Search term reports optimisation can be a tedious process. A PPC Specialist will invest many hours do dive in into each campaigns trying to extract the most and worst performing keywords. In this presentation, we will share how the heavy lifting of the Search Term report can be automated and how technology can help to identify trends invisible to the human eye.
17. EXA
ORDER
EXA
TYPOS
2010
Broader Reach or Less Control?
BMM
2014 2017
EXA
MEANING
2018
BMM &
PHRASE
MEANING
2019
*Check Frederick Vallaey Article on Search Engine Land
18. N. of Triggering Queries per KW
EXA
ORDER
EXA
TYPOS
2010
Broader Reach or Less Control?
BMM
2014 2017
EXA
MEANING
2018
BMM &
PHRASE
MEANING
2019
EaseofControl
19. We live in a Fuzzy World
BEFORE
QUERYKW
CAR VEHICLE
+CAR +VEHICLE
1.0 1.0
20. We live in a Fuzzy World
BEFORE TODAY
QUERYKW
CAR VEHICLE
+CAR +VEHICLE
1.0 1.0
CAR VEHICLE
+CAR +VEHICLE
0.8 0.7
0.2 0.3
24. DI =
Identify queries triggering certain keywords that are not containing the initial
keyword
Dispersion Index
SUM OF IMPRESSIONS
from Queries not containing the initial KW
TOTAL KW IMPRESSIONS
27. Dispersion Index
1000
800+30
a) +CAR Tot Impressions = 830
b) Impressions from Queries not
containing the keyword + CAR = 30
CAR VEHICLE
+CAR +VEHICLE
0.8
0.3
100
28. Dispersion Index
1000
830
a) +CAR Tot Impressions = 830
b) Impressions from Queries not
containing the keyword + CAR = 30
c) DI +CAR
= 30/830 = 3.6%
CAR VEHICLE
+CAR +VEHICLE
0.8
0.3
100
34. Close
Variant
Dispersion Index | The not-so-easy Way
Accuracy
Complexity
Levenshtein
Distance
Jaro-Winkler
Jaccard
DistanceHamming
Distance
Token
Based
35. Close
Variant
Dispersion Index | The not-so-easy Way
Accuracy
Complexity
Levenshtein
Distance
Jaro-Winkler
Jaccard
Distance
Wordnet /
SpaCy
Hamming
Distance
Semantic
Based
36. CI =
Cannibalization Index
COUNT Keywords triggered by the same Query
COUNT TOT KWs
Identify the frequency of a certain query is triggering different Keywords
39. Cannibalization Index
a) COUNT Keywords triggered by
the same CAR = 2
b) COUNT TOT KW = 3
CAR VEHICLE
+CAR +VEHICLE
TRUCK
+TRUCK
40. Cannibalization Index
a) COUNT Keywords triggered by
the same CAR = 2
b) COUNT TOT KW = 3
c) CI CAR
= 2/3= 67%
CAR VEHICLE
+CAR +VEHICLE
TRUCK
+TRUCK
48. 0 PERF.
STR
REPORT
SETTINGS
- Time Range
- ROAS/CPA
- Min. Clicks
- Min. Impressions
- Target
- Conversion Action
MAGIC CANNIBALIS
ATION
BUDGET
BURNERS
MONEY
PRINTER
CTR
KILLERS
MISPLACED
LOW PERF.
TOP PERF.
AD
REPORT
49. CANNIBALISATION Identify and fix Misplaced Queries
A Query can be considered Misplaced when a “Closer” Ad Group for
the same Query does exist in the Account
50. CANNIBALISATION Identify and fix Misplaced Queries
A Query can be considered Misplaced when a “Closer” Ad Group for
the same Query does exist in the Account
Conditions to be matched:
● Query ≠ Triggered KW & Triggered KW Variations
● Query = KW Account Variations (NLTK)
● Triggered Ad Group ≠ Matched Ad Group
● Query ≠ Triggered Ad Group KWs
● Query ≠ Negative Matched Ad Group KW
53. CANNIBALISATION
Conditions to be matched:
● Query = KW Account Variations (NLTK)
● Triggered Ad Group ≠ Matched Ad Group
● Query ≠ Triggered Ad Group KWs
● Query ≠ Negative Matched Ad Group KW
cheap car car BMM_car
BMM_cheap car UK | Search | Generic | Car cheap car
54. A Query can be considered Budget Burner when is underperforming
either in terms of Conversion or CTR
BUDGETBURNERS Identify Underperforming Queries
55. A Query can be considered Budget Burner when is underperforming
either in terms of Conversion or CTR
BUDGETBURNERS Identify Underperforming Queries
CTR KILLERS
Query CTR < AVG Ad Group CTR
LOW PERFORMERS
CPA > n*tCPA or ROAS < ROAS/n
0 PERFORMERS
Conversion = 0
56. A Query can be considered Money Printer when is delivering enough
volume both in terms of Clicks and Conversion above Target
Identify New Performing KeywordsMONEYPRINTER
57. ● KW is generated from the Triggering Query
MONEYPRINTER
58. ● KW is generated from the Triggering Query
● Headline1 as well is generated from the Query
MONEYPRINTER
59. MONEYPRINTER
● KW is generated from the Triggering Query
● Headline1 as well is generated from the Query
● All other elements are copied from the Triggered Ad
65. n-Gram
how to do online advertising
how to advertise online
online advertising
how to advertise
ad online
STR
66. n-Gram
how to do online advertising
how to advertise online
online advertising
how to advertise
ad online
STR
67. how (3)
to (3)
online (4)
advertising (2)
...
n-Gram
how to do online advertising
how to advertise online
online advertising
how to advertise
ad online
STR UNIGRAM
68. how (3)
to (3)
do (1)
online (4)
advertising (2)
...
n-Gram
how to do online advertising
how to advertise online
online advertising
how to advertise
ad online
STR
how to (3)
online advertising (2)
...
UNIGRAM BIGRAM
72. how to do online advertising
how to advertise online
Smarter than n-Gram...Smart n-Gram
how
online
advertise
73. how to do online advertising
how to advertise online
Smarter than n-Gram...Smart n-Gram
how
online
advertise
WORD
ROOTS
74. how to do online advertising
how to advertise online
Smarter than n-Gram...Smart n-Gram
how
online
advertise
WORD
ROOTS
PROs
> Ignore grams order
> Aggregate grams in Words Root
80. “All roads lead to Google Ads”
AdWords API
BQ Service
Transfer
Google Ads API
81. PROs
> Faster GET than GAds API
> Structured Data Format (CSV)
> Tons of Documentations
The Old Fashioned Way AdWords API
82. PROs
> Faster GET than GAds API
> Structured Data Format (CSV)
> Tons of Documentations
CONs
> Clunky and Complicated POST
> Soon Deprecated
The Old Fashioned Way AdWords API
83. PROs
> Exhaustive Data Export
> BigQuery Based (Computational
Power)
BigQuery Data Transfer Service
84. PROs
> Exhaustive Data Export
> BigQuery Based (Computational
Power)
CONs
> Maintenance Cost ($5/M/Account)
> Refresh Window (Max 30 Days)
> High Initial Investment
BigQuery Data Transfer Service
85. PROs
> Structured POST (gRPC)
> Powerful Query Builder
> Easier access to Granular Data
Future Proofed Google Ads API
86. PROs
> Structured POST (gRPC)
> Powerful Query Builder
> Easier access to Granular Data
CONs
> Very little Documentation
> Back to Beta… :-(
> Slow slow GET (2x vs AW API)
Future Proofed Google Ads API
91. 3+1 BigQuery (Unfair) Advantages
• Direct Integration with Google Ads
• Detached from your Server Architecture
92. • Direct Integration with Google Ads
• Detached from your Server Architecture
• Computational Power
3+1 BigQuery (Unfair) Advantages
93. 3+1 BigQuery (Unfair) Advantages
• Direct Integration with Google Ads
• Detached from your Server Architecture
• Computational Power
+ Plug & Play ML Solution
108. 0 PERF.
STR
REPORT
SETTINGS
- Time Range
- ROAS/CPA
- Target
- Min. Clicks
- Min. Impressions
- Conversion Action
MAGIC CANNIBALIS
ATION
BUDGET
BURNERS
MONEY
PRINTER
CTR
KILLERS
MISPLACED
LOW PERF.
SONS OF
BMM
AD
REPORT THE BIG PICTURE
123. Resources & Materials
Wind of Change
● Reach more of the right customers with phrase match and broad match modifiers
● Timeline: The long, slow death of Exact Match in Google Ads
● The impact of ‘close variants’ in exact match types
● SKAG vs The World
● Search query reports show a huge increase in Google close variant traffic
Scaling Up
● Google Ads Query Builder
● gRPC
● Google Ads Transfers
● Google Ads report transformation
● What is Kubernetes