intelligent bidding using non-sem sources by pravin thampi
TRANSCRIPT
#SMX #23B2 @PravinThampiIncrease revenue and reduce cost faster
Intelligent Bidding Using
Non-SEM Sources
#SMX #23B2 @PravinThampi
100s of thousands of such deals
Over a 1MM Ads Active/day in US
Bid computation is complicated by additional modifiers– RLSA– Device– Location
SEM @ Groupon: Handle Deals that have different locations, prices and time (expiry)
#SMX #23B2 @PravinThampi
100s of thousands of Slot Machines
Unknown probability of success
Target for rate of return on spend
Limited time window for slot machines
Existing Solutions– Head Vs Tail Models– Statistical Models– Machine Learning Models
Bidding Challenge: Identify the deals that generate max revenue before the deal expires
#SMX #23B2 @PravinThampi
1 2 3 4 5 6 7 8 9 1011121314151617181920212223242526272829300
5
10
15
20
25
30
35
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
Non SEM SEM
Sample Data
Use performance from free sources
Strong correlation with paid sources
Free sources Data (Direct/SEO) : leading indicator of performance for deals
Outside the box: Leverage all sources of info
#SMX #23B2 @PravinThampi
Increase/Decrease bids for Deals based on their overall performance
Factor Calculation: 1+(VPC_deal - VPC_avg) / VPC_avg
Less than 1 for below avg performers, greater than 1 for above This factor was multiplied on top of existing bid
calculation Scaling multipliers & upper/lower thresholds can be
used to adjust sensitivity
Simple Model: Best model, used a linear model
*VPC: Value per click/visitor
#SMX #23B2 @PravinThampi
Best deals started rising to the top faster Higher bids saw more conversions Saved cost on low potential deals Maintained ROAS Targets
Double Digit lift in performance!!
#SMX #23B2 @PravinThampiLEARN MORE: UPCOMING @SMX EVENTS
THANK YOU! SEE YOU AT THE NEXT #SMX