Moderator: Matt Van Wagner
Q&A Moderator: Joseph Kerschbaum
Speakers: Alex Cohen, Adam Goldberg, Frank Kochenash, Wister Walcott
PPC & Analytics today! The goal today is to provide a different way to think about the way you’re currently looking at analytics.
Wistor Wolcatt, the tallest guy in search, is up first.
Paid Search Forecasting
Keywords Per Ad Group
Forecasting for paid search involves “predicting clicks, conversions, and profit for different CPA/margin targets, bidding to an overall spending target (budget-based bidding), finding optimal CPA to balance conversion volume and cost, maximizing profit, targeting specific positions, and maximizing revenue for a CPA target.” These guys modeled the auction landscape at the keyword level.
Certain days of the week performed differently than others.
Fit each keyword to a logistic regression:
If you’re at the top of an auction you won’t get more by bidding more.
If you’re at the bottom of the auction you won’t get any traffic at all.
If you’re in the middle your bids will make or break the traffic.
Keywords per ad group:
Test for CTR and QS differences by number of keywords per group.
Compare CTR and QS to the group size.
Look for R2.
High volume keywords should probably get a dedicated ad group regardless.
There isn’t a correlation between quality score and number of keywords in a group and CTR.
Look at the ad groups that have an above average keyword count, below average CTR / QS, start with high click volume groups first.
Questions to ask:
- Any broad match terms getting more traffic than their exact and phrase siblings?
- Any groups with just one active ad?
- Identify loser ads based on CTR or conversions?
- Any budget constrained campaigns?
- Adjusting bids for day of week / time of day? (requires time of click)
- Delete keywords with no impressions for 13 months?
I really wanted to get this guy’s section, but he talked too fast for me to type.
What is attribution? Attribution is giving proper credit to all the advertising-driven consumer activity that results in a desired outcome. Attribution allows for continuous increase of sales and media profitability.
Out-of-the-box attribution models:
The even attribution model takes the conversion and distributes it evenly across all of the ad models that were involved in the profit.
In the even and exclusion model we are excluding credit for a certain type of ad.
With the even and exclusion model we have the ability to:
- Set maximum days in a path
- Aggregate ad network impressions
- Set hours between views and apply exclusions by ad type, ad position, and by time factors.
PPC% orders by path:
PPC accounted for 60% of all single click paths.
When PPC was in a path it received 40% of the credit while 60% of the credit went to other formats.
When PPC was an Introducer and Influencer it received 27% of the revenue credit. This credit would not have been given under a last click model.
“The perfect is the enemy of the good” – Voltaire
Advanced ROI attribution is about predicting probability of conversions.
The prediction of probability of conversion and how it’s used weighs the calculation of ROI.
View the customer journey as an evolution of conversion probability.
The keyword-product sales attribution and optimization problem:
In some cases, the keyword-product correlation is straightforward. In other cases (where variations or substitutes exist) the correlation is more complex.
Questions To Ask:
- How correlated are the keywords to unique SKUs?
- Does the correlation increase with the keyword specificity?
- Is the correlation between keyword and product brand more meaningful?