Analyzing Brand Reputation with Oracle Analytics

[email protected] www.rittmanmead.com @rittmanmead Analyzing Brand Reputation  with Oracle Analytics Federico
Venturin, Rittman Mead

[email protected] www.rittmanmead.com @rittmanmead 2 •Principal Consultant at Rittman Mead •12
years experience in BI/Analytics •Based in Trebaseleghe Venice, Italy • [email protected] • https://speakerdeck.com/barretbse Federico Venturin

[email protected] www.rittmanmead.com @rittmanmead 3 •Oracle data and analytics services provider
•40+ sta ff helping clients to get the most out of their investment in Oracle's software and services •Signi fi cant web presence with the Rittman Mead Blog:  https://www.rittmanmead.com/blog Rittman Mead

[email protected] www.rittmanmead.com @rittmanmead 4 A Long Time Ago

[email protected] www.rittmanmead.com @rittmanmead 5 •Italy used to qualify for the
FIFA World Cup A Long Time Ago

[email protected] www.rittmanmead.com @rittmanmead 6 •I was a big fan of
Oracle Endeca Information Discovery A Long Time Ago

[email protected] www.rittmanmead.com @rittmanmead 7 Nothing Lasts Forever

[email protected] www.rittmanmead.com @rittmanmead 8 •Italy failed to qualify for the
World Cup twice in a row Nothing Lasts Forever

[email protected] www.rittmanmead.com @rittmanmead •Endeca didn't achieve the success it deserved
and disappeared 9 Nothing Lasts Forever

[email protected] www.rittmanmead.com @rittmanmead 10 Why Am I Here Today?

[email protected] www.rittmanmead.com @rittmanmead 11 •To remind you that Italy is
still the European champion Why Am I Here Today?

[email protected] www.rittmanmead.com @rittmanmead 12 •To remind you that several features
seen in Endeca have been included in Oracle Analytics •I'm going to show you how to use them to perform a brand reputation analysis without ML or marketing experience Why Am I Here Today?

[email protected] www.rittmanmead.com @rittmanmead 13 Brand Reputation is the perception of
the brand and how the particular brand of the company is viewed by customers, stakeholders, and the market as a whole. It is the culmination of ideas and emotions that a customer associated with the brand with the customer service experienced during the purchase of goods and services, whilst using them, and after-sales services provided by the company. — Marketing91 Brand Reputation

[email protected] www.rittmanmead.com @rittmanmead 14 •Identify the brand •Collect the data
•Prepare and enrich the data •Explore the data •Improve the analysis Activities

[email protected] www.rittmanmead.com @rittmanmead Identifying the Brand

[email protected] www.rittmanmead.com @rittmanmead 16 eFootball 2022

[email protected] www.rittmanmead.com @rittmanmead 17 •Free to play football videogame released
on 30 September 2021 •At lauch it was heavily criticized and became the worst-rated game on Metacritic and Steam eFootball 2022

[email protected] www.rittmanmead.com @rittmanmead Collecting the Data

[email protected] www.rittmanmead.com @rittmanmead 19 •Metacritic ‣ Website that aggregates reviews
of fi lms, TV shows, music albums and video games •Twitter ‣ Microblogging and social networking service on which users post and interact with messages known as tweets Where to Collect the Data?

[email protected] www.rittmanmead.com @rittmanmead 20 •All eFootball 2022 user reviews for
each platform What to Collect from Metacritic?

[email protected] www.rittmanmead.com @rittmanmead 21 •Collect tweets mentioning the o ff
i cial eFootball account •Collect tweet replies for the o ff i cial eFootball account What to Collect from Twitter?

[email protected] www.rittmanmead.com @rittmanmead 22 •Oracle Analytics does not allow to
collect data from Metacritic and Twitter out of the box How to Collect the Data?

[email protected] www.rittmanmead.com @rittmanmead 23 •Use Beautiful Soup How to Collect
Reviews from Metacritic?

[email protected] www.rittmanmead.com @rittmanmead 24 Reviews.xlsx

[email protected] www.rittmanmead.com @rittmanmead 25 •platform is the environment in which
the game is executed •name is the unique identi fi er of the user who wrote the review •date is the date when the review has been posted •rating measures how good the game is according to the user •review is the text of the review •total_ups is the number of users who found the review helpful •total_thumbs is the number of users who evaluated the review Reviews.xlsx

[email protected] www.rittmanmead.com @rittmanmead 26 •Apply for a Twitter developer account
•Set up your Twitter API project and an application •Use Search Tweets v2 •Convert the data (JSON) to Excel How to Collect Tweets from Twitter?

[email protected] www.rittmanmead.com @rittmanmead 27 •Create your twitter_keys.yaml credential fi le
Using Search Tweets v2

[email protected] www.rittmanmead.com @rittmanmead 28 •Execute the search_tweets.py application Using Search
Tweets v2

[email protected] www.rittmanmead.com @rittmanmead 29 Tweets.xlsx

[email protected] www.rittmanmead.com @rittmanmead 30 •id is the unique identi fi
er of the tweet •author_id is the unique identi fi er of the user who wrote the tweet •created_at is the creation time of the tweet •text is the actual UTF-8 text of the tweet •retweet_count, reply_count, like_count and quote_count are public engagement metrics for the tweet at the time of request Tweets.xlsx

[email protected] www.rittmanmead.com @rittmanmead 31 Users.xlsx

[email protected] www.rittmanmead.com @rittmanmead 32 •id is the unique identi fi
er of the user •name is the name of user as it appears in the user's pro fi le •username is the Twitter screen name, handle, or alias that the user identi fi es themselves with •location is the location speci fi ed in the user's pro fi le •followers_count, following_count, tweet_count and listed_count are public metrics about activity for the user at the time of request Users.xlsx

[email protected] www.rittmanmead.com @rittmanmead Preparing and Enriching the Data

[email protected] www.rittmanmead.com @rittmanmead 34 •Use a custom script to the
detect the language of each review and fi lter out non- English ones Filtering Out Non-English Reviews

[email protected] www.rittmanmead.com @rittmanmead 35 •Allows you to extend the machine
learning and data curation capabilities of Oracle Analytics •Python/R scripts can be embedded in a simple pre-de fi ned XML format, uploaded into DV and invoked from a data fl ow •Custom script feature must be enabled (Doc ID 2675894.1) Custom Script Feature

[email protected] www.rittmanmead.com @rittmanmead 36 Example: py.DetectLanguage

[email protected] www.rittmanmead.com @rittmanmead 37 •<scriptname> is the name of the
script •<scriptlabel> is the name of the script displayed to end users •<target> is the execution target •<type> must be set to execute_script for the script to show up in data fl ows •<scriptdescription> is the description of the script •<outputs> contains the outputs returned by the script •<options> contains input parameters to the script •<scriptcontent> contains the actual Python/R script Important Tags in the XML Wrapper

[email protected] www.rittmanmead.com @rittmanmead 38 •Add a calculated column to determine
the sentiment (Positive, Negative or Mixed) for each review using the rating information Analyzing the Sentiment (Metacritic)

[email protected] www.rittmanmead.com @rittmanmead 39 •Use the Analyze Sentiment step to
detect the sentiment (Positive, Negative or Neutral) for each tweet Analyzing the Sentiment (Twitter)

[email protected] www.rittmanmead.com @rittmanmead 40 •Use a custom script to break
text down into distinct words, count their occurences and save the results in a new dataset Tokenizing Reviews and Tweets

[email protected] www.rittmanmead.com @rittmanmead 41 • As an alternative, you can
use  the Text Tokenization database  analytics function ‣ The source data must be available  in an Oracle Database ‣ Straightforward implementation ‣ Returns only top N tokens ‣ Does not handle mentions  and hashtags in tweets Tokenizing Reviews and Tweets

[email protected] www.rittmanmead.com @rittmanmead •Save the data to a database connection
and create datasets with multiple tables 42 Creating the Data Model

[email protected] www.rittmanmead.com @rittmanmead Exploring the Data

[email protected] www.rittmanmead.com @rittmanmead 44 •Shows the number of brand mentions
over the selected period •Tip: Investigate peaks and drops to identify what caused such an increase or decrease in mentions Mentions Over Time

[email protected] www.rittmanmead.com @rittmanmead 45 Mentions Over Time (Metacritic)

[email protected] www.rittmanmead.com @rittmanmead 46 Mentions Over Time (Twitter)

[email protected] www.rittmanmead.com @rittmanmead 47 Mentions Over Time (Twitter)

[email protected] www.rittmanmead.com @rittmanmead 48 •Shows the general attitude towards the
brand •Provides context for mentions and helps to understand the data Sentiment Analysis

[email protected] www.rittmanmead.com @rittmanmead 49 Sentiment Ratio (Metacritic)

[email protected] www.rittmanmead.com @rittmanmead 50 Sentiment Ratio (Twitter)

[email protected] www.rittmanmead.com @rittmanmead 51 Sentiment Over Time (Metacritic)

[email protected] www.rittmanmead.com @rittmanmead 52 Sentiment Over Time (Twitter)

[email protected] www.rittmanmead.com @rittmanmead 53 •People with the ability to in
fl uence potential buyers of a product or service by promoting or recommending the brand In fl uencers

[email protected] www.rittmanmead.com @rittmanmead 54 Top 10 In fl uencers by
Tweet Count

Followers Count

Engagement

[email protected] www.rittmanmead.com @rittmanmead 57 •Shows the concepts people usually associate
with the brand •Source of inspiration for creating content, titles, and campaigns Word Cloud

[email protected] www.rittmanmead.com @rittmanmead 58 Word Cloud (Metacritic)

[email protected] www.rittmanmead.com @rittmanmead 59 Word Cloud (Twitter)

[email protected] www.rittmanmead.com @rittmanmead 60 Mentioned Users Cloud (Twitter)

[email protected] www.rittmanmead.com @rittmanmead 61 Hashtag Cloud (Twitter)

[email protected] www.rittmanmead.com @rittmanmead 62 •Leverage data actions to interact with
people and data sources ‣ Access a user pro fi le on Metacritic:  https://metacritic.com/user/${valuesForColumn:username} ‣ Access a user pro fi le on Twitter:  https://twitter.com/intent/user?user_id=${valuesForColumn:author_id} ‣ Access a tweet on Twitter:  https://twitter.com/twitter/status/${valuesForColumn:id} Data Actions

[email protected] www.rittmanmead.com @rittmanmead Demonstration

[email protected] www.rittmanmead.com @rittmanmead 64 •Automate the data gathering process •Update
the stop list with the commonly used words in the context of the brand (e.g. efootball, game, konami) •Extend the search query on Twitter by including popular hashtags (e.g. #efootball, #efootball22, #efootball2022) •Identify major competitors (e.g. EA SPORTS FIFA) and research their mentions to evaluate strengths and weaknesses •Monitor all relevant data sources What's Next?

[email protected] www.rittmanmead.com @rittmanmead 65 #EOF email   [email protected] web  
https://www.rittmanmead.com/blog twitter   @barretbse presentations   https://speakerdeck.com/barretbse

Analyzing Brand Reputation with Oracle Analytics

Analyzing Brand Reputation with Oracle Analytics

More Decks by Federico Venturin

Other Decks in Technology

Featured

Transcript