An expert in PPC optimization discusses how to maximize visibility and minimize waste through granular targeting and automated bidding at scale. Granular ad texts and bidding the correct amount for every keyword can increase ROI and reduce waste. Managing many products, brands, sizes, and categories makes full granular targeting difficult to do manually at scale. Automated systems are needed to flexibly create, update, and manage keywords, ads, bidding, and more across many campaigns daily using product data and structured logic while integrating with ad platforms. Comparison groups and statistical smoothing can help determine accurate conversion rates for low volume keywords.
6. Granularity
Granularity means:
– More specific, better performing ad texts
– Bidding the correct amount for every possible
keyword
– Higher ROI, less waste, better returns
8. Example
A company specialising in Red Armani Fedora Hats could
end up with ad groups such as these:
– Red Armani Fedora Hats
– Armani Fedora Hats
– Red Fedora Hats
– Red Fedoras
– Red Hats
– Fedoras
– Hats
– Headwear
We also might show ad text such as “Up to 25% off Red
Fedora Hats” or “Armani Red Fedoras from £30”, etc.
9. The problem of scale
But what do we do if the company also sells:
– 10 colours?
– 10 brands?
– 5 sizes?
Plus a similar range of top hats?
… And caps too?
Very quickly the numbers get too big to deal with
10. How do we solve it?
More people or less granularity?
Equivalently:
– More operating costs or lower return?
How about a third option…
13. Item Id Brand Product Classification Colour RRP Price
A0001 Armani Headwear > Hats > Fedoras Red £40.00 £30.00
A0002 Nike Headwear > Hats > Fedoras Red £38.00 £10.00
A0003 Haat Headwear > Hats > Fedoras Red £27.00 £27.00
A0004 D&G Headwear > Hats > Fedoras Red £412.00 £299.00
A0005 Prada Headwear > Hats > Fedoras Red £38.00 £30.00
A0006 Kangol Headwear > Hats > Fedoras Red £29.99 £29.99
A0007 Boss Headwear > Hats > Fedoras Red £99.00 £67.00
A0008 Nike Headwear > Hats > Fedoras Red £18.00 £14.00
Process overview
Red Armani Fedora Hats
Armani Fedora Hats
Red Fedora Hats
Red Fedoras
Red Hats
Fedoras
Hats
Headwear
14. Extract fields
from feed
•Convert into usable
text-strings
•Manually build
Synonyms
•Derive categories
from taxonomy or
classification
Plan targeting
item types
•Combinations of
fields
•Multipliers
Derive keyword
& ad logic
•Synonyms
•Multipliers
•Templates
•Thresholds
Write automation
•Parse feed into
targeting items
•Generate campaigns,
keywords, ads, etc.
•Sync with Ad Platform
Process overview
15. Process overview – key points
Product level alone is not normally enough
– Cover higher level categories
It’s not enough to have business-specific
templates
– we need business-specific logic
Setup needs a high level of human input
Ask the right questions!
17. The bidding problem
What is the conversion rate of a
keyword with 1 conversion from 7
clicks?
Only 95% sure that the Conversion Rate is
between 0.36% and 57.87%
18. The bidding problem
Determining the conversion rate on broad high-
volume terms is easy
The problem occurs in low-volume long tailed terms
How can we get the most accurate measure of
conversion rate?
– Look to other similar terms
– External data
19. Similar terms
Group keywords based on user intent:
“Red Armani Fedora”
“Blue Armani Fedora”
“Green Armani Fedora”
“Cheapest Armani Fedora”
“Discount Armani Fedora”
“Armani Fedora”
“Armani Fedora Hat”
20. Comparison groups - Smoothing
Pool statistics from the comparison group
– But don’t ignore the keyword’s own stats:
Keyword 3%
150 clicks
Group 1.5%
10,000 clicks
2%
2.5%
Keyword 3%
300 clicks
Group 1.5%
10,000 clicks
21. Comparison groups - Smoothing
So how do we incorporate external data? We believe our keyword out-
performs the group by 50%
2.5%
Keyword 3%
300 clicks
Group 1.5%
10,000 clicks
2.75%
Keyword 3%
300 clicks
Adjusted Group
Rate 2.25%
22. BidLabTM – How it works
Custom tree structure based on
intention
All keywords will use the most
relevant data
Very tightly-grouped in a natural
fashion
25. Summary
Maximising visibility from PPC
campaigns depends on:
Granularity
Up-to-date, focussed,
relevant ad copy
Full product & category
coverage
Accurate bidding
Minimising waste means knowing how
to automatically:
Add or update hundreds of new
campaigns
Add or update hundreds of
thousands of ad texts & keywords
Accurately manage bids on
millions of low-volume keywords
Every single day!
Stress these ads were all live & showing for the term “purple scarf”
Highlight – matching what the user searched for.
M&S Ad is still less than perfect. Can’t we say something about the purple scarves M&S stock rather than a generic line1 & line 2?
Short v long tail
Stress that all these groups & structure are selling ONE product and possibilities are endless. Size terms, ‘bold hats’, etc.
Highlight that this hypothetical company’s target market is still VERY ‘niche’ but already we could be running to thousands of ad groups
Note that lots of volume will exist in crossovers such as “Red hats”, etc. We need to retain the ability to bid on these with ad texts saying “Choose from over x red hats” or “Up to 25% off Red Hats”
This dilemma is faced by PPC account managers all over the world
MINIMAL human intervention on day to day basis
Flexibilty to change ad messages, add kw multipliers etc. quickly & easily
Define “Higher-Level”
Make case for logic. “Choose from 8 different Red Fedoras” sounds good – “Choose from 8 different tshirts” does not!
Lots of people do it at product level but not higher level categories
Introduce the issue first!
Need to value a click
so need CVR
This is WHY we don’t resell it
Lot of manual work goes into optimising this