waited 30% longer for fixes
54% of users abandon a site that loads slowly
Cutting sales 7% per second of delay
$9,000/minute loss

I made finding clues fast for agents

Change I made

I saw that searching for stats took the longest (4 minutes!),  so I turned the messy list into easy tabs to make them quick to find:

Custom events

Special actions

—like clicking a button grouped as “how many, how many fail, how quick”

JS Errors

Mistakes in the code

—like when a button doesn’t work.

Speed

How fast pages load.

—like waiting for a game to start.

Long Tasks

Things that take too long

—like a page getting stuck.

Network

How the website talks to its computer brain.

—like sending messages.

Cluttered menu confused teams—I iterated to a clean menu.

Grouped stats into a short list to make them easier to find.

Reduces from 14 options to 5 one(4 Front-end, 1 Back-end)

Early try

Cluttered tabs with all stats, captioned “Confusing tabs, like a jumbled puzzle.”

Teams select a tab, such as “Speed,” to identify issues , like Chrome's 9,200 visitors and a poor LCP of 5 seconds.

Final fix

So I tailored the menu into tabs—front-end and back-end teams get their own stats which align with the exisiting system

UX Metrics

Stats about how the website feels

—like how fast it loads

Front-end

Stats for teams who make the website look nice and work fast

—like when a button doesn’t work.

Back-end

Stats for teams who keep the website’s engine running

—like waiting for a game to start.

Custom Events

Special actions

—like clicking a button—that teams track.

Tabs fit like a clear puzzle picture.

How agents check metrics to fix websites fast

Count the Visitors

(Page Views/Route Changes)

Look at how many people visit pages or switch routes.

—like when a button doesn’t work.

Time the Page Load

Page Load/Route Change Duration

See how long it takes for pages to show up.

If it’s too long (like over 4 seconds), users get annoyed.

Watch the Big Picture Load

Largest Contentful Paint - LCP

Check how fast the main stuff (like pictures or text) appears.

A 5.92-second LCP means it’s too slow—users leave!

Test the First Click

First Input Delay - FID

See if users can click buttons right away.

If it takes over 100 milliseconds, it feels laggy.

Spot the Jumpy Pages

Cumulative Layout Shift - CLS

Look if the page jumps around.

If it shifts more than 0.1, it’s confusing.

Tabs fit like a clear puzzle picture.

How they decide

The agent checks all these clues in the tabs I made.
If LCP is 5.92s, FID is high, and pages jump, they know the website’s a mess.
They pick the biggest problem—like the slow LCP—and fix it first.

Impact

What I Shortened

Cut fix time from 5 minutes to 1 minute by making stats easy to find.

Clicks Reduced

Reduced clicks from 14 to 4—agents now find stats faster.

How I Measured Success

Users waited 30% less for fixes, and sales loss dropped—54% of users don’t abandon the site anymore, saving $9,000/minute.

What I learned

I learned design can save businesses millions and make users’ lives 30% faster by fixing chaos with clear solutions.

Key takeaway

Quicker Insights:

Issues seen in seconds.

Tailored Fixes

Job-based stats fit their flow.

30% Faster User Experience

Reduced wait times

Revenue Saved

Stops big losses

Clear insights from chaos

—Speeding up website fixes