marketing analytics · director data science and analytics fintech startup former uber technologies...
TRANSCRIPT
M A R K E T I N GA N A L Y T I C S
M A R I O A V I N A S C O
D I R E C T O R D A T A S C I E N C E A N D A N A L Y T I C S
F I N T E C H S T A R T U P
F O R M E R U B E R T E C H N O L O G I E S
F O R M E R F A C E B O O K
. . . . . .
T O P I C S
Brief Intro to AI and ML
Models used in Marketing
Use Cases
Q&A
T H E M O S T P O P U L A R U S E O F A I I S I M A G E R E C O G N I T I O N
M O D E L S A T U B E R : " C H U R N "
0%
100%
True Positive
80% Threshold
METRIC
Accuracy how often is the classifier correct?
(TP+TN)/total
Precision When it predicts yes, how often is it correct?
TP/predicted yes
Recall When it's actually yes, how often does it predict yes?
TP/actual yes
O T H E R B R E A K T H R O U G H S I N A I
- Embedding: vector representations of user behavior, words, sounds
- Automated feature engineering: Transformation of the input data using math
- Reinforced learning: multiple neural nets providing feedback to each other
- Reverse bootstrap: data set contain many labels, with noise. This technique
partitions the data, uses 1st partition to score second partition and remove
ambiguous labels
C A S E S. . . . . . +
Communication Frequency
Channel Efficiency
Q&A
C A S E S. . . . . . +
Communication Frequency
Channel Efficiency
Q&A
H O W M A N Y E M A I L S S H O U L D W E S E N D T O O U R R I D E R S / D R I V E R S ?
What is the optimal level so we avoid un-subscription and
yet communicate well?
For a long time, we had a ‘One size’ fits all
We capped the number of marketing emails to our Riders, Drivers and Eaters at a fix number per week
Clearly, some segments expected less communications and unsubscribed accordingly, while others appreciated higher numbers.
Its not an easy problem to identify these segments
U S E S E V E R A L M L M O D E L S A S S E G M E N T A T I O N T O O L S
. . . . . .
By crossing predictive scores we
identify cohorts of high and low
risk for comms.
W E F O R M U L A T E D A N E W C O M M U N I C A T I O N S T R A T E G Y
. . . . . .
We recommended a reduction for the red
cohort and increase for the green one.We estimated that approximately 500K un-subscriptions were
avoided by implementing this model
M O D E L P E R F O R M A N C E – C R O S S S E L L
. . . . . .
The model identifies a segment
with +15X the conversion rate
Avg Conversion
High Conversion 15X
C A S E S
Communication Frequency
Channel Efficiency
Q&A
H O W D O E S M A R K E T I N G I N V E S T M E N T M O V E M Y K P I ?
Where and how much should I invest?
Optimization is the process of allocating resources (usually money) under constraints in order to maximize (or minimize) an outcome
For us, it usually means marketing spend to maximize driver’s trips or acquisition
THIS IS TYPICALLY A PROBLEM OF LOW DATA VOLUMES AND LOTS OF PARAMETERS:
Coefficients
Trend terms
Ad Stock
W H A T D O E S A M E D I A M I X M O D E L D O ?
. . . . . .
It answer: How does
marketing investment move
my KPI?
FORECAST/BENCHMARK
ATTRIBUTION PERFORMANCE
BUDGET OPTIMIZATION
C I T Y M O D E L S A R E C R E A T E D U S I N G T I M E S E R I E S F I T T I N G
SPEND VARIABLES RESPONSE
T I M E S E R I E S F I T T I N G W I T H P Y D L M
THE MODEL IS COMPRISED OF:
Trend
Seasonality
Regressors
P R O B L E M F O R M U L A T I O N
Excel prototype
M A R K E T I N G O P T I M I Z A T I O N R E S U L T S
Optimization is now automated and it reduces 2 weeks of planning and iterations to minutes
In Q3 – 2018 we were able to reduce marketing spend by ~15% in the USA and achieve our targets in terms of First Trips
Ad Stock
Coefficients
Trend terms
We developed a rapid iteration and prototyping process, and we can onboard new models in weeks or less as opposed to months
T H A N K Y O U !
M A R I O A V I N A S C O
M A R I O _ V I N A S C O @ A L U M N I . S T A N F O R D . E D U