I have made a slide for a marketing analytics course that I have recently completed on Coursera. The slide covers topics such as: different types of marketing analytics and their uses, brand value & brand architecture, customer lifetime value and its applications, before-after and full factorial marketing experiments and regression analysis.

1. Marketing Analytics
Prepared by
Mahir Mahtab Haque

2. 3 types of marketing analytics:
• Descriptive analytics – It involves looking at the past. It’s basically ad
hoc/standard reports, where you’re looking at the data and seeing what
had happened and how long ago. It may even be alerts, which says
something abnormal is happening and what actions are needed to tackle
this abnormality. In a nutshell, descriptive analytics is looking at the
history and seeing what happened in the past.
• Predictive analytics – It deals with what will happen, say, if you reduce the
price of a product. Randomized testing also falls under predictive
analytics, which looks at, how can we look at changing the price or
increasing promotion advertising. And what will happen through
experiments and AB testing.
• Prescriptive analytics – Optimization falls under prescriptive analytics,
which looks at, what’s the best that can happen of all the options out
there.

3. Marketing Process
• Objectives:
- Customers
- Company
- Competitors
- Collaborators
- Context
• Strategy:
- Segmentation
- Targeting
- Positioning
• Tactics:
- Product
- Price
- Place
- Promotion
• Financials:
- Gross margin
- ROI
- Customer life time value

4. Airbnb marketing process
• Objective: How to improve customer experience?
- Customers: Everyone who is looking to travel is a customer of Airbnb
- Company: Airbnb portal
- Competitors: Hotels
- Collaborators: Airbnb allows people to put up their houses for rent on the portal
- Context: Sharing economy, it allows customers to transact with each other and share something they have, to be rented by others using
the portal
• Strategy:
- Segmentation: Could be based on location, adventure, price, vacation, family, students, etc
- Targeting: Deals with which segment to focus on
- Positioning: Value proposition
• Tactics:
- Product: A house/apartment
- Price: Rent rate
- Place: Location of the house/apartment
- Promotion: Customer reviews describing their experience
• Financials: It’s about how Airbnb makes money
- As a portal that makes hosts and guests come together, Airbnb charges both sides of this platform
- Airbnb’s strategic challenge: How do we improve the rental prospects for our hosts and identify better rental options for our guests?
- Airbnb needs to see if they can use all the user-generated data (reviews coming from the promotional part) and the price
- By using this information of price and reviews, Airbnb can impact their bookings and achieve their goal of renting more properties

5. Airbnb marketing process
• A Mental model is the first step used in the analytics
process to outline the factors that influence the target
metric. It is used as a hypothesis to test the data.
• Start by identifying the target metric Airbnb is trying to
maximize. This is the ‘Profit per property’ = gross margin X
price charged X number of rentals X minimum stay
• Factors influencing customers on which property to rent:
- Star rating
- Review
- Product attributes

6. Airbnb marketing process
• In order to put text into a mathematical/predictive model,
we must process text data (Reviews data) into a number.
This is commonly known as text analytics. Software like R
use text data and convert it to a review sentiment. Higher
the review sentiments score, more positive the review.
• We need to take all the reviews for each of the property
and run it through a code in R to get the review sentiments
score. After this, the data collected can be plugged into a
predictive model to predict the number of times it is saved
on a wish list or rented.
• What Airbnb needs is a region-based strategy

7. Airbnb marketing process
• Descriptive analytics is really about looking at historic information
that Airbnb has on its website that looks into reports that says how
often a property is rented, in which place, and why and what are
happening in terms of some abnormalities in terms of rental rates
over time. This is basically summarizing past information.
• In predictive models, we looked at how we take all the review
sentiments, star ratings and property attributes to predict sales of a
property.
• In prescriptive, we are going to look at now that in Miami price is
important, how can we optimize the price in Miami to improve the
rental prospects of properties in Miami?

8. Snapple brand value
• Snapple’s marketing mix:
- Product: 52 different flavors in glass bottles, pop-top
- Price: No discounting
- Place: Mom and pop/cold channel
- Promotion: Real people/ Wendy
• By merging Snapple with Gatorade, Quaker destroyed $ 1.4 billion of value. This was a
clear example of how not understanding what a brand means to the consumer, and
treating all brands equally, and merging them without thought, can really destroy value.
• Later, Triarc buys Snapple from Quaker and identifies 3 wrong decisions that Quaker
made and reversed them:
- Revamp the Snapple ad campaigns - Back to quirky advertising, rehired Wendy
- Launch new products and product lines – Back to quirky products
- Make friends with distributors again
Gatorade Snapple
• 8 flavors
• Warehouse
distribution
• Associated with
Sports
• Science
• Endorsed by
Michael Jordan
• 52 flavors
• Distributed door to
door
• Associated with
Dixie peach
• Mango madness
• Endorsed by Howard
Stern

9. Developing brand personality
• What is a brand:
- The capitalized value of trust between the company and the consumer
- A consumption tax for customers who want what a brand promises to
supply
- A relationship with customers
- The extra money a customer is willing to pay to get what the brand stands
for
• Brand personalities:
- Sincerity – down-to-earth, honest, wholesome, cheerful
- Excitement – daring, spirited
- Competence – reliable, intelligent, successful
- Sophistication – upper-class, charming
- Ruggedness – outdoorsy, tough

10. Brand architecture
• Brand personality is part of a larger view of brands called
brand architecture.
• Marketing uses this brand architecture to connect product
features and attributes to the emotional connection of
brands with their consumers.
• Brand architecture identifies aspects of consumer response
that need to be influenced by marketing actions.
• Marketers use analytics to evaluate how marketing affects
aspects of the brand architecture.

11. Southwest airlines brand architecture
Freedom
Friendly, fun, down-to-
earth
Less hassle, pleasant
Less time, reliable, convenient
Point-to-point, low price, downtown
Brand core/essence
Brand personality
Emotional benefits
Product benefits
Product attributes

12. Measuring brand value
• Interbrand valuation model
Financial
analysis
Market
analysis
Brand
analysis
Residual
earnings &
forecasts
Role of
branding
Brand
strength
score
Brand value
Brand
earnings
Risk rate

13. Measuring brand value
• Interbrand carries out both market analysis and brand analysis
through consumer surveys. The financial analysis comes from a
company’s balance sheet. They evaluate the residual earnings and
forecast, and then merge it with market analysis to see how
important a brand is in the market.
• Next, in brand analysis, they look at the strength of the brand to
see how strong is the relationship of the brand with its consumers.
This is done through consumer surveys and the risk rate is
identified, i.e., how strong the brand is going to be in the future.
• Finally, the brand earnings and the risk rate are merged to form the
brand value.

14. Measuring brand value
• Young and Rubicam brand asset valuator: It is totally based on consumer
surveys and it has four different elements: how the brand is differentiated,
is it relevant, what is the esteem and knowledge? All of these concepts
packaged into what is called the brand asset.
Differentiation Relevance
Brand stature
(emotional
capital)
Brand asset
Brand strength
(vitality)
KnowledgeEsteem

15. Measuring brand value
Aspiring
brands (high
on d,; low on
r, e & k)
Power
brands (high
on d, r, e & k)
New/fading
brands
(some d: low
on r, e & k)
Eroding
brands (high
on k: low on
d, r & e)
Brandstrength
(differentiation&relevance)
Brand stature (esteem & knowledge)

16. Measuring brand value
• A Y&R brand asset valuator gives you diagnostics on
what you need to do with the brand. It gives you the
relationship of the brand with the consumer.
• But it doesn’t give you the dollar value/financial value
of the brand, which the Interbrand rankings do.
• Interbrand gives you the financial value, and Y&R gives
you what the value means in terms of relationship with
the customers and gives good diagnostics. As both of
them together complete the picture in terms of how to
value a brand.

17. Measuring brand value
• Brand equity:
- Provides a long term estimate of the value of a brand
- Measures, such as Interbrand and revenue, provide a
financial estimate
- The Y & R provides a diagnostic estimate
• What is the value of knowing Brand Value:
- Know the amount to pay for a company during mergers
and acquisitions
- Trade off marketing investment between long term brand
value and short term price promotion pressures

18. Revenue premium as a measure of
brand equity
• It combines the measures of Interbrand brand ranking and Y & R brand asset valuator, to give
you both the financial value of the brand along with some diagnostics.
• It is more suitable for brands that are in the grocery stores, like Colgate or Snapple.
• It is based on data that is collected when you scan the product(s) in the grocery store during
the check-out. Companies like Information Resources Inc. collect this sort of data when you
scan the product in the check-out counter. They know who you are. What brand you
bought. What price you paid for the product. And what were the features of the brand and
were there any marketing or display of feature surrounding the brand in the grocery
store. This measure uses all that information to come up with a measure of brand equity.
• Assumptions of this method:
- Brands make decisions that are best for themselves to maximize the brands’ profits.
- This method assumes that branded and private label products are similar in all aspects
except their brand names.

19. Revenue premium as a measure of
brand equity
• Equity = [Revenue premium – Add. Variable cost] X (1+D)/(1+D-R)
• Here,
- Revenue premium is how much more revenue the branded product makes over the
private label
- Additional variable cost is the cost the branded product has to get over the private label
- (1+D)/(1+D-R) is the long-term multiplier, where D is the discount rate; whenever we
make projections into the future, you want to include the discount rate when you are
taking the future money to the present day's terms. R is the stability factor, which
measures the riskiness of the brand.
• If the sales of the brand is stable, you're okay with getting the cash down the road
versus today. If the brand is very risky, you want this money from the brand today.
• So if R goes up, that is if R is stable, 1+D-R goes down. If that happens, equity increases.
So equity is higher for stable brands and vice versa.
• Note: Annual brand equity = Revenue premium – Add. Variable cost

20. Calculating Snapple brand equity

21. Customer lifetime value
• Customer lifetime value (CLV) can be defined as the discounted sum of all
the future customer revenue streams minus product and servicing costs
and remarketing costs.
- It computes the dollar value of an individual customer relationship
- It is both backward looking and forward looking, i.e., computing value of
past customers and using that information to project forward.
• CLV is used to:
- determine how much to spend to acquire a customer
- determine how aggressively to spend to retain a customer/group of
customers
- even value a company
• With its knowledge of CLV, Netflix is able to spend money where it matters
(technology, retention) most to enhance customer value.

22. Customer lifetime value for Netflix
a) Expected customer lifetime in months – 20 months
b) Average gross margin per month per customer - $50
c) Average marketing costs per month per customer
(assuming) - $0
d) Average net margin per month per customer = b - c = $50
e) Customer lifetime value = a*d = $1000
• Netflix should not spend more than $1,000 in marketing
to acquire a new customer because this is the break even
point between the cost of acquiring a new customer and
the customer's lifetime value.

23. Calculating CLV
• Remember: CLV is the net present value of future cash flows from any
customer.

24. Calculating CLV

25. CLV – Time Horizon
Percent of CLV accruing in first 5 years
Discount
rate
Retention rate
40% 50% 60% 70% 80% 90%
2% 99 97 93 85 70 47
4% 99 97 94 86 73 51
6% 99 98 94 87 76 56
8% 99 98 95 89 78 60
..
..
20% 100 99 97 93 87 76
Percentage of CLV accruing in the first 5 years states that if most of the CLV that a
customer provided comes within the first 5 years, then it may make sense to do just
those initial 5 years.

26. CLV – Time Horizon trends
• As retention rate goes up, the percent of CLV
accruing in the first 5 years decreases. Increased
retention rate implies that the customers are
more likely to stay with the company for a
longer time. Hence, the percent of overall CLV
accrued decreases for the first five years.
• As discount rate goes up, the percent of CLV
accruing in the first 5 years increases.

27. Two types of services & Initial margin
Customer pays before using the service Customer pays after using the service
Apartment rentals, Netflix, Hulu Credit cards
CLV = [M-R]*(1+d/1+d-r) CLV = [M-R]*(r/1+d-r)
The company that collects CLV after the
service is used is always one margin
behind a company that receives payment
before the service is used.

28. CLV – cohort and incubate
• Cohort = customers acquired at the same time period (month, quarter or year)
• Since retention changes with time since acquisition, CLV calculations are better if they are done separately
for each cohort.
• Typical customer retention curve
Retentionrate
Time

29. CLV – contractual vs. non contractual
• Xfinity and Netflix have a contract with their customers. So these firms
sign a contract with their customers and the customers have to call up
these firms to cancel their subscription. What this means is that the firms,
like Xfinity and Netflix, know when a customer unsubscribes to their
service. This really helps in knowing lifetime duration and retention rate.
• In case of grocery stores, you don't have to sign a contract with them. You
can just happily walk in, buy something, and go home. And then they will
know you're still a customer if you walk back in again. But if somebody
doesn't come back for a long time, that could even mean the customer is
just dormant. It doesn't mean the customer left the relationship, they may
come back again. So what does it mean for calculating lifetime duration
and retention rate? What this means is you'll have to use empirical
models. You'll have to use regression to calculate this retention rate, to
use historic data, to predict expected retention rates. So the retention rate
calculation is much more complex when you use a noncontractual setting.

30. Using CLV to make decisions
• CLV is a complex and sophisticated tool, not just a simple calculation. It can yield
surprising insights into how to allocate spending to boost revenue. IBM identified
customers with a high CLV that had not been called before 2004 and spent
marketing dollars on them, getting the resources by pulling marketing dollars from
customers with a low CLV.
• Where should the firm be looking? What are the metrics that lets them
identify customers who are going to be profitable in the future?
• Backward looking metrics:
- Share of wallet
- Past customer value
- Past period revenue
• Forward looking metrics:
- Customer lifetime duration
- Customer lifetime value

31. Brand equity and CLV
Marketing actions
Customer mindset:
awareness and
associations
Customer behavior:
acquisition and
retention
Brand equity CLV

32. Brand equity and CLV
Marketing
actions:
advertising,
innovation,
promotions,
market
presence, price
Brand equity:
differentiation,
relevance,
esteem,
knowledge
Behavior:
acquisition,
retention, profit
contribution
CLV

33. Experiments: What establishes
causality?
• Change in marketing mix produces change in sales, i.e., increasing
advertising will lead to more sales.
• No sales increase when there’s no change in marketing mix, i.e.,
no increase in advertising will leave sales unchanged.
• Time sequence, i.e., an increase in advertising will lead to an
increase in sales tomorrow.
• No external factor, i.e., when advertising was increased, there was
no change in the market, none of the competitors left the market
nor did they reduce their prices.

34. Experiments: Designing basic
experiments
Choose 1000
customers
Control group
(500)
Exposed to old ad
for a month
Test group (500)
Exposed to ad
highlighting new
packaging for a
month
Control group sales
(1000 units)
Test group sales
(1200 units)
Sales lift (test
control 200
units)
Randomization can
match test and control
groups on all
dimensions
simultaneously, given
a sufficient sample
size.

35. Experiments: Before-after design
Choose 1000 customers
Old ad
Same old ad
Old ad
Sales 1000 units
Sales 1200 units
Exposed to new ad
Sales 1100 units
Sales 1000 units
Sales lift test control
[(1200-1000)-(1100-1000)]
= 100 units
Test group Control group
In this more sophisticated version of a basic experiment, BOTH the test and control groups are first
exposed to the existing marketing to see how that impacts sales within each group and uncover
any pre-existing conditions. Then ONLY the test group sees the new marketing.

36. Web design – Full Factorial Design
Ad copy Price
$1.59 $1.89 $2.15
Lasts longer $1315 $1112 $1206
Tastes better $957 $1030 $1500
Good for
you
$930 $820 $770
As web experiments are cheap and fast, it also provides the additional benefit of
manipulating a lot of variables at the same time. This type of multiple variables being
manipulated simultaneously is called the Full Factorial Design.

37. Web design – Full Factorial Design
• Now Cheerios is priced, at the moment, at $1.89. And they want to see
how sales changes if the price either decreases to $1.59 or increases to
$2.15. At the same time they also want to test the ad copy. They have also
added two other ad copies that they want to test. So, Cheerios is going to
see what happens if they change both price and ad copy simultaneously.
• Findings:
- If the price is kept the same but the ad copy is changed, then the best ad
copy seems to be at $1.89 with sales of $1112
- If the ad copy is kept the same but the price is changed, then the best
decision would be to reduce the price to $1.59 with sales of $930
- However, if both price and ad copy are changed, then the best decision
would be to increase the price to $2.15 which would lead to sales of
$1500

38. Analyzing an experiment: Etch-A-
Sketch
Etch a sketch Doodle Doug
Test product Control product
No. of
weeks
Cincinnati
units
Control
units
Cincinnati
Shares %
Cincinnati
units
Control
units
Cincinnati
Shares %
Pre-test 5
Dec 2005 –
26 Nov 2006
12 162 1526 9.6 1517 6742 18.4
Test 26 Nov
2006 – 16
Dec 2006
3 240 1598 13.1 816 3780 17.7
Lift 136.1 96.7
Net lift 39.4 %

39. Analyzing an experiment: Etch-A-
sketch results
a)Retail price $10
b) Retail margin 36%
c) Mfg. Selling price $6.4 (a x (1-0.36)
d) Mfg.
Contribution
margin %
58%
e) Mfg.
Contribution
margin
$3.712 (c x d)
f) National budget $5000000
g) Break-even unit 1346983 (f/e)
h) Base unit 3100000
i) Base unit test period 1085000
j) Break-even lift % of
Base
124% (g/i)*100
The break even lift is the amount of lift necessary to make sense of the national
investment in TV advertisement.
Here, it is 124%. The net lift from TV ads was 39.4%. So clearly, it does not make any
economic significance to invest in a national TV advertising campaign for Etch A
Sketch because the net lift is much lesser than the break even lift possible, a break
even lift that is necessary to make this any economic sense.

40. Analyzing an experiment: Betty
Spaghetty
Arizona California
Color crazy Go Go Glam Color
crazy
Go
Go
Glam
Total/store/wee
k 17 Jun – 17 Jul
2007
1.8 2.2 0.3 1.2
Lift 267%
(1.8+2.2/0.3+
1.2)*100

41. Analyzing an experiment: Betty
Spaghetty results
Ad budget $3000000
Retail selling price $15
Retail margin % 36%
Mfg. Suggested price $9.6
Mfg. contribution margin % 58%
Mfg. contribution margin $5.568
Break-even units 538793
Note: For projecting lift, we have to go from Arizona Test  Arizona Chain Sales 
National Chain Sales  All retail  Selling season

42. Analyzing an experiment: Betty
Spaghetty results
a)Test % of California sales 10%
b) Total California units 1420
c) California % of national sales 12%
d) National retailer sales 11833 (b/c)
e) Retailer share 25% (d/f)
f) National units 47332 (d/e)
g) Test % of annual sales 5.5%
h) Annual sales 860606 (f/g)
i) Holiday % of annual 45%
j) Holiday units without ads (Base sale
units)
387273
k) Lift from ads 267%
l) Units from ads (expected) 1034018 (j*k)
m) Break-even lift % of Base (expected) 52%

43. Takeaways from marketing
experiments
• Experiments assess the cause and effect
• Pay attention to:
- Design
- Gap between test results and field implementation
- Difference between test and campaign contexts
• Web experiments are cheaper and faster
- Costs of experiments are variable rather than fixed
• Experiments provide forecasts of expected ROI
- This can help with determining campaign budgets

44. Regression analysis: Diagnosing market response
Regression statistics
Multiple R 0.775
R-squared 0.601
Adjusted R-
squared
0.586
Standard
error
2.566
Observations 29
df SS MS F Sig F
Regression 1 267.28 267.28 40.60 0.00
Residual 27 177.75 6.58
Total 28 445.03
Coefficients Standard
error
t Stat P-value
Intercept 9.90 0.85 11.60 0.00
No. of
promotions
1.42 0.22 6.37 0.00
ANOVA Table
P-value can be considered as ‘the confidence in the regression;’ thus, a
p-value of 0.00 can mean that the regression is very unlikely to
change, i.e., low p-value=higher confidence
In marketing, 60% is a good R-squared because marketing has a lot of factors that influence
our consumers to go and buy products in the store.

45. Regression analysis: Multivariable regression & omitted variable

46. Regression analysis: Multivariable
regression & omitted variable

47. Regression analysis: Multivariable
regression & omitted variable

48. Regression analysis: Multivariable
regression & omitted variable
As marketers, what is the big issue if
a regression model has an omitted
variable bias?
Since the model does not include the
omitted variable, the impact of the
omitted variable is dispersed among
the rest of the variables. Such a
model might lose out on identifying
the real driver influencing the
dependent variable.

49. Using price elasticity to evaluate
marketing
• PED = [% change in sales/ % change in price] X [price/sales]
• Note: The coefficient from a log-log model is equal to the elasticity, which
shows how price effects sales.

50. Understanding Log-Log models
• Logarithmic transformation are particularly useful when the rate of
change of a variable (sales) is relative to other variables (price).
• Log-Log models look at:
- Percent change
- Regression of the first log
- Price elasticity
• First difference of natural LOG = percentage change
- Logging converts absolute differences into relative (i.e., percentage)
differences
- The series DIFF (LOG(Y)) represents the percentage change in Y
from period to period

51. Marketing mix models
• Variables to include in a marketing mix model:
- Product = product quality + brand lifecycle
- Price of the product
- Place (distribution)
- Promotion (marketing campaign)
- Carryover effect (this takes into account the impact of an ad
campaign after the campaign has ended)
• Statistical vs. economic significance:
- Statistical significance is the relationship observed in the sample
which is likely to be observed in the population as well. Look for p-
value<0.1 for the coefficient of interest.
- Economic significance checks to see if the benefit from a marketing
intervention (i.e., the size of the coefficient) justifies the expense.

52. Marketing mix models
• Calculating economic significance:
- A unit increase in no. of promotions increases
units purchased by 1.42 (coefficient from
previous slide)
- Assume gross profit per unit is $5
- Cost of promotion is $0.5
- Profit = (units purchased*gross profit) – (cost of
promotion* no. of promotions)
- Profit = (1.42*5) – (0.5*1) = $6.6

53. The End!