W3 L1 Flashcards
4 main pillars in online marketing
website placement, social media, search engine, mobie webapp
social media: 7 levels of Customer engageemnt
inactives, spectators, joiners, collectors, critics, conversationalists, creaatoes
What Drives Influence (Influence Model):
The probability that a person (I) choose potential influencer A over B will increase with the number other people choose A (social proof)
what are level 2 followers
followers of followers
2 main parameters to determine influencers
1 Number of K (2) (chain) level followers
2 Likelihood (activation) level of each (1+…+K) level follower to share the content
virality rate percentage berekenen
shares/impressions x 100
What Makes an Online Content Go Viral ? 4
Positive content High-Emotional Arousal, Positive emotions of amusement, excitement, inspiration, and warmt, Drama elements such as surprise, plot, and characters, including babies, animals, and celebrities arouse emotions.
Social Media Analytics: Integrating
Customer ID + Social ID
what is customer id
information about one customer, purchase history, adress etc
social id:
social media, activitty, emotion social identity
what delivers measurable marekting ROI
f ACTUAL SHARING OF CONTENT
buyers jounrey 5
awareness - considration - decision - adoption - advocacy
Share of Voice berekenen
of conversations mentioning your brand / total # of industry conversation
Social Sales Effectiveness berekenen
of sales coming from social channels / # total sales
conversation rate
of comments or replies / # of posts
amplification rate
of shares or retweets / # of posts
applause rate
of likes or favourits / # of posts
WHAT makes a Facebook (brand) post more POPULAR? what do we need to answer this question?
Remember we need some X (predictors) and outcome/performance (Y) variables to answer this question and to build our model.
FB page popularity: what are your x and y variables?
1 What do we want to improve/make better ? These are your outcome Y variables
2 What can we change to achieve this objective? These are your predictors X variables
“What kind of social media content drives people’s engagement on social media : In other words: What kind of content should we post ? “
what could be our x and y variable here?
Y = Dependent Variable
Model 1 Y = Number of Likes
Model 2 Y = Number of Comments
Two separate models for (i) likes and (ii) comments
X variables = predictors (excplanatory variables)
Structured Data
: We can add, sum, get averages and make mathematical statistical operations (number of likes, number of reshares etc)
Unstructured Data:
In tekst, photo or video format. Very common. Then how do we deal with it?
Unstructured Data – Definition:
Data unit in which the information offers a concurrent representation of its multifaceted nature without predefined organization or numeric values
1 Non-Numeric:
Text, Photo or Video Data. To be converted into numeric data to be analyzed.
2 Multifaceted
Possesses multiple facets, each offering unique information.
Example: A photo = brightness, mode, contrast, colors or a Voice: pitch, speech rate, intensity
Concurrent Representation
Each facet represents a different phenomena at the same time.
Example: An online customer review has a positivity (valence) but also represents a topic (price, service, service etc)
3 facts about unstrcted data
1 Non-Numeric: 2 Multifaceted: 3 Concurrent Representation
Unstructured Data 1: Tekst Data Analysis 1 Classifiers
group or tag data into a defined category (by sentiment, emotion, topic, etc.)
Unstructured Data 1: Tekst Data Analysis 2 Extractors
: retrieve pieces of information (like keywords, entities, phrases, numbers, etc.)
Sentiment Analysis
s is a sub domains of text-analytics based on the classification of statements being +/- (positive or negative)
Some measures to be used in visual (video and photo) analytics:
Resolution, Aspect Ratio, Hue (warm vs cool colors), Brightness, Saturation, Contrast, Smoothness
clustering images dichotomies example
Example 1: Clustering of social media images based on different dimensions (dichotomies)
How to cluster image 2
(1) Supervised and (2) Unsupervised Learning
Supervised learning
model like ‘Teacher’ or ‘Supervisor’ who tells the machine the Label so that the machine knows which output from the given input.
(Mere) presence of an image has a …… impact on …..
Mere) presence of an image has a positive impact on customer engagement on Twitter
High-quality and professionally shot pictures lead to
s lead to higher engagement on Instagram and Twitter
Presence of human face and image–text fit can lead to
ead to higher user engagement on Twitter but not on Instagram
(1) Feature complexity:
i.e., unstructured pixel-level variation; color, luminance, and edges
(2) Design complexity
i.e., structured design-level variation; number of objects, irregularity of object arrangement, and asymmetry of object arrangement
results Li & Xie: is a piture worth a thousand words
Results*: (read the full paper for detailed results and managerial implications)
Inverted u-shape between feature complexity and consumer liking
Regular u-shape relationship between design complexity and consumer liking
Sentiment Analysis
: the process of computationally identifying and categorizing opinions expressed in a piece of text, in order to determine whether the writer’s attitude towards a particular topic, product, etc. is positive, negative, or neutral.
In sentiment analysis, social conversations are classified based 3
n positive, negative, neutral language
2 types of scores sentinment anlysis textual data
- begative or positive 2. Each word is scored on the basis of its sentiment based on a Lexicon database: every word has a score
Multi layered tecxtual seniment analys 3 different dimensions
1 Valence positivity-negativity
2 Subjectivity (Emotionally) to what extent the tekst is factual or subjective: emotion-opinion based
3 Polarity (Extremity) how strong the positivity in (1) is expressed
Paralanguage
Aspects of speech that are not actual verbal prose,
gives contextual information that allows interactors to more
appropriately understand the message being conveyed
3 types of paralanguage
1 Auditory Non-Verbal Communication:2 Visual Non-Verbal Communication:3 Tactile Non-Verbal Communication:
1 Auditory Non-Verbal Communication:
information communicated by aspects of speech such as pitch,
rhythm, tempo, vocal qualities and vocalizations
2 Visual Non-Verbal Communication:
conscious or unconscious bodily movements that possess communicative
value, including human gestures and body language.
Tactile Non-Verbal Communication:
nonverbal communication related to physical, haptic interaction with
another individua
Topic Models extracts and classifies
Topic Models extracts and classifies (clusters) the prominent TOPICS mentioned in a textual data (reviews, social media posts etc)
Topic and Sentiment analysis: used to
used to manage the pleasure-pain points along CUSTOMER JOURNEYS
Topic and Sentiment analysis:
The rate (%) of positive-negative comments in each stage of customer journey based on each TOPIC
Sentiment & Topic Analysis Used Together
Brand Reputation Tracking Using Social Media
Positive-Negative words (by lexicons) are matched/associated with certain extracted topics (
Facial Expressions are relatively easier to detect through Machine Learning/Deep Learning techniques
Research focuses on six basic categories of emotions
happiness, surprise, anger, sadness, fear, and disgust.
Challenges in Visual Sentiment Analysis: 3
Visual semantic is hidden in images (not expressed with words – but facial expression <easy>, body language or ambience: to be detected)
Visual sentiment = Visual semantic, and there is no high level visual semantic dictionary like text-analysis lexicons
Visual sentiments from images, requires high level visual semantic ontology features instead of low-level visual features</easy>
