This is the presentation from the lecture I gave at MDI Gurgaon on November 8, 2017. While examples are not mentioned in it, this deck covers various important concepts that will be useful to anyone trying to acquire a better knowledge of digital marketing and the analytics around it. I particularity focused on Performance Marketing. Please comment / ask questions if any.
Using Analytics and Customers’ Behavioral Data in Digital Marketing
1. Using Analytics and
Customers’ Behavioral Data
in Digital Marketing
Presented by Tarun Babbar
Guest Lecture | MDI Gurgaon | 8th November, 2017
2. We will try to answer the following questions today...
• Digital performance marketing: Why is it so analytics driven?
• First goal: How do we maximize performance for every rupee
spent?
• Second goal: How do ensure everything is incremental?
• Difficulties: What challenges we need to address to remain on top
of the game?
3. Digital Performance Marketing runs on Analytcis
first mindset
BROAD CLASSIFICATION
• Brand Marketing
Engagement
Story telling
• Performance Marketing
Action oriented
Intent driven
PERFORMANCE GOALS
• Maximize return on every dollar
Optimize for various metrics at
different stages of a customer’s
journey in a multi screen format
• Ensure incrementality
Identify what is the real worth of a
specific marketing channel or tactic
5. It’s like selecting the Super Hero you believe will save the earth, every time!!!
There is only one problem - you don’t know yet!
6. 1) While budgeting, look beyond averages
Spends Orders CPO (cost per order) Marginal Cost
Channel 1 3,00,000 13,000 23 50
Channel 2 2,00,000 8,000 25
5,00,000 21,000 24
Spends Orders CPO (cost per order) Marginal Cost
Channel 1 2,50,000 12,000 21
Channel 2 2,50,000 9,500 26 33
5,00,000 21,500 23
Spends Orders CPO (cost per order)
Channel 1 2,00,000 10,000 20
Channel 2 2,00,000 8,000 25
4,00,000 18,000 22
Spends Orders CPO (cost per order) Marginal Cost
Channel 1 2,50,000 12,000 21 25
Channel 2 2,00,000 8,000 25
4,50,000 20,000 23
1. Identify opportunities in
interim metrics but
optimize for final cost of
achieving your goal
2. Optimize for marginal
cost and average cost
across portfolio
3. Diminishing returns
usually apply in Digital
Marketing
7. Segmentation in digital marketing is all about
connecting myriad dots
User Identification
Users Tracking (Cookies,
Pixels, Tags)
Analytical Layer
Delivery Layer (Ad Tech, in
house)
9. 2) Segmentationshouldbe all about puttingthe
customerback to the buyingjourney
• Minimize path to purchase (e.g. triggers for abandoned cart)
Analyze around engagement metrics (time, # of steps, value)
• Capture repeat behavior (e.g. replenishment)
Average frequency analysis
• Present more choices (e.g. fashion categories)
Number of items browsed before buying (# of views, types, category switching)
10. 3) Customersbehavedifferently in different stages
Upper Funnel
Middle
Funnel
Lower
Funnel
1. Prospecting: Where do I go
traveling?
• Exploratory content
• Beach vs. mountains
• Engagement
2. Consideration: Things to do
in Goa?
• Specific, relatable
stories
3. Action: Best offers for Goa?
• Lead capture forms
• Best price guarantee
• Traffic and
keywords analysis
• Time required to
convert
• Bounce rate
analysis
• Type and # of
pages reviewed
11. 4) Funnel analysis is also useful to identifybrokenlinks
and opportunities – specially for end conversion
Advertisement
Landing Page
Site / App / App Install Prompt
Registration / Login
Homepage landing
Product Browsing / search
Add to cart rate
Cart to conversion rate
Shipped order %
Cancellation
/Returns
12. 5) Understandthat different channels may suggest
different customerbehaviours
• Google vs. Facebook (e.g. category, AOV,
stage of buying)
• World of affiliates (e.g. stage of buying
funnel, coupons)
• Some channels drive specific objectives
(e.g. App install vs. transactions)
Typical Analysis
1. Time to convert
2. AOV
3. Category specific conversion
4. Coupons vs. no coupons
5. Price point for first time
customers
6. Customer Lifetime Value (CLTV)
13. 6) Use customerincentivesonly when it helps in
increasingCLTV
• Incentives target customers
• Amount of incentives
• Redemption % with spill over effect
• Delayed buying behavior
Typical Analysis
1. Controlled experiments driven
by hypothesis
2. Margin/Cost analysis
3. Overall lift analysis
4. Time series analysis
14. 7) If youwant to prevent fraud, start with identifying
variations
• E.g. Low click to order time vs. Large click to order time (click injection)
16. Source: http://www.nytimes.com/1999/04/25/weekinreview/ideas-trends-sham-surgery-returns-as-a-research-tool.html
Birth of a
genius!
Circa 1939, long before high-tech
drugs came along to treat the
chest pain known as angina, an
Italian surgeon named Fieschi
devised a simple technique.
Reasoning that increased blood
flow to the heart would ease his
patients' pain, he made tiny
incisions in their chests and tied
knots in two arteries. The results
were spectacular. Three
quarters of all patients
improved.'' One third were
cured.
17. Dr. Leonard A. Cobb
Two decades later, the National Institutes of
Health paid a young cardiologist in Seattle, Dr.
Leonard A. Cobb, to conduct a novel test of the
Fieschi technique. Dr. Cobb operated on 17
patients. Eight had their arteries tied; the other
nine got incisions, nothing more. In 1959, the
New England Journal of Medicine published his
findings: the phony operations worked just
as well as the real thing.
Source: http://www.nytimes.com/1999/04/25/weekinreview/ideas-trends-sham-surgery-returns-as-a-research-tool.html
This is how Sham Surgeries were born!
18. Controlled studies remain gold standard
• Test vs. Control (gold standard)
• Version A vs. Version B
• Existing Baseline vs. New Approach (also called Pre vs. Post)
• Geo Experiments
• Long term control groups
19.
20. It’s getting more complicated…
• Storing, organizing, analysing big data accurately
• Tracking multi screen behaviour
• Blurring of lines between customer behavior (online/offline, segment types)
• Dealing with walled gardens like Facebook
• Uncertainty while dealing with auction markets
• Sophisticated frauds in online marketing
• Data exchange with partners
• Privacy issues
21. Thank you for coming.
You can find me at LinkedIn https://www.linkedin.com/in/tbabbar/
Editor's Notes
There are various ways to prioritize. Run experiments and focus on intent + recency.