The Black Box — The Sleep Company OS

The Experiment

I ordered one mattress and went through the entire journey as a consumer.

The Sleep Company

Order confirmed! Delivery between Aug 14 – Aug 18 🎊

a five-day "window" — be home on all of them, any day could be the day

Update: your order will arrive on Aug 19 (outside the window)

Track your order here → bit.ly/track-id

opens a page with a different date entirely — and no real tracking

…called support. "I'll email you back." (they didn't)

Engagement

12 msgs

Window

5 days

Accuracy

The Benchmark

Order confirmed — arriving Thursday, by 9 PM

Out for delivery — arriving by 6 PM today

Delivered — a day early

Optimized workflows

Immediate chat support Self-return Self-scheduled delivery Real-time GPS tracking

Precision

1 date

Efficiency

Early

Friction

0 calls

The Impact of Invisible UX

By eliminating the "anxiety window" and providing deterministic updates, we reduce support ticket volume by 42% and increase repeat purchase intent by 2.4x. This isn't just shipping; it's trust engineering.

The Reports

The dashboards said everything was fine.

The dashboards celebrated; the customers kept calling. I saw vanity metrics live and at scale.

Logistics reported

98% delivery adherence

→

Actually

3,000 open tickets nobody was counting

Support reported

90% CSAT

→

Actually

3.5 calls per unit sold — surveys only reached the happy endings

Leadership heard

NPS through the roof

→

Actually

Nobody owned post-purchase — a black box, end to end

That black box became my job.

The First Move

I Observed, I Questioned, I Wrote, I Planned — then I Built.

I researched and understood every action, the effect, the cause — and documented it for the first time in the company's history. That map led to where the money was leaking, which two numbers mattered, and the roadmap required.

01Document the lifecycle → 02Map the cost centres → 03Set the North Stars → 04Build the roadmap

North Star · Customer

Calls per unit sold

was 3.5

now 1.4

Get this one number down, and everything downstream — cost, trust, CSAT — follows it.

North Star · Business

People per process

Headcount was the biggest cost driver. Every process redesigned to need fewer hands — then, wherever possible, none.

The Roadmap

The flashy thing waited until the boring things were true.

Set up the roadmap

AI on chaos is just faster chaos.

Build the boring foundation

Unified data, defined flows, click-button automations.

Build the team

Product Support, APMs, product analysts

✓

AI ships — on solid ground

If a human agent can read the right data and act on it, an AI agent can too.

The Build

One operating system. Eight layers. In order.

Each layer created the data that revealed the next problem. Click a layer to explore it — built bottom-up, like any OS.

08 Agentic AI The culmination

07 The New OMS

06 RTO · Edit Order · Hold Order

05 Refunds & The Ledger

04 Returns & Replacements

03 The My Orders Page

02 Communication

01 The Unified Data Layer Foundation

add images/layer-01.png

Layer 01 · Foundation

The Unified Data Layer

Order data lived in four systems that didn't talk. I unified their webhooks and APIs into one master database — a single source of truth that everything else would stand on.

4 systems → 1 truth

↑ auto-playing — click any layer to pause and explore

The Range

For a year, every role this needed.

🎯Strategy & Roadmap

Found the problems, proved they mattered, owned the prioritisation.

📄Every PRD

Wrote every requirement doc, end to end.

✏️Every Design

Drew every screen and flow myself.

🔄Scrum Master + PO

Ran the sprints and owned the backlog.

🧪QA

Tested every release before it shipped.

🎓Trainer of 1,000+

800 store staff + 200 call agents, trained personally.

🛡️Risk & Fraud

Found the fraud gaps, closed them, dropped the policies that didn't serve us.

🚀Then: the team

Built the Product Support team, APMs, and analysts who run it now.

I was the first person on this team — so everything it needed was mine to do, until I built the people it could be handed to.

The Impact

What actually changed.

₹7Cr

total post-purchase
cost saved

Calls per unit sold

3.5 → 3.5 ↓ 60%

Time to process a return

100 hrs → 100 hrs ↓ 99.9%

Open tickets (exception cases only)

3,000 → 3,000 ↓ 83%

Returns-to-origin (RTO)

baseline → less than half ↓ 50–60%

CSAT — real this time, not the vanity number

· → ~90%

Accuracy of the delivery dates we promise

· → ~95%

Two years of work · Still compounding

I joined to fix post-purchase. I ended up building the operating system it runs on — and the team that keeps it running. All of it started with one bad delivery experience I couldn't stop thinking about.

Sharanya Chaturvedi · Senior Product Manager · The Sleep Company

The Black Box.

I ordered one mattress and went through the entire journey as a consumer.

The Sleep Company

The Benchmark

The Impact of Invisible UX

The dashboards said everything was fine.

I Observed, I Questioned, I Wrote, I Planned — then I Built.

Calls per unit sold

People per process

The flashy thing waited until the boring things were true.

Set up the roadmap

Build the boring foundation

Build the team

AI ships — on solid ground

One operating system. Eight layers. In order.

The Unified Data Layer

For a year, every role this needed.

What actually changed.

Want to talk about this?