The Fifth Elephant 2019

Gathering of 1000+ practitioners from the data ecosystem

Leveraging Power of Analytics for Martech

Submitted by Jacob Joseph (@jacjose) on Feb 11, 2019

Session type: Full talk of 40 mins Status: Confirmed & Scheduled


Marketing Technology has undergone a technological revolution over the past 10 – 15 years. Today marketers are able to track the smallest of digital footprint like scrolls on mobile or web apps. Armed with the digital trove of user behavior data, marketers are trying to nudge and retain their users across the customer lifecycle.

But, why are conversions per campaign in low single digits? Why does an app lose 90% of the newly acquired users in the first month?

A marketer of a medium sized app (MAU ~ 20,000 and 50 events per user per month) has to analyze atleast 1 million data points monthly. In order to be agile, marketers tend to cut corners and take sub-optimal data driven decisions. As a result, conversions rates are poor. A sub 1% conversion rate is quite common i.e. 99 users out of 100 are not interested in the app’s messaging. This not only means that the marketer has poorly allocated the resources and runs the risk of underachieving the target KPIs but also antagonize the user experience which could result in churn or uninstalls.

CleverTap is a leading Customer Lifecycle Management and Engagement Platform which help apps retain their users for life. The solutions that we develop have to be best-in class, generic (usable by any app) and highly actionable. This talk will discuss our approach to solve 2 keys issues faced by marketers:

  1. Identify that elusive segment that gives the maximum ROI based on pre-defined goal within a pre- defined time
  2. Hyper personalized messages that resonate with their users resulting in increased conversions

The key takeaway for the audience is the use of temporal user behavior to create:

  1. Dynamic User Clusters
  2. Recommender Engine


  • Brief about CleverTap
  • Current State of the Industry
  • Challenges faced by Marketers on Segmentation
  • CleverTap’s solution for Intelligent Segmentation with Machine Learning
  • Case studies showing the real impact of CleverTap’s solution
  • Challenges faced by Marketers on Campaign Content
  • CleverTap’s solution to Automating Campaign Content with Recommender Engine
  • Case studies showing the real impact of CleverTap’s Recommender Engine


Machine Learning enthusiasts eager to solve business problems

Speaker bio

An Investment Banker by accident and a Data Scientist by choice, Jacob has successfully transitioned from the world of finance to the world of nerds in analytics. Jacob brings over 15 years of combined experience in analytics, consulting, portfolio management to solve complex business problems faced by marketers with the help of cutting-edge industry-first solutions powered by Data Science.

Jacob is a proud recipient of 40 under 40 Data Scientists (2019) awarded by Analytics India Magazine. He has written multiple articles on data science which have been picked up by renowned sites like kdnuggets, datasciencecentral.



Preview video


  • Anwesha Das (@anweshasrkr) a year ago

    Thank you for submitting this proposal. We require slides and preview video by 11th March, latest, to evaluate your proposal and make a decision.

    • Jacob Joseph (@jacjose) Proposer a year ago (edited a year ago)

      Hey Anwesha, Kindly review my submission. I have included the slides.

  • Anwesha Sarkar (@anweshaalt) a year ago (edited a year ago)

    Submit your preview video by 21st March (latest). Without a preview video it is difficult for us to have a fair evaluation to your proposal.

  • Zainab Bawa (@zainabbawa) a year ago

    Thanks for the submission, Jacob. Couple of questions:

    1. What is the takeaway for participants from this talk? What do they learn and what can they apply at their work?
    2. Not everyone in the audience will be marketeers. In which case, what is the takeaway for particpants who don’t work with marketing, either directly or at all?
    • Jacob Joseph (@jacjose) Proposer a year ago

      Hey @zainabbawa ,
      I have added the key takeway to the abstract. Basically the takeaway for the audience is the use of temporal user behavior to create dynamic clusters and recommender engine

  • Zainab Bawa (@zainabbawa) a year ago (edited a year ago)

    Hello Jacob,

    We have evaluated the proposal. Thanks for the slides.

    Here is the review and suggestions for restructuring this:

    1. The slides are currently a company pitch for Clevertap. The presentation should focus on domain knowledge and concepts rather than your company. It is obvious to the listener that your company has the capabilities you are talking about.
    2. Without compromising a client’s confidentiality, illustrate your concepts with more concrete examples.
    3. What do “event” or “goal” mean? This is unclear in the slides.
    4. Can you take one case out of the clients you have cited. For example, you could talk about Hotstar trying to predict which of its customers can be converted from a free to paid account and what kind of campaign and pricing is required. Alternatively you could talk about predicting which of its paid customers will renew their account based on their usage data. A clear business goal in terms of the profit/loss problem at hand puts things in good perspective.
    5. Quantify the impact of your intervention, ideally with $$$ figure.
    6. A brief intuition on why recency and frequency are important variables capturing a lot of information. This may not be clear to everyone.
    7. Also why do you not consider money spent by customers in each event? Usually, Recency Frequency and Monetary value (RFM) are considered jointly in customer analysis (see point above).
    8. What you mean by “manual segmentation”? ML-based techniques such as cluster analysis have been implemented by big companies for 30+ years now. The scale has of course increased with faster computers, more memory and (occasionally) more sophisticated techniques.
    9. Focus on general intuitions rather than specific functions. For example, xgboost is an implementation of gradient boosting. Participants will get more value from knowing the intuition behind the algorithm than the specifics of xgboost.

    If you are interested in pursuing the proposal, you have to change it substantially to structure it as per the above requirements. Upload revised slides by 27 May latest, to close on the decision on this proposal.

    • Jacob Joseph (@jacjose) Proposer a year ago

      Hey Zainab,

      Please have a look at the updated slides which is almost entirely restructured. Some of the points mentioned in feedback will be easier to explain in the talk. I have inputted relevant slides to cover the same. But the talk will cover those in greater depth. Let me know your feedback.

    • Jacob Joseph (@jacjose) Proposer a year ago

      Thanks Zainab for the detailed feedback. I will try to meet your deadline but may overshoot it as it requires a complete overhaul.

      • Abhishek Balaji (@booleanbalaji) a year ago

        Hi Jacob,

        Let me know if you had a chance to work on this. If not, will move it for a future conference.

        • Jacob Joseph (@jacjose) Proposer a year ago

          Please give me couple of days more. I will upload the reworked slides

        • Jacob Joseph (@jacjose) Proposer a year ago

          Hey Abhishek,

          I have uploaded the revised presentation

          • Abhishek Balaji (@booleanbalaji) a year ago

            Hi Jacob,

            I did take a look at the slides again and it still reads like a company pitch. It does not contain any technical details which would be useful for someone in the audience. While I understand the importance of Martech, the slides are just about ‘Cleartap’ and how your solutions are the way to go.

            Unfortunately it does not meet the criteria for a community submission at The Fifth Elephant.

            You may choose to update the content so it can be considered for a future event. Here’s an example of a talk which balances discussing martech and advertising tech for a technical audience:

          • Zainab Bawa (@zainabbawa) a year ago

            Jacob, you have to change the permission settings on your slides to public. I am unable to access the slides.

  • Zainab Bawa (@zainabbawa) a year ago

    Jacob, you have to change the permission settings on your slides to public. I am unable to access the slides.

  • Zainab Bawa (@zainabbawa) a year ago

    Thanks for sharing the revised slides, Jacob. I will have these reviewed in the next two days. From a quick view of the slides, the content becomes interesting from slide 7 onwards. The earlier slides are about Clevertap which makes the proposed talk look like an advertisement for Clevertap. Whereas, as I have mentioned in the feedback, the proposed talk should be about a general problem and not about Clevertap.

    • Jacob Joseph (@jacjose) Proposer a year ago

      Hey Zainab, Slides 3 to 5 deal with who is our customer and the kind of problems that they face. This a typical of the any growth marketer or product experts who use retention marketing tool like CleverTap. This is not limited to just our customers but is general in nature.

      • Zainab Bawa (@zainabbawa) a year ago

        Got it! Thanks Jacob. We will complete the review of your proposal today, and share an update by tomorrow.

  • Zainab Bawa (@zainabbawa) a year ago

    The revised slides are definitely much better. The presentation is shaping up quite nicely.

    There is one area to work on though. In the “how we do it” slide, any segmentation – or for that matter any analysis in any context – must be dictated by a clearly specified question. For example, how you segment your customers will vary greatly on whether you were doing this to acquire or retain customers, or whether you were doing this for an ad campaign or aligning your sales force. This is one major thing missing in the workflow. All sophisticated analysis is useless unless you ask the right question. Adding this detail will make the presentation far more richer.

    We have confirmed your talk, and will be setting up a pre-event rehearsal followed with logistical details.

    • Jacob Joseph (@jacjose) Proposer a year ago

      Hey Zainab, Slide on “How we do it” has to be read with the earlier slide “What our customer wants”. That slide speaks of the problem statements such who, when, what, why, where. In the current talk, I am discussing how we solved the “Who” part with Automated User Segmentation and “What” part with content recommendation. I thought it best to explain this during the talk when i get to the slide “What our customer wants”

Login to leave a comment