is about web analytics, data science and marketing strategy

This article takes only 2 minutes to read

Linear Regression vs Logistic Regression vs Poisson Regression


Generalized Linear Models (GLMs) extend the ordinary Linear Regression and allow the response variable y to have an error distribution other than the normal distribution. GLMs are great because:

  1. They are easy to understand.
  2. Most statistical packages contain functions to fit and interpret the resulting model.
  3. Linear regression or logistic regression is sufficient for a lot of real-life applications.

Click here to continue reading “Linear Regression vs Logistic Regression vs Poisson Regression”…

This article takes only 3 minutes to read

Modelling responses to market shocks

impulse response,

impulse response,

One of the frequently asked questions by marketers is what shall we expect from any market turbulences. This can be an increase in competitors’ spend on advertising, beginning of pricing war or any outlier of current market conditions. Let us address this issue on an example of a competitors high discount price using impulse response functions.
Click here to continue reading “Modelling responses to market shocks”…

This article takes only 4 minutes to read

Bayesian Modeling of Click-through Rate for Small Data

(cc-sa) Matt Buck

(cc-sa) Matt Buck

After a long break we get back to the topic of Small Data problems (read the original article HERE). This time we evaluate a modelling framework for metrics like Click-through Rate and Conversion Rate. In most business scenarios these are estimated with a fraction of Clicks to Impressions and Orders to Visitors. Most reporting suites use this methodology and most analysts are conditioned to ignore the occasional 0% CTR. Single value estimates do not work well when the amount of data is very small. What should we do if an ad has only 4 impressions and no clicks? Is the Click-through Rate actually zero? What if we had 4 impressions and 3 clicks? Is CTR close to 75%? Not really! When we have thousands of visitors, an accurate estimate can be achieved with a simple fraction. Once we enter the Small Data World, a different apparatus is needed to deal with uncertainty. The solution comes again from the world of Bayesian statistics.

Click here to continue reading “Bayesian Modeling of Click-through Rate for Small Data”…

This article takes only 1 minute to read

5 Hot Links for Week 38

Dear readers,
For your convenience, we have put together a list of informative and inspiring articles that caught our eye last week. Please let us know if such a weekly post is useful or just creates more information-noise. Good luck in the new week!

  1. Magnificent Mobile Website And App Analytics: Reports, Metrics, How-to! by Avinash Kaushik – As mobile becomes the primary channel for some businesses, marketing tools and strategies struggle to keep up. Avinash presents a thorough (as always) review of tracking strategies for both mobile webpages and mobile apps using Google Analytics.
  2. Digital Dashboards: Strategic & Tactical: Best Practices, Tips, Examples by Avinash Kaushik – At Marketing Distillery we believe that dashboard design is an art form. Having the right amount of detail while providing business-focused insight is more than critical in our data-overloaded world. The difference between tactical and strategic reporting levels often blur to produce a pulp of charts and numbers.
  3. YCombinator 2014 Data Science Startups – Checkout this list of Data Science startups for ideas and inspiration. Data visualization and machine learning are still at the top of the list of most sought job skills.
  4. D3.js Step by Step – The D3 library is almost a standard for data visualization. If you are a data scientist, or are thinking about becoming one, the ability to “sell” your insight is absolutely critical. A picture is worth more than a thousand words and this set of tutorials is a good starting point for building interactive and rich data visualizations.
  5. Predicting What User Reviews Are About with LDA and Gensim – Topic classification in reviews is a big problem when your app or website gains traction. Manually reviewing and classifying user feedback can be a boring process. This LDA-based methodology is a good starting point for an automated solution. Yes, the sources code is included.