The Fifth Elephant 2019

The eighth edition of India's best data conference

Participate

Optimisation using Julia

Submitted by Ginette V (@ginettev) on Friday, 12 April 2019


Preview video

Session type: Lecture Session type: Short talk of 20 mins

Abstract

While planning their marketing campaigns, our clients had to understand how their marketing spend affects their KPIs. We created models to understand the effect of individual marketing channels such as TV, Radio, Digital etc on KPIs like sales, qualified reach or profits. We had to help them to build optimised brand plans and campaign plans that use the allocated budget effectively.

Knowing how the marketing spend affects the KPIs enables us to optimise the marketing spend for maximal result. We also had to suggest the optimal plan to achieve desired business outcomes within user defined constraints, e.g. user defined % range of permitted spend changes.

We used Adbudg S-curves to optimise the marketing spends as they have several useful characteristics for optimisation and it is easier to find the point where the ROI is maximised. Using the response curves for each of the individual channels (TV, Radio, Digital), we could find the optimal spend for each individual channel with an easy to solve optimisation problem. The result is an optimal marketing mix that maximizes the chosen KPIs.

In this talk we will delve into how to find the optimal marketing mix using S-curves and Julia.

Outline

  1. Introduction to the Marketing domain and the optimisation problem
  2. Model and data considerations
  3. How we went about choosing Julia
  4. Walkthrough of the solution
  5. Challenges faced
  6. Conclusion

Requirements

Basic understanding of Data Sciences

Speaker bio

Ginette is working as Senior Developer at ThoughtWorks for 9 years. She has worked on solving problems across multiple domains like retail, marketing and publishing.

Links

Slides

https://drive.google.com/file/d/1uDlyMJd--ibGGlhICWQL9fP8GdXPl__I/view?usp=sharing

Preview video

https://drive.google.com/file/d/1QAk2zkJ-CamEyQMf1m4gCgD7HJqFY93L/view?usp=sharing

Comments

  • Anwesha Sarkar (@anweshaalt) Reviewer 2 months ago

    Thank you for submitting the proposal. Submit your slides and preview video by 20th April (latest) it helps us to close the review process.

  • Ginette V (@ginettev) Proposer 2 months ago

    Yes, will upload soon.

  • Ginette V (@ginettev) Proposer a month ago

    Slides have been uploaded, but they are work in progress with outline in place.
    I will be filling in and polishing the slides soon.

  • Zainab Bawa (@zainabbawa) Reviewer a month ago (edited a month ago)

    Thank you for the slides and video, Ginette.

    Below are some of the comments and questions (responses of) which you have to add to your slides so that we can close the evaluation:

    1. Since this is an implementation for a single customer, show how the problem is more general, and not limited to a single use case.
    2. Why did you choose Julia for the solution? Which other options did you evaluate and what were the criteria for evaluation? Why did Julia emerge as a better option?
    3. What is the before-after scenario? What are the outcomes of the implementation with Julia?
    4. How did the customer/customer’s team adapt to Julia? What resources, including skills, are required to adopt Julia in such scenarios?
    5. How can participants – at The Fifth Elephant – make decisions on whether Julia is an option for them if they are facing similar problems?

    Upload your revised slides, with the above details, by or before 21 May, so that we can close the decision on your proposal.

  • Ginette V (@ginettev) Proposer a month ago

    Thanks for the comments. Will update the slides and upload soon

  • Ginette V (@ginettev) Proposer 28 days ago

    I have uploaded the slides with updates wrt points 2 and 5.
    For point 1, I wanted to focus more on optimisation in general and not just the optimisation with context to marketing, hence did not spend much time showing it is general as time is limited for the talk.
    For point 3, I hope to be covering it with the code which I will be demoing.
    For point 4, as the customer did not go to production with the Julia implementation, I will not be able to suggest much in this situation.

Login with Twitter or Google to leave a comment