Analyse your SEO Data with R and Kibana

Analyse your SEO Data with R and Kibana
June 10th, 2016
Vincent Terrasi

Vincent Terrasi
--
SEO Director - Groupe M6Web
CuisineAZ, PasseportSanté, MeteoCity, …
--
Join the OVH adventure in July 2016
Blog : data-seo.com

Agenda
Mission : Do a Real-Time Log Analysis Tool
1. Using Screaming Frog to crawl a website
2. Using R for SEO Analysis
3. Using PaasLogs to centralize logs
4. Using Kibana to build fancy dashboards
5. Test !
3
“The world is full of obvious things which nobody by any chance ever observes.”
Sherlock Holmes Quote

Real-Time Log Analysis Tool 4
• Screaming Frog
• Google Analytics
• R
Crawler
• IIS Logs
• Apache Logs
• Nginx Logs
Logs

Screaming Frog : Export Data 6
When the crawl is
finished, click the
export button and save
the XLSX file
Add your url and click
the start button

Screaming Frog : Data ! 7
"Address"
"Content"
"Status Code"
"Status"
"Title 1"
"Title 1 Length"
"Title 1 Pixel Width"
"Title 2"
"Title 2 Length"
"Title 2 Pixel Width"
"Meta Description 1"
"Meta Description 1 Length“
"Meta Description 1 Pixel Width"
"Meta Keyword 1"
"Meta Keywords 1 Length"
"H1-1"
"H1-1 length"
"H2-1"
"H2-1 length"
"H2-2"
"H2-2 length"
"Meta Robots 1“
"Meta Refresh 1"
"Canonical Link Element 1"
"Size"
"Word Count"
"Level"
"Inlinks"
"Outlinks"
"External Outlinks"
"Hash"
"Response Time"
"Last Modified"
"Redirect URI“
"GA Sessions"
"GA % New Sessions"
"GA New Users"
"GA Bounce Rate"
"GA Page Views Per Sesssion"
"GA Avg Session Duration"
"GA Page Value"
"GA Goal Conversion Rate All"
"GA Goal Completions All"
"GA Goal Value All"
"Clicks"
"Impressions"
"CTR"
"Position"
"H1-2"
"H1-2 length"

Why R ?
Scriptable
Big Community
Mac / PC / Unix
Open Source
 7500 packages
9
Documentation

WheRe ? How ?
 https://www.cran.r-project.org/
10
Rgui RStudio

Using R : Step 1
Export All Urls
11
"request“;"section“;"active“;
"speed“;"compliant“;"depth“;"inlinks"
Packages :
Stringr
Ggplot
Dplyr
Readxl

R Examples
 Crawl via Screaming Frog
 Classify URLs by :
 Section
 Load Time
 Number of Inlinks
 Detect Active Pages
 Min 1 visit per month
 Detect Compliant Pages
 Canonical Not Equal
 Meta No-index
 Bad HTTP Status Code
 Detect Duplicate Meta
12

R : read files 13
# Read xlsx file
urls <- read_excel("internal_html_blog.xlsx",
sheet = 1,
col_names = TRUE,
skip=1)
# Read csv file
urls <- read.csv2("internal_html_blog.csv", sep=";", header = TRUE)

Detect Active Pages 14
#default
urls_select$Active <- FALSE
urls_select$Active[ which(urls_select$`GA Sessions` > 0) ] <- TRUE
#factor
urls_select$Active <- as.factor(urls_select$Active)

Classify URLs by Section 15
schemas <- read.csv(“conf.csv”,header = FALSE, col.names = "schema", stringsAsFactors = FALSE)
urls_select$Cat <- "no match"
for (j in 1:length(schemas))
{
urls_select$Cat[ which(stri_detect_fixed(urls_select$Address , schemas[j]) ) ] <- schemas[j]
}
/agenda/sorties-cinema/
/agenda/parutions/
/agenda/evenements/
/agenda/programme-tv/
/encyclopedie/
Conf.csv

Classify URLs By Load Time 16
urls_select$Speed <- NA
urls_select$Speed[ which(urls_select$`Response Time` < 0.501 ) ] <- "Fast“
urls_select$Speed [ which(urls_select$`Response Time` >= 0.501
& urls_select$`Response Time` < 1.001) ] <- "Medium“
urls_select$Speed[ which(urls_select$`Response Time` >= 1.001
& urls_select$`Response Time` < 2.001) ] <- "Slow“
urls_select$Speed[ which(urls_select$`Response Time` >= 2.001) ] <- "Slowest"
urls_select$Speed <- as.factor(urls_select$Speed)

Classify URLs By Number of Inlinks 17
urls_select$`Group Inlinks` <- "URLs with No Follow Inlinks"
urls_select$`Group Inlinks`[ which(urls_select$Ìnlinks` < 1 ) ] <- "URLs with No Follow Inlinks"
urls_select$`Group Inlinks`[ which(urls_select$Ìnlinks` == 1 ) ] <- "URLs with 1 Follow Inlink“
urls_select$`Group Inlinks`[ which(urls_select$Ìnlinks` > 1
& urls_select$Ìnlinks` < 6) ] <- "URLs with 2 to 5 Follow Inlinks“
urls_select$`Group Inlinks`[ which(urls_select$Ìnlinks` >= 6
& urls_select$Ìnlinks` < 11 ) ] <- "URLs with 5 to 10 Follow Inlinks“
urls_select$`Group Inlinks`[ which(urls_select$Ìnlinks` >= 11) ] <- "URLs with more than 10 Follow Inlinks"
urls_select$`Group Inlinks` <- as.factor(urls_select$`Group Inlinks`)

Detect Compliant Pages 18
# Compliant Pages
# Canonical Not Equal
# Meta No-index
# Bad HTTP Status Code
# Not Equal
urls_select$Compliant <- TRUE
urls_select$Compliant[ which(urls_select$`Status Code` != 200
| urls_select$`Canonical Link Element 1` != urls_select$Address
| urls_select$Status != "OK"
| grepl("noindex",urls_select$`Meta Robots 1`)
) ] <- FALSE
urls_select$Compliant <- as.factor(urls_select$Compliant)

Detect Duplicata Meta 19
urls_select$`Status Title` <- 'Unique'
urls_select$`Status Title`[ which(urls_select$`Title 1 Length` == 0) ] <- "No Set"
urls_select$`Status Description` <- 'Unique'
urls_select$`Status Description`[ which(urls_select$`Meta Description 1 Length` == 0) ] <- "No Set"
urls_select$`Status H1` <- 'Unique'
urls_select$`Status H1`[ which(urls_select$`H1-1 Length` == 0) ] <- "No Set"
urls_select$`Status Title`[ which(duplicated(urls_select$`Title 1`)) ] <- 'Duplicate'
urls_select$`Status Description`[ which(duplicated(urls_select$`Meta Description 1`)) ] <- 'Duplicate'
urls_select$`Status H1`[ which(duplicated(urls_select$`H1-1`)) ] <- 'Duplicate'
urls_select$`Status Title` <- as.factor(urls_select$`Status Title`)
urls_select$`Status Description` <- as.factor(urls_select$`Status Description`)
urls_select$`Status H1` <- as.factor(urls_select$`Status H1`)

Generate CSV 20
urls_light <- select(urls_select,Address,Cat,Active,Speed,Compliant,Level,Inlinks) %>%
mutate(Address=gsub(“http://moniste.fr","",Address))
colnames(urls_light) <- c("request","section","active","speed","compliant","depth","inlinks")
write.csv2(“file.csv”, filename, row.names = FALSE)
Package dplyr : select and mutate
Edit colnames
Use write.csv2

R : ggplot2 command 21
DATA
Create the ggplot object and populate it with data (always a data frame)
ggplot( mydata, aes( x=section,y=count, fill=active ))
LAYERS
Add layer(s)
+ geom_point()
FACET
Used for conditionning on variable(s)
+ facet_grid(~rescode)

R Chart : Active Pages 23
urls_level_active <- group_by(urls_select,Level,Active) %>%
summarise(count = n()) %>%
filter(Level<12)
Geometry Aesthetic
p <- ggplot(urls_level_active, aes(x=Level, y=count, fill=Active) ) +
geom_bar(stat = "identity", position = "stack") +
scale_fill_manual(values=c("#e5e500", "#4DBD33")) +
labs(x = "Depth", y ="Crawled URLs")
#display
print(p)
# save in file
ggsave(file=“chart.png")

R Chart : GA Sessions 24
urls_cat_gasessions <- aggregate( urls_select$`GA Sessions`,
by=list(Cat=urls_select$Cat, urls_select$Compliant), FUN=sum, na.rm=TRUE)
colnames(urls_cat_gasessions) <- c("Category","Compliant","GA Sessions")
p <- ggplot(urls_cat_gasessions, aes(x=Category, y=`GA Sessions`,
fill=Compliant))+
theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
labs(x = "Section", y ="Sessions") +
scale_fill_manual(values=c("#e5e500","#4DBD33"))
#display
print(p)
# save in file
ggsave(file=“chart.png")

R Chart : Compliant 25
urls_cat_compliant_statuscode <- group_by(urls_select,Cat,
Compliant,`Status Code`) %>%
summarise(count = n()) %>%
filter(grepl(200,`Status Code`) | grepl(301,`Status Code`))
p <- ggplot(urls_cat_compliant_statuscode, aes(x=Cat, y=count,
fill= Compliant ) ) +
theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
facet_grid(`Status Code` ~ .) +
labs(x = "Section", y ="Crawled URLs") +
scale_fill_manual(values=c("#e5e500","#4DBD33"))

R : SEO Cheat Sheet 26
Package Dplyr
select() allows you to rapidly zoom in on a useful subset using operations that usually only work on numeric variable positions
mutate() a data frame by adding new or replacing existing columns
filter() allows you to select a subset of rows in a data frame.
Package Gplot2
aes - geom
ggsave()
Package Readxl
read_excel()
read.csv2()
write.csv2()

Architecture 28
Hard to monitor and optimize host server performance

PaasLogs 32
164 noeuds au sein du cluster Elastic Search
180 machines connectées
Entre 100 000 et 300 000 logs traités par seconde
12 milliards de logs transitent tous les jours
211 milliards de documents enregistrés
8 clicks and 3 copy/paste to use it !

PaasLogs : Streams 36
The Streams are the recipient of your logs. When you send a log with the
right stream token, it arrives automatically to your stream in a awesome
software named Graylog.

PaasLogs : Dashboards 37
The Dashboard is the global view of your logs, A Dashboard is an efficient
way to exploit your logs and to view global information like metrics and
trends about your data without being overwhelmed by the logs details.

PaasLogs : Aliases 38
The Aliases will allow you to access directly your data from your Kibana or
using an Elasticsearch query
DON’T FORGET TO ENABLE KIBANA INDICES AND WRITE YOUR USER PASSWORD

PaasLogs : Inputs 39
The Inputs will allow you to ask OVH to host your own dedicated collector
like Logstash or Flowgger.

PaasLogs : Network Configuration 40

PaasLogs : Plugins Logstash 41
OVHCOMMONAPACHELOG %{IPORHOST:clientip} %{USER:ident} %{USER:auth} [%{HTTPDATE:timestamp}]
"(?:%{WORD:verb} %{NOTSPACE:request}(?: HTTP/%{NUMBER:httpversion_num:float})?|%{DATA:rawrequest})"
%{NUMBER:response_int:int} (?:%{NUMBER:bytes_int:int}|-)
OVHCOMBINEDAPACHELOG %{OVHCOMMONAPACHELOG} "%{NOTSPACE:referrer}" %{QS:agent}

PaasLogs : Config Logstash 42
if [type] == "apache" {
grok {
match => [ "message", "%{OVHCOMBINEDAPACHELOG}"]
patterns_dir => "/opt/logstash/patterns"
}
}
if [type] == "csv_infos" {
csv {
columns => ["request", "section","active", "speed",
"compliant","depth","inlinks"]
separator => ";"
}
}

How to send Logs to PaasLogs ? 43

Filebeat : Install 45
Install filebeat
curl -L -O https://download.elastic.co/beats/filebeat/filebeat_1.2.1_amd64.deb
sudo dpkg -i filebeat_1.2.1_amd64.deb
https://www.elastic.co/guide/en/beats/filebeat/current/filebeat-installation.html

Filebeat : Edit filebeat.yml 46
filebeat:
prospectors:
-
paths:
- /home/ubuntu/lib/apache2/log/access.log
input_type: log
fields_under_root: true
document_type: apache
-
paths:
- /home/ubuntu/workspace/csv/crawled-urls-filebeat-*.csv
input_type: csv
fields_under_root: true
document_type: csv_infos
output:
logstash:
hosts: ["c002-5717e1b5d2ee5e00095cea38.in.laas.runabove.com:5044"]
worker: 1
tls:
certificate_authorities: ["/home/ubuntu/workspace/certificat/key.crt"]

Filebeat : Start 47
Copy / Paste Key.crt
-----BEGIN CERTIFICATE-----
MIIDozCCAougAwIBAgIJALxR4fTZlzQMMA0GCSqGSIb3DQEBCwUAMGgxCzAJBgNVBAYTAkZSMQ
8wDQYDVQQIDAZGcmFuY2UxDjAMBgNVBAcMBVBhcmlzMQwwCgYDVQQKDANPVkgxCzAJBgNVB
AYTAkZSMR0wGwYDVQQDDBRpbi5sYWFzLnJ1bmFib3ZlLmNvbTAeFw0xNjAzMTAxNTEzMDNaFw0
xNzAzMTAxNTEzMDNaMGgxCzAJBgNVBAYTAkZSMQ8wDQYDVQQIDAZGcmFuY2UxDjAMBgNVBA
cMBVBhcmlzMQwwCgYDVQQKDANPVkgx
-----END CERTIFICATE-----
Start Filebeat
sudo /etc/init.d/filebeat start
sudo /etc/init.d/filebeat stop

How to combine multiple sources ? 48

Paaslogs : Plugins ES 49
Description : Copies fields from previous log events in Elasticsearch to current events
if [type] == "apache" {
elasticsearch {
hosts => "laas.runabove.com"
index => "logsDataSEO" # alias
ssl => true
query => ‘ type:csv_infos AND request: "%{[request]}" ‘
fields => [["speed","speed"],["compliant","compliant"],
["section","section"],["active","active"],
["depth","depth"],["inlinks","inlinks"]]
}
}
# TIP : fields => [[src,dest],[src,dest]]

Kibana : Install 51
Download Kibana 4.1
• Download and unzip Kibana 4
• Extract your archive
• Open config/kibana.yml in an editor
• Set the elasticsearch.url to point at your Elasticsearch instance
• Run ./bin/kibana (or binkibana.bat on Windows)
• Point your browser athttp://yourhost.com:5601

Kibana : Edit Kibana.yml 52
Update Kibana.xml
server.port: 8080
server.host: "0.0.0.0"
elasticsearch.url: "https://laas.runabove.com:9200"
elasticsearch.preserveHost: true
kibana.index: "ra-logs-33078"
kibana.defaultAppId: "discover"
elasticsearch.username: "ra-logs-33078"
elasticsearch.password: "rHftest6APlolNcc6"

Kibana : Line Chart 53
Number of active crawled from google over a period of time

Kibana : Vertical Bar Chart 54

How to compare two periods ? 56

Final Architecture
PassLogs Kibana
Filebeat
@
58
@
Soft RealTime
--
Old Logs
IIS
Apache
Ngnix
HA Proxy

Test yourself 59
Use Screaming Frog Spider Tool
www.screamingfrog.co.uk
Teach R
www.datacamp.com
www.data-seo.com
www.moise-le-geek.fr/push-your-hands-in-the-r-introduction/
Test PassLogs
www.runabove.com
Install Kibana
www.elastic.co/downloads/kibana

TODO List 60
- Create a GitHub Repository with all source code
- Add Plugin Logstash to do a reverse DNS lookup
- Schedule A Crawl By Command Line
- Upload Screaming Frog File to web server

Thank you
Keep in touch
June 10th, 2016
@vincentterrasi Vincent Terrasi

Analyse your SEO Data with R and Kibana

More Related Content

Viewers also liked (20)

Similar to Analyse your SEO Data with R and Kibana (20)

More from Vincent Terrasi (13)

Recently uploaded (20)

Analyse your SEO Data with R and Kibana