Marketing Analytics

Final Project Guidelines

MKTG 3528:
Marketing Analytics

Mengzhou (Austin) Zhuang

[email protected]

1
Assign. 3 due

Data selection

Assign. 4 due
Intermediate
report

Presentation

Final report due

2
Parts to cover
• Executive summary (R*: 5%)
• Provide a synopsis / abstract of the entire report

• Problem Identification and Data Summary (P: 30%; R: 20%)

• Background information
• Data and Variable descriptions
• Summary of data
• Research Question

• Analyzing Methods and Results (P: 40%; R: 60%)

• Method Selection
• Data Analysis Description
• Results Interpretation

• Effective communication
• Effectiveness of Presentation (P: 30%)
• Effectiveness of Writing (R: 15%)

*P-presentation, R-report
3
General flow

1. What are the information in your dataset?

2. What are the key information and interesting questions?

3. What could you learn from relevant online information?

4. What are your predictions?

5. How to test your predictions?

6. What are the prediction models?

7. How to interpret your results?

4
A few tips

• Focus on storytelling with evidence

• Logic
• Evidence
• Analysis

• You don’t need to cover everything

• Depth matters more than width
• Support for your predictions should come from
background
• Keep your report consistent

• Precise and concise

• Visualization helps
• Too many redundant elaborations hurt

5
Requirements

• Check out your syllabus

• TLA3 for objective and general requirements
• AT3 for parts you need to include and rubrics

• You are required to root your analysis with the

knowledge, methods, and skills learned from this
class

• DO NOT copy others’ work without appropriate

citation
• Refer to MKTG3528-APA Reference &
Common mistakes, available in Moodle
6
Timeline

• Intermediate report (to give feedbacks) due Apr. 5

• One-page bullet points of your plan
• A direction of your plan (e.g., Variables, questions,
methods.)

• Presentation slides due on Apr. 26

• 20 mins oral presentation
• 10 mins live Q&A
• All group members have to participate
• Upon completion, you will receive a customized badge

• Final project report due May. 12, Mid-night

7
Data Description

8
Data Description

• Dataset will be provided after picked up

– Please do no share datasets
– Last year’s dataset are different – pay attention!

• To pick dataset
– please rank your priority of 6 datasets (1 with the top priority)
– Email me your list
– First-come-first serve (email received time)
– Please make your selection no later than March 28

• No easy or hard dataset

– Follow your interests
– Use your imagination
– Collect additional information if needed
9
Data Description

• Panel
– Variables are aggregated every certain period (e.g., day, week, month,…)

Product_ID Day Sales Unit

1 1 10 1
1 2 5 1
1 3 15 3
1 4 20 2
2 1 20 2
2 2 10 1
2 3 0 0
2 4 5 1

10
Data Description

• Scanner
– Raw transactions
– Multiple observation across time

Product_ID User_id Day price Unit

1 1 1 10 1
2 2 2 5 1
8 3 3 15 1
10 1 2 20 1
2 2 2 20 2
1 3 2 10 1
25 1 1 0 1
27 2 4 5 1

11
Data Description

• Cross-sectional
– Transactions aggregated to some time unit
– One observation across time

Product_ID price Units sold Revenue

1 10 20 200
2 5 19 …
3 15 25 …
4 20 1 …
5 20 32 …
6 10 44 …
7 5 60 …
8 5 52 …

12
Data Description

• Available datasets:
1. Online shopping data
2. Search advertisement data
3. House appliance retail data
4. Mobile app usage during Covid
5. Fake listings in home rental market

6. Social media native ad

7. Social network referral campaign
8. Mobile direct marketing campaign
9. Online-to-offline coupon campaign
10. Smart customer presentative data

13
1. Online shopping data

14
1. Online shopping data

15
1. Online shopping data

• 5,000 consumers from Aug 2013 to June 2015 (23 months)

• ~91,000 transaction records (consumer-seller-product-day)

– PC or Mobile
– Buy and return
– Consumer, Product, Seller types

16
1. Online shopping data

Variable Definition
product_id product indicator
user_id consumer indicator
store_id store indicator
date date of the transaction
time time of the trasaction
price unit sale price (after discount) of the product
original_price original unit price of the product
pay_amount payment amount of the transaction
order_num units sold
mobile_order mobile order (=1) vs. PC order (=0)
gender gender of consumer
age age of consumer
membership_duration the membership duration of consumer
flagship flagship store (=1) vs. regular store (=0)
category_id indicator of general category of the product
openyear store open since
main_brand whether the product is the major brand of the seller: major brand (=1) vs. others (=0)
rating product rating
store_sales cumulative sales of store
hedonic product: hedonic (=1) vs. utilitarian (=0)
certification number of official business certification
category general category of the product
return whether the order is returned

17
2. Search advertisement data

18
2. Search advertisement data

19
2. Search advertisement data

• 679 keywords from Sept to Nov 2018 (60 days)

• ~64,000 Daily observations (product-keyword-day)

– Top 8 positions are shown (first page)
– Consumer response: Click-through rate & Conversion rate
– Industry, keyword, seller, product features
– Advertising strategy?

20
2. Search advertisement data

Variable Definition
keyword_id keyword indicator
seller_id seller indicator
product_id product indicator
date record date
position ad position (1 top, 8 bottom)
search number of searches
click number of clicks
conversion number of purchases
price advertised price
discount_rate discount rate
original_price original price
brand keyword: include brand (=1) vs. others (=0)
promo keyword: include promotion (=1) vs. others (=0)
keyword_length number of words in keywords
review_number number of reviews of the product
rating reviews rating of the product
title_length number of words in ad title
bookmark number of consumers bookmarked the product
previous_units_sold number of units sold previously
keyword_number number of keywords bid by the sellers
product_number number of products advertised by the sellers
list_id list index (a set of search results)
industry the general category (industry) the keyword belongs to

21
3. House appliances retail data

22
3. House appliances retail data

• 167 products from Apr 2016 to Sep 2017

• Panel data: ~90,000 Daily transactional records

– Online channel and offline channel
– Sales and promotion
– Product/brand characteristics
– Online review

• The retailer provides price-match (online and offline) policy from Feb. 1,
2017

23
3. House appliances retail data

Variable Definition
product_id product indicator
date record date
match The price match policy is implemented (=1) vs. not (=0)
online_price unit price online
offline_price unit price offline
unit_cost unit cost
competition number of competing products
category_id product category index
category product category
multinational multinational brand (=1) vs. domestic brand (=0)
listing Public-held (=1) vs. private-held (=0)
product_number number of products within the same brand
units_sold daily units sold
units_sold_online daily units sold online
units_sold_offline daily units sold offline
order_number number of transactions
order_number_online number of transactions online
order_number_offline number of transactions offline
brand_id brand index
rating rating of the product in online store
priorsales the cumulative sales in online store

24
4. Mobile App Usage during Covid

25
4. Mobile App Usage during Covid

26
4. Mobile App Usage during Covid

• 100 apps from Jan 2019 to Jul 2020 in 10 cities

• Panel data: ~80,000 weekly usage summary

– Week active user
– Local covid cases and policy
– City and App features

27
4. Mobile App Usage during Covid

Variable Definition
appid App indicator
cityid city indicator
category category of the app
week week number starts since data recording
date recording date
offline Whether the app involves offline services (=1) or online only (=0)
hedonic Whether the app hedonic (=1) or utilitarian (=0)
free Whether the app is free of charge (=1) or with paid component (=0)
platform Whether the app is a platform app (=1) or not (=0)
activeratio mobile order (=1) vs. PC order(=0)
wau weekly active users
usage proportion of weekly active users within the focal city
marketshare Marketshare of the app within the category
gdp gdp of the city
gdpcapital gdp per capital of the city
population population of the city
lockdown_level The indicator of lockdown level (1 is the weakest and 4 is the tightest)
exist number of existing covid cases
confirmed Total number of confirmed covid cases
cured number of cured
dead number of deaths
start general category of the product
policy whether the order is returned
lock whether the city is being locked down
added number of newly added covid cases
density population density of covid cases in the focal cities

28
5. Fake listing in home rental market

29
5. Fake listing in home rental market

• One of the major online home rental matching platforms

– House rental listings in 1,343 communities
– Across 5 cities
– From 2017 to 2019

• ~48,000 monthly aggregates

– The platform intentionally added some fake listings
– Fictitious (fake) or deceptive (exaggerated)
– How effective?

30
5. Fake listing in home rental market

Variable Definition
area_id Community index
date The reporting date
city/city_id City (index) of the listings
fic Number of fictitious listings in the community
fic_price Average rental price of fictitious listings
fic_size Average rental size of fictitious listings
dec (dec_price, dec_size) Number of deceptive listings in the community
report Number of consumer complains received
distance Distance between the community and subway station
listing (listing_price, listing_size) Number of listings in the community
landlord Number of landlords involved
Profession_landlord Number of professional landlords (agents) involved
new (new_price, new_size) Number of new listings in the community
match (match_price, match_size) Number of matches in the community
call Number of phone enquires received in the community
visit Number of visits received in the community
city_avg_price The average rental price in the city
market_quality The quality score of the community in the city
market_share The share of listings of the community in the city

31
6. Social media native ads

32
6. Social media native ads

• 223 ads of a mobile gaming app distributed in a social media from Feb.
2018 to May. 2018

• ~20,000 daily aggregates

– Impression -> click -> download
– 5 types of ads
– Target customers: subscribers vs. strangers
– The effectiveness of the ad

33
6. Social media native ads

Variable Definition
adid Ad indicator
date Reporting time
week Week index of reporting time
weekday Mon to Sunday
textid The indicator of the text content
imageid The indicator of the image
subscription Target audience, subscriber=1 and stranger=0
ad_type Type of the ad
display Number of ad impressions received
click Number of ad click-throughs received
conversion Number of downloads received
cost Daily cost of ad
narrative Whether the content is narrative (Yes=1, no=0)
words Number of words in the text content
sentence Number of sentences in the text content
positive(negative) Number of words expressing positive (negative) attitude
The gender of the characters in the image (0=female, 1=male, 2=mixed or
gender unknown)
pic Number of images
human Whether the character in the image is human (yes=1 and no=0)
morecharacter Whether 2 or more characters are involved (unknown=2, yes=1, and no=0)
interaction Whether the characters in the images are interacting with readers (yes=1 and no=0)
autoplay Whether the video (if nay) will be autoplayed (yes=2 and no=1)
screenshot Whether the images are screenshot of the gameplaying (yes=2 and no=1)
numeric_word Number of numeric words in the text content
past_word Number of words related to the past in the text content
current_word Number of words related to the current situation in the text content
future_word Number of words related to the future in the text content
34
7. Social network referral campaign

35
7. Social network referral campaign

• 48 social network ads from Jun 2017 to Jan 2018

• ~100,000 distributions
– Ad: repost with reward vs. regular
– Consumer decisions: Repost and read
– Ad poster (who share the ads) and ad receiver (who were exposed to ads)
– Level of distribution (level=1 for seeds; level>1 for C2C distribution)

36
7. Social network referral campaign
Variable Definition
ad_id ad indicator
user_id user indicator
common_friend number of friends shared by ad poster and receiver
repost receivers' decision of reposting (=1) or not (=0)
repost_social receivers' decision of reposting to social media (=1) or not (=0)
repost_friend receivers' decision of reposting to friends (=1) or not (=0)
read receivers' decision of reading the ad (=1) or not (=0)
brand ad: brand-focused (=1) vs. others (=0)
incentive ad: repost for reward (=1) vs. others (=0)
deal ad: deal-focused (=1) vs. others (=0)
fdegree ad poster's number of friends
degree ad receiver's number of friends
read_num number of reads by ad receiver
moment view time: morning (=1), afternoon (=2), evening (=3), midnight (=0)
workday view time: workday (=1) vs. weekend (=0)
dwell_time time spent in reading the ads
title_length number of words in ad title
theme_unrelated Theme of ad: unrelated (=1) vs. others (=0)
theme_fitting Theme of ad: fitting (=1) vs. others (=0)
theme_rendering Theme of ad: rendering (=1) vs. others (=0)
theme_progress Theme of ad: progress report (=1) vs. others (=0)
theme_style Theme of ad: style (=1) vs. others (=0)
theme_technology Theme of ad: technology (=1) vs. others (=0)
level level of posts (1=close to advertisers)
read_date read date (if readed)
read_time read time (if readed)
post_date post date
post_time post time
root The original ad poster
37
rootdegree The original ad poster’s number of friends
8. Mobile direct marketing

38
8. Mobile direct marketing

• Direct messages campaign in WeChat official account in July 2021

– To promote a cosmetic product

• ~87,000 customers received the direct message

– Customized to reflect customer’s name
– Customized to reflect the sales representative’s name
– Customized the product recommendation to the specific customer
– The effectiveness of the message

39
8. Mobile direct marketing
Variable Definition
user_id Customer indicator
region Geographically location of the customer
region_id Location indicator
register_time The date of customer registered as a member
gender Gender of the customer
tenure Number of weeks since the customer become a member
household Whether the customer shop for a family
age Age of the customer
os Operational systems of the customer mostly used devices
pay_total Total expenditure of the customer in the past year
view_total Total page views of the customer in the past year
order_num Total order number of the customer in the past year
priorcategory Total number of unique SKUs the customer bought in the past year
firstorder Whether the customer has bought anything from the retailer (yes=0, no=1)
personal_recommendation Whether the message features personalized recommendation (yes=1, no=0)
personal_name Whether the message features personalized consumer names (yes=1, no=0)
personal_seller Whether the message features personalized representative names (yes=1, no=0)
click Whether the customer click the message (yes=1, no=0)
buy Whether the customer buy the product (yes=1, no=0)
buy_amt The amount of the order (if buy=1)
gdp GDP of the city of the customer
gdpincrease percentage increase in GDP of the city of the customer
citysize Size of the city of the customer
pop Population of the city of the customer
40
9. O2O Coupon campaign

41
9. O2O Coupon campaign

• ~84,000 distributed online-to-offline (O2O) Coupon in April 2020

– ~80,000 consumers across 337 cities national wide
– Distribution (mobile) → claim (mobile) → redemption (offline)
– Coupon: games with friends vs. regular
– 3 types of coupons: combo buy 1 get 1 free, free chicken bucket, free drink

42
9. O2O Coupon campaign

Variable Definition
coupon_id coupon index
user_id user index
gender gender
location_id location index
gdp GDP of the city of the coupon receiver
gdpincrease percentage increase in GDP of the city of the coupon receiver
citysize Size of the city of the coupon receiver
pop Population of the city of the coupon receiver
redeem coupon: redeemed (=1) vs. not redeemed (=0)
claim coupon: claimed (=1) vs. not claimed (=0)
displaydatetime coupon distributed date&time
claimdatetime coupon claim date&time (if claimed)
redeemdatetime coupon redeem date&time (if redeemed)
displaydate coupon distributed date
claimdate coupon claim date (if claimed)
redeemdate coupon redeem date (if redeemed)
prior_campaign_participation consumer: number of previous campaign participated
invitation_send_number consumer: number of invitation sent
invitation_receive_number consumer: number of invitation received
previous_game_participation consumer: number of previous game participated
shared coupon: shared by friends (=1) vs. by advertiser (=0)
game coupon: involves game (=1) vs. others (=0)
coupon coupon type

43
10. Smart customer representative

44
10. Smart customer representative

• ~80,000 consumers who contact the retailer through live chat system
– Three task types: fit, promotion, trending (ask for popular products)
– Two customer representatives: human vs. chatbot
– The representatives will recommend a product at the end
– The consumers could click and buy the products in the live chat system

45
10. Smart customer representative
Variable Definition
user_id Customer indicator
reg_time The date of customer registered as a member
tenure Number of weeks since the consumer registered as a member
ai Whether a chatbot is used (yes=1 and no=0)
Task_type Type of the task
purchase Whether the consumer bought the recommended product (yes=1 and no=0)
purchase_amt The size of the order (if purchase=1)
click Number of times the consumer clicked the recommended product
cart Whether the consumer added the recommended product to shopping cart (yes=1 and no=0)
gender Gender of the consumer
age Age of the customer
os Operational systems of the customer mostly used devices
orderamt Total expenditure of the customer in the past year
ordernum Total order number of the customer in the past year
avgprice Average price of the past purchases
precategory Total number of unique SKUs the customer bought in the past year
pageview Number of pages viewed in the past year
pagestay Number of pages stayed (viewed for more than 5 seconds) in the past year
shareproduct Number of products shared to friends in the past year
commentread Number of reviews viewed in the past year
discountnum Number of coupon used in the past year
discountamt Total amount of promotion used in the past year
pop City population
size City size
gdp City GDP in the past year
gdpcapita GDP per capita of the city
unemployment Unemployment rate of the city
cpi CPI of the city
46
Notes

• Please make your selection no later than Mar. 28.

– First come first serve
– List no less than 5 selections
– Send an email directly to me

• Think about your final project

– Intermediate report by Apr. 18