When does multivariate testing (MVT) become an absolute necessity?

posted 28 Jan 2015, 02:24 by Dimitris Tsomokos   [ updated 16 Jul 2016, 02:39 ]

There are many possible statistical testing methodologies. It can be hard for marketers in the online marketing space to know which one to use; depending on the situation one of the methods could be better than another. In this Maxymiser blog post, I go over the two most basic testing methods for digital marketing - A/B (or split) testing and multivariate (MVT) testing. 

Patent on Customer Community Analytics

posted 19 Jan 2015, 09:04 by Dimitris Tsomokos   [ updated 22 Jan 2015, 03:11 ]

Maxymiser Inc. - a new method of customer segmentation using community detection in customer networks is now published, e.g. here: http://patentscope.wipo.int/search/en/WO2014177821



In this patent we have devised a new algorithm for community detection in complex networks and applied it to online & offline customer networks, which emerge naturally from customer affinities based on transactional and behavioural data. 

The method relies on the principal eigenvector of the adjacency matrix corresponding to the customer network. First we remove a random link in the base network and calculate the principal eigenvector of the new adjacency matrix (with the missing link). The vector of relative shifts between the first principal eigenvector and the second tells us about the community structure of the whole network, assuming we repeat the procedure for several missing links.

Maxymiser Blog - 5 Steps to Setting up a Successful Test: Statistics Best Practices

posted 15 Jan 2015, 09:57 by Dimitris Tsomokos   [ updated 16 Jul 2016, 02:39 ]

If you’re reading this, chances are that you’re either running a testing optimization program, or are about to. And with every a/b or multivariate test that you have run, or intend to run, there are some nagging questions that are sure to follow. These kinds of questions are very common and can be answered easily with the right statistical approach. I’m talking about questions like:

  • How long should my test run for?
  • I need to include at least 16 variants and I want to conclude the test within 2 weeks. Do I have enough traffic to do this?
  • My test hasn’t reached statistical significance – how much longer should I keep it running?

These sorts of questions can hinder your day-to-day work and possibly even disappoint your stakeholders. A major problem is that many people tend to think that the goal of testing is to identify ‘winners’ – that is, to find alternative content that drives higher conversions or revenue. Working under this false notion makes the questions above even more stressful and disconcerting...


Maxymiser Blog: Running multiple CXO tests simultaneously

posted 5 Sep 2014, 07:27 by Dimitris Tsomokos

Here’s a question we get a lot: “Is it okay to run multiple tests at the same time on a page or across pages and if so, how should it be done to ensure valid conclusions?” 

It’s a great question, but the response will be different depending on who you ask. Mathematically speaking, there are significant risks to your results (and ultimately your performance metrics) if you don’t carefully consider your options and their respective outcomes.

Read more here for the whole blog post.

Patent on Personalization & Testing in Online Marketing

posted 14 Dec 2013, 08:53 by Dimitris Tsomokos   [ updated 14 Dec 2013, 08:55 ]

Patent application is now published

Our innovation in personalization and content testing is about providing a new "Method of operating a server apparatus for delivering website content".

Talk at UCL's complex networks interest group

posted 14 May 2012, 05:18 by Dimitris Tsomokos

14 May 2012: Presented my latest work at the networks theory group at UCL, about link failures in complex networks and related considerations on quantifying complexity in networks (mainly from the point of view of quantum networks).

Title: "Trapped in Communities - detecting structure in complex networks using random walks and link failures"

Visit and Seminar at Harvard University

posted 27 Apr 2011, 14:46 by Dimitris Tsomokos   [ updated 23 Jun 2012, 02:15 ]

1-week research visit to the group of Alán Aspuru-Guzik in the Chemistry and Chemical Biology department at Harvard University.


28 Apr. 2011: Seminar - "Quantum walks on complex networks with connection instabilities and community structure" about this paper.


Slides from the talk are provided below.



APS March Meeting 2011

posted 26 Mar 2011, 06:25 by Dimitris Tsomokos   [ updated 26 Mar 2011, 06:35 ]

Sahel Ashhab (RIKEN & Michigan) summarized our recent work on the quantum simulation of exotic lattice systems with superconducting qubit circuits. 

The abstract can be found here and the talk itself is attached below.

Talk - Colleges of London Quantum Information Meeting

posted 15 Dec 2010, 12:21 by Dimitris Tsomokos   [ updated 28 Feb 2011, 14:09 ]

15 Dec. 2010: "Quantum dynamical probing of community structure in complex networks."
 
Talk given at the inaugural COLLOQUI meeting (UCL), introducing the basic ideas of my proposal on the use of quantum random walks as probes of complex network structure.

Slides from the talk can be viewed in the file below:

Seminar - Imperial College London

posted 12 Oct 2010, 11:47 by Dimitris Tsomokos   [ updated 5 Feb 2012, 12:10 ]

12 Oct. 2010: "Stability of topological order against random perturbations", presented at the Quantum Information Group, Imperial College; discussing work which led to this paper.

For further information, see the IC group's events page. Alternatively, you can access a video of the talk by downloading the attachment below.

1-10 of 17