The 12% Drop Case
A metrics story

The morning orders fell 12%

Eight questions took me from a red number to one broken step.

Orders placed, dailyAlert
-12%vs. before

By Geetika Gupta

Chapter 1

Everyone has a theory

A competitor sale? Bad weather? Just a slow week? Prices? Festival season?

I chased none of them. I shrank the problem one question at a time instead.

Chapter 2 · Question one

When did it start?

Daily orders placed, indexed to 100. The shape is illustrative.

A cliff, about two weeks ago. A slow slide means a habit changed. A cliff means something changed.

Chapter 3 · Question two

Is the number real?

REAL
Analytics dashboard
Backend orders
Payment records
BeforeNow

The backend and the payment records agree with the dashboard. A broken tracking tag didn't cause this. The orders really are missing.

Chapter 4 · Question three

Cut it open

A 12% fall is an average. Averages hide things.

-12%

Each dot is a slice of orders. Hollow red dots are the orders we lost.

PlatformCityNew vs returningTime of dayOrder type
Chapter 5 · Turning point

It was Android

Android

-19%

iOS

-1.5%

Competitors, weather and prices would hurt both phones. This hurt only one.

Chapter 6 · Question four

Follow one order

Open app Browse restaurants Add to cart Checkout Payment Order placed They reach the cart, then leave here.

Everything before checkout looked normal. The orders were falling out of checkout.

Chapter 7 · Turning point

Inside checkout

✓Cart
✕Confirm delivery locationCould not confirm
Payment
Place order
9% of location checks fail in large cities

Address and map handling is harder in big cities. That is where the new step broke.

Chapter 8 · Turning point

What shipped?

An Android-only checkout redesign went out just before the drop. Sudden, one platform, one step: it usually traces back to something we shipped.

Chapter 9

Connecting the dots

  1. 1The drop is real.
  2. 2It is concentrated on Android.
  3. 3It lives inside checkout.
  4. 4It matches the date of the Android checkout redesign.
  5. 5One step in that redesign fails often in big cities.
Cause found
Chapter 10 · The response

Stop, fix, prove

The order of the response matters as much as the diagnosis.

NEWOLD
First

Stop the loss

Revert the Android checkout or hotfix the location step. Every day of delay costs orders and trust.

Second

Fix the cause

Find out why location confirmation fails in big cities, and repair it.

Third

Re-launch slowly

Give the fixed redesign to a target group. A control group keeps the old checkout. Compare completion before going wider.

Checkout completion on Android Location-step failure rate Orders back toward the earlier level
The end

What I'd carry to the next drop

  1. Don't guess first.Structure removes whole families of causes at once.
  2. Always cut the average.One number can hide the real problem.
  3. A sudden drop is a change.Check releases, campaigns and pricing before the market.
  4. Stop it, fix it, then prove it.Fix before you experiment.
Checklist for any metric drop

0 of 7 checked

Orders are down 12%. Let's not guess.