Bidding has become more difficult. With over 1000 changes in AdWords last year, there is a new level of complexity that cannot be managed by any humanoid. With automation now a necessity, Daniel will run through a number of real-world examples of bidding automation. Once you've finished this session you'll have all the tools you need to bid better than 90% of advertisers.
7. @danielgilbert44
• AdWords restricted to 6 bidding
windows
• 3rd party tools make 4 changes max
• Overall ROI increase of 11% for
Domino’s Pizza
http://bit.ly/1GwKN1Z
24/7 Bidding
10. @danielgilbert44
• AdWords is an auction
• 3rd party tools make 4 changes max
• Overall ROI increase of 67% for
alpharooms
• http://bit.ly/1IcQDAF
RTB to position
11. @danielgilbert44
• Call centre demand changes
• 3rd party tools make 4 changes max
• Overall CR increase of 17% for Chill
Insurance + 32% reduced spend
• http://bit.ly/1IcQDAF
Bid to respondto demand
14. @danielgilbert44
• Released in Marin March 2014
• Accessible via OpenWeatherMap
• Overall Conversion increase of 14% for
alpharooms
• http://bit.ly/1KEKGwc &
http://bit.ly/1PMCVcO
Adjustbids basedon the weather
15. @danielgilbert44
• Still not visible in most 3rd parties
• None ready for custom audiences
Bid for locations & audiences
19. @danielgilbert44 19
Element Keyword Level Optimisation Account Level Optimisation
Bid change decision
Each keyword is bid according
to its own value prediction.
Optimize target metric total for portfolio of
keywords.
Ability to set portfolio level budget and
efficiency constraints.
Total cost and target predicted for each
keyword at different bid levels.
Rapid reaction to seasonal trends in data.
Longtail management Clustering algorithms used to create predictions for tail terms.
Mobile Can be optimised at AdGroup level
Data Requirements
10 conversions in 90 days
1 conversion in last 7 days
10 conversions/day
100 clicks/day
What about for algorithms?
21. 21
PID example
o Make your own lookback windows
oThese sample weights gives us a weighted predicted conversion rate of 13.5%
oTherefore, MaxCPC = Conversion Rate * MaxCPA = £5.40 * 13.5% = 73p
Conversion rate period
7 Days 14 Days 30 Days
[pokemon
meetups]
7% 14% 28%
Weighting 9% 90% 1%
0 5 10 15 20
Relative
weighting
Look back period
Choose your ownlookback
22. 22
o For keywords without significant historical data our bespoke clustering algorithms group relevant
keywords.
o Longtail terms such as [increase my pokemon collection anonymously] present a data problem.
o Use three-dimensional clustering for longtail terms. A simplistic illustration:
Data from campaign
level – or product
category
Data from behavioural
clustering
Data from semantic
clustering
Not enough data
at keyword level
Make yourownclusters
23. 23
Test weightings
o We construct weightings across the different clustering dimensions.
o This table shows suggested weightings for the keyword [dkny dress for my lovie dovie].
Cluster
dress dkny Semantic(in/for)
Conversation 8% 13% 2%
Sampleweighting 15% 80% 5%
o Calculated weighted conversion rate: 17%
o MaxCPC = £7.80 * 17% = £1.36
0 5 10 15 20
Relative
weighting
Look back period
24. 24
The magic – A/B test it
0 5 10 15 20
Relative
weighting
Lookbackperiod
ClusteringA
0 5 10 15 20
Relative
weighting
Lookbackperiod
ClusteringB
0 5 10 15 20
Relative
weighting
Lookbackperiod
PID Loop B
0 5 10 15 20
Relative
weighting
Lookbackperiod
PID Loop A
http://selnd.com/1Chx7PH
25. @danielgilbert44 25
The magic – A/B test it
• I know I get better conversion rates in London but
should I increase bids there?
• Which is the best bid management software for
me?
• Am I using the best campaign structure?
• Is my PPC agency any good?
http://www.brainlabsdigital.com/brainlabs-tech-stack