Scribd
Upload
43 views6 pages

Scenarios: Mihir Jethwa 13 OCTOBER 2021

The marketing division of a manufacturing firm wants to understand if it needs to revamp its marketing program. They have data for different types of campaigns over 5 years. Analyzing the Always On, Volume Discounts, and Targeted Campaigns data could provide insights about customers, top dealers/products in different areas, and how promotions attract customers. This could help determine how to optimize the discounts and campaigns.

Uploaded by

Mihir Jethwa
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
43 views6 pages

Scenarios: Mihir Jethwa 13 OCTOBER 2021

The marketing division of a manufacturing firm wants to understand if it needs to revamp its marketing program. They have data for different types of campaigns over 5 years. Analyzing the Always On, Volume Discounts, and Targeted Campaigns data could provide insights about customers, top dealers/products in different areas, and how promotions attract customers. This could help determine how to optimize the discounts and campaigns.

Uploaded by

Mihir Jethwa
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 6

SCENARIOS

MIHIR JETHWA
13TH OCTOBER 2021
1: The marketing division for a manufacturing firm wants to understand if it needs to revamp its marketing program. The business is setup in a
Manufacturer to Dealership model where the funds are allocated by the manufacturer to the dealership for specific promotions. Assume you have data
available for 5 years for the following types of campaigns
• Always On: Ongoing promotions that provide the dealer a discount when they purchase above a certain amount in a month/year
• Volume Discounts: One-off discounts allocated to the dealer for purchases above a certain ticket size, during a promotional window (assume a 3-month
window at the end of each financial year)
• Targeted Campaigns: Localized campaigns targeting sales for specific products wherein the dealer provides a % discount to an end consumer and charges
the manufacturer back

Always On campaign data would help figure out the retaining customers under different dealers. So based on the ROI I can boil it down to specific products and
in specific areas where this campaigns can be executed.

Volume Discounts would help figure out the bigger sharks in market where bulk orders take place. Will figure out from those bigger sharks how many of them
actually avail the promotion in that window size. Accordingly window size can be altered to retain a set of customers around a dealer.

Targeted Campaigns would help us understand the pattern of a particular product in a particular area and the amount of consumers that buy on the basis of
discount %. Accordingly we can alter if specific products are to be marketed which heavy discounts and normalise the demand and supply scale.

If we combine all of them, it can revamp focusing which kind of campaigns for different locations as well as limited number of dealers too. Like if the data shows
80% of Chicago people buy product at 30% discount and they fall into the same category of dealer discount, we don’t need to provide extra discount. It can
reduce the implication charge.

Revamping these campaigns can be decided on the basis of ROI of every campaign as well as frequency can be altered on the basis of two things:
• Number of customers that are being retained because of the campaigns and
• New customers adding up during these campaigns

So on the basis of how customers are getting attracted either at a particular period only or a particular campaign works only in a particular area, we can revamp
the discounts ( amount and window).
1: The marketing division for a manufacturing firm wants to understand if it needs to revamp its marketing program. The business is setup in a
Manufacturer to Dealership model where the funds are allocated by the manufacturer to the dealership for specific promotions. Assume you have data
available for 5 years for the following types of campaigns
• Always On: Ongoing promotions that provide the dealer a discount when they purchase above a certain amount in a month/year
• Volume Discounts: One-off discounts allocated to the dealer for purchases above a certain ticket size, during a promotional window (assume a 3-month
window at the end of each financial year)
• Targeted Campaigns: Localized campaigns targeting sales for specific products wherein the dealer provides a % discount to an end consumer and charges
the manufacturer back

Insights that can be generated from the data :


• Customers focused in different areas in terms of monetary value and volume
• Top dealers in terms of monetary value and volume of products
• Top products and the lagging products that are sold in different areas where targeted campaigns are done
• Areas where promotions attract customers

ROI can be calculated as:

(Total Revenue generated from a campaign)


(Amount invested to set up the campaign)
2: Your team created a churn prediction model last year that would provide a list of Top N customers that are likely to churn in the next month. For the last
2 months the model performance has been de-grading leading a significant number of false positives. Please explain in your own words what could be the
possible reasons for such a change

I find it hard to believe that a model we built would have a problem but in case the model is having increasing number of false positives it might possibly
because of following reasons:
• New parameter being introduced and wasn’t considered in the model
• The present data is having more number of unusual cases or the model was trained on limited amount of data

We can introduce a workflow where every time new variables are added, weights would be revisited by training the model again.
3: A new product launched by your team required users to submit images of particular types of food and upload it on your website. We wanted to
understand a user’s dietary preferences using the image data in order to potentially recommend restaurants to the user. Currently none of the images
provided by the users contain any other information besides the image. Can you design and explain a process using which we can efficiently utilize this data
and achieve our goal?

I think following approach can be taken into consideration:

Image data can be used to identify the dish present in the image by using a Google API and then a word vector library which will put each identified dish into a
cuisine such that we would be able to understand the cuisine a user prefers at a regular basis as well as some unusual food.

For more information, we can further break it down further into calories where it provides the calories a particular food has.

Create a model for


Identify the dish Differentiate the dish into
understanding user diet and
using Google API different cuisines
map it with restaurants

Suggest a restaurant based


on user’s preference of
image/choice
4: The above equation is the mathematical basis of a popular classification algorithm. Please name the algorithm and explain conditions where it may not be
suitable to use this algorithm

The equation is of Logistic Regression.

Biggest Disadvantage is that in an unsupervised problem, Logistic Regression won’t work at times. So for those kind of clustering where we ourselves don’t know
the pattern, we will have to use a clustering technique like K-NN or K-means.

Another disadvantage is the amount of data, it should not be less than the inputs variables we provide

You might also like