Harley
“Ok so, this seems odd. Let me share it with Deepti to see her opinion.”
Opens Notes → clicks Add Screenshot → annotates the suspicious timestamp directly on the chart.
Every 10 seconds, a sensor fires.
Every missed alert costs millions.
Across thousands of Walmart stores, IoT sensors monitor refrigeration, HVAC, lighting, and equipment 24/7. When something goes wrong, an energy analyst has minutes to diagnose it — or the alert ages into a work order, the work order becomes product loss, and the product loss becomes wasted electricity. $400 million a year was on the line. The tools were the bottleneck.
$120M/yr
Electricity Saved
Projected annual savings directly attributable
66%
Faster Resolution
Down from 15-20 mins to 5-7 mins
39%
Fewer Alerts
Achieved via intelligent auto-grouping
76%
SUS Score
Usability benchmark significantly exceeded
01
Meet the user
S
Santiago Ortiz Fernandez
Energy Analyst · 32 · Bentonville, Arkansas
“I usually end up resolving an alert which is already taken up by someone else. I have to move across multiple tabs and copy paste the data. I often forget what I am resolving for — and if I go back, I lose my progress.”
Santiago is an expert. He knows his domain cold — refrigeration cycling, compressor anomalies, HVAC patterns. He doesn't need the system to teach him. He needs it to stop getting in his way.
02
The number that changed everything
60%of users abandoned the task
when it was accidentally closed.
when it was accidentally closed.
The Cascade of Failure
Closed tab
→Lost data
→
Aging alert
→Work order
→
Product loss
8,243
Alerts per day
Constant noise masking true signals
$400M
Annual product loss
The cost of delayed interventions
15–20 min
To resolve one alert
Unscalable manual workflow
03
What Santiago was fighting
The interface was the enemy of the expert.
✕ Too little space to actually analyse
The graph — the most important tool — was squeezed into a corner. No room to think.
→ Collapsible segments + full-screen chart mode. The UI adapts to the investigation.
✕ Related alerts were invisible
Same root cause, five separate tickets. Santiago resolved one. Someone else resolved another. Nobody knew.
→ Auto-grouping related alerts into one ticket. One resolution for all. 39% fewer alerts overnight.
✕ Memory load across tabs
Data lived in one tab, notes in another, the graph in a third. Close one by accident — start over.
→ Everything on one page. Graph overlay to compare multiple series without switching.
✕ Collaboration happened over email
Screenshot the graph. Open Outlook. Paste. Write context. Hope they understand. They usually didn't.
→ Annotated screenshot sharing with auto-drafted email + deep link to exact alert with same filters.
04
Kill your darlings
The first design looked impressive. Testing said otherwise.
✕ Killed in A/B testing
Top Metrics Dashboard Bar
It looked like a proper enterprise dashboard. It ate the real estate analysts needed for actual investigation. Users didn't care how their team was performing — they needed space to diagnose their own alerts.
The screen belongs to
the investigation.
Not the dashboard.
05
The design
Everything Santiago needs. One screen. His way.
01
Quick View
Side panel for related alerts and notes. Never lose context. Never lose progress.
02
Graph Overlay
Stack multiple data series on one chart. Find the real cause without switching tabs.
03
Auto-Grouping
Related alerts collapse into one ticket. One resolution fixes all. 39% fewer alerts.
04
Kepler Belt
Defrost and suppression periods overlaid directly on the timeline. No more switching systems.
06
The head-turning feature
A collaboration story that designed itself.
1
2
System
Auto-drafts an Outlook email with the annotated screenshot + a deep link to this exact alert with the same filters Harley was using.
One click. Zero manual context-building.
3
Deepti
“Harley needs my help. Let me open and check this alert.”
Clicks the email link → lands directly on the alert with Harley's filters preserved → analyses → adds her comment in the notes panel.
✕ Before
Screenshot → Outlook → paste → write context → hope they understand → they don't → call → explain → re-share.
→
✓ After
Annotate → one click → Deepti lands on the exact same view. The investigation never breaks.
07
The impact
Zero aging alerts.
$120 million in projected annual savings.
A system that finally worked as fast as the people using it.
76%
SUS Score
Usability benchmark exceeded
~39%
Fewer Alerts
8,243 → ~5,000 / day
~66%
Faster Resolution
15–20 min → 5–7 min
100%
Aging Alerts Gone
Zero remaining in the system
$20–25M
Product Loss Prevented
Per year, directly attributable
$120M
Electricity Saved
Projected annual savings
08
What I took away
Design for the expert, not the tourist
Santiago didn't need onboarding. He needed speed. The best design decision was getting out of his way and giving him every pixel for investigation.
Collaboration is a design problem
Harley and Deepti were already collaborating — just badly. The work wasn't inventing collaboration. It was removing the friction from what they were already doing.
Kill your darlings with data
The metrics bar looked impressive in reviews. A/B testing killed it in 48 hours. Testing saved us from our own instincts — and gave the screen back to the investigation.
Real estate is respect
Every pixel given to decoration was a pixel taken from the analyst's investigation. In operational tools, space is the most valuable design material you have.