Customer Analytics – UPenn Coursera Notes


Descriptive Analytics (keep managerial goal in mind)
* links the market to the firm through information
* information needed for actionable decisions
* systemically collecting and interpreting data that can aid decision makers
* Exploratory research (ambiguous problem) – why sales drop?
* develop initial insight, first step in understanding a broader managerial problem
* focus groups (8-10 individuals, free flow conversation)
* Market Research Online Community (MROC), e.g. C Space, 6 months to 1 year
* caveat: ROI is hard to determine, may have no findings
* Descriptive research (Aware of problem) – who are buying us?
* Casual research (problem clearly defined) – will buyers purchase our product with website change?
Active Data Collection
* Surveys (Qualtrics, SurveyMonkey, Mixpanel – mobile survey – track by location/context, capture customer reaction in-situ, caveat – fatigue)
* Type of questions
* Predictive validity (can help valid) and test-retest reliability (result won’t change volatile)
* Cons: questions biased, get right respondents, require use?
* Self-Reports of customer behaviour
Type of Questions
* Itemised-Category (very satisfied…very dissatisfied), compare to what?
* Comparative (both comparatives might not be that great)
* Ranking (too many comparisons)
* Paired Comparison (e.g. Honda v Toyota) – might hate both, other brands?
* The Likert Scale (very common – agree/disagree statement)
* Continuous scale
* (R squared – between 0 – 1, proportion of the variance in the dependent variable that is predictable from the independent variable)
Net Promoter Score
* track health of brand – customer satisfaction -> profitiability by one single Q
* Not linear relationship – increase NPS may not have measurable change in profitability
* Self-report (Info Scout)
* word-of-mouth dynamics – capture comments in dairy form (Keller Fay)
* Typical response rate is 5%

Survey Design
* Exploratory – open-ended Qs and use this to pre-code close-ended quantitative surveys
* Order bias – use randomised order. Use proven questions, pretest questionnaire
Passive Data Collection: Scanner data – POS data (SPINS, IRi, Nielson)
Media Planning – audience engagement of TV/ Radio show | Social Media Analytics (Hoodsuite/Sprout Social/Web data (Comscore) / Mobile Data (Foursquare)
* Audience engagement for a campaign
* Brand mentions as compared to competitors
* Sentiment analysis (how?positive?)
* Location-based coupons / information to show
Casual Data Collection
* Field experiment (A/B Testing) e.g. click through rate of Landing Page A and B
* Correlation / causation (producing an effect)
* Correlation
* Temporal antecedence (X must occur before Y)
* No third factor driving both (Control other possible factors)

Week 3 – Predictive Analytics
* Prediction in fixed period based on past data
Regression (short term)
* R2 usually 70%-80%
* Can do multiple regressions on multiple variables
* recency, frequency, monetary value (RFM)
* Regression is limited to periods beyond in the future
Probability Model (long future model)
* Buy till you die (forecast customer lifetime value CLV)
* Recency trumps frequency
* Heat map of RF distribution
* Comparison of “cohort” – similar acquisition characteristics (e.g. time)

Prescriptive Analytics
– A problem will have a goal to optimize, actions to be taken and a model linking the actions and the goal.
– Provide recommendation on what actions to take to achieve some objectives / goals.
– Goal: maximise quantity sold? Maximize profit?
– Action: change the price
– MR (marginal revenue) = MC (marginal cos) —> optimal price to maximise profit
– WTP = willingness to pay – how much one would pay for an additional item
– Online targeting – show ads to people after they visited a specific website
– Advertising Attribution problem – see same ads on multiple websites
– Make profit at one customer at a time
* Data
* Data exploration
* Predictive Models (churn models – when leave)
* Optimization
* Firm Decisions (Amazon Prime)
– Collection, management, analysis and strategic leverage of an organisation’s granular data about the behaviour of its customers
– individual-level, behavioural, forward-looking, multi-platform (data fusion), broadly applicable, multidisplinary
– store level—> DM —> Store scanner —> Internet (last-click attribution)
GRP (Gross Rating Point) = Reach*Frequency
– total engagement / profitability – not inter-platform cannibalisation
– 1/3 of discount marketing value is for social network / word of mouth
– FB has great short-term effect / TV has longer carry-over effect
– New set of data:
* Shopping plan (intention survey)
* 60% purchase is unplanned, 40% planned but not purchase
* Shopping path (RFID) / heat map — shelve space (cover 25% only – middle paths are cold)
* Travel & order deviation (not efficient path) – wandering around
* Higher order deviation tend to have more purchases
* Field of vision (Eye-cam)
* 5 foot 6 is optimal shelf level on left hand side (woman) / 5 foot 9 for men
* Recall/consideration/choice higher with eye fixation
* Purchase (scanner data)
* Data-mining to create content by mega-tagging (Netflix)
* Avoid future illnesses – forward-looking (Health care)
* Turn infrequent customers to loyal ones – ROI is highest (Starbucks)
* Call Centre / Amazon
* CLV = customer lifetime value
* Target customers with highest marketing effectiveness
* Lowest churn propensity rate is the most valuable

One thought on “Customer Analytics – UPenn Coursera Notes”

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: