Transportation Electricity Industry Agriculture Commercial
Cars
2970 kg
Trucks
1210 kg
Coal
2310 kg
Natural
gas
1880 kg
Chemicals
790 kg
Refining
750kg
Cattle
700 kg
Soil
management
960 kg
Others
Metal
Others
1
2
3
4
5
0
Yearly tCO2e per capita - source United States Environmental Protection Agency
Residential
Planes
495 kg
1200 kg 1020 kg
Personal car is the
#1 cause of GHG
emissions worldwide
USA cars: 18%
Product beekly
Build a machine learning component
from scratch
with a product-first approach.
Money and Time 💸
Hard to find the right price,
Time-consuming to update regularly
Smart Pricing 📈
Machine learning component to
recommend optimal prices to
owners
Context: Smart Pricing
13
Scope: EU
1
It’s a product marketing
challenge 📣
2
3
16
Ok, It’s also a technical challenge ⚙
(but…)
It’s a design challenge
Launching a machine
learning component is
a technical challenge.”
“
The Marketplace Dynamics squad
18
Its mission is to maximise long term revenues of
the marketplace
2 Data Scientists
2 Full Stack developers
2 Pricing Managers
1 Product Manager
1 Product Designer
Meet:
Product
Manager
Designer
Engineer Data
24
“I hope the algorithm does its job well enough so that I don’t lose
money”
“I would have tried Smart Pricing just to see prices that the algorithm
suggest.”
“There is a button to deactivate at anytime. Reversibility in one click,
that’s reassuring”
Learnings - It’s all about trust.
They don’t trust the
“machine”
Owners understood how smart
pricing works, but are suspicious on
the fact that the algorithm can do
better than their manual work.
👉 Reinforce trust with
transparency (Preview)
They are afraid to lose
money
Owners need proof they would earn
more with smart pricing.
More bookings ?
More earnings ?
👉 Give more info before
activating Smart Pricing (copy)
Learnings - It’s all about trust.
25
They want to keep some
control
Owners feared the switch would be
a one-way thing.
They wanted a preview of prices
before making their mind.
👉 Leave control on smart pricing
(min price, price edit)
Follow the language used by owners
26
“Area”
“Renters”
“Client”
“Modulate prices” “Booking stream”
“My car part”
“Edit prices”
“How much I get / In my
pocket”
“High tier”
“Old option”
“You set prices”
“Minimum price”
34
Product Marketing
1
1 - Announcement
Announce the new offer when
opening the app
2 - Banner
Promote the feature in-product
(here, pricing page)
3 - Emailing
To all eligible owners, with
repeated campaigns
4 - Blog post
On the owner community blog
2
1
3
It’s a technical challenge.
(But it’s also a people and process challenge)
57
If you are starting with machine learning,
start narrow but vertically integrated.
58
Data org.
Prediction
Model precision
Adjustments
Production releases
Monitoring
Pedagogy
Benchmark
Narrow perimeter
Large perimeter ✔
🚫
Data org.
Prediction
Model precision
Adjustments
Production releases
Monitoring
- limited region
- limited car types
- limited owners typology
- …
Pedagogy
Benchmark
- all regions
- all cars
- all owners
- …
Start narrow, vertically integrated