1. Home
  2. Knowledge Base
  3. Getting Started
  4. Which events are contributing to the thalox algorithm?

Which events are contributing to the thalox algorithm?

We are asked this question again and again in conversations with our customers and potential customers. That’s why we decided to write a short article about it.

It’s nice that we can see under the bonnet of thalox for marketers today.

We see a lot of confusion when it comes to machine learning and artificial intelligence. Having said that, we see a clear mandate to shed some light into it and clear up with some general concepts of machine learning.

To understand the topic a little better, you first must understand the term machine learning better. What is it actually?

At a high level, you can describe it with the use and development of computer systems that are able to learn and adapt without following explicit instructions, by using algorithms and statistical models to analyze and draw inferences from patterns in data. Especially the statistical dimension is important to understand. A machine learning model needs to work with data that has a positive correlation to a certain event.

What does this mean in the context of thalox?

We want to give our users the likelihood of their customers interacting with their brand through email marketing initiatives and in this context answering the question who will very likely click a call-to-action (CTA) within an email. By giving insights into this we help marketers to increase one of their main KPIs the so called click-through-rate.

Can you give us some insights into how thalox provides an answer to this question? 

Giving insights into this kind of question you need to follow these steps:

  1. Get the relevant data – in our context contact data and behavioral data on email interaction. Our algorithm combines contact data with email events like email send, email open, and email click to determine the likelihood of a contact clicking a CTA in a certain email.
  2. Clean the data and make it readable for machine learning.
  3. Identify the data that has a positive correlation to a certain event – in our context identify those attributes that positively correlate with clicking a CTA.
  4. Train a custom machine learning model with the identified data to generate results.
  5. Run the customer specific trained model and generate results.

This doesn’t sound very complicated. What are your customers’ main questions about this way of processing and what is your answer to it? 

We are often confronted with the question why certain data is not included in our model. Topics like “We think for our business the data about e.g. job title is very important and why is this not taken into consideration in the machine learning model” are regular concerns.

Our answer: It is all about statistics and positive correlation to a certain event.

To make it simple: Does wearing flipflops have an impact on or influence on the weather being sunny and nice?

The answer is a clear no. Wearing flipflops might be a result of but not a criterion for good weather.

In a nutshell, identifying the right attributes that have a positive impact on the event on interacting with a brand is key and not always easy to understand. Understanding and accepting this is key to being successful when using machine learning techniques. It is about understanding the result of vs. influence on a certain event.

Furthermore, in most of the cases there is not enough data across the whole data set available for a certain attribute which makes it irrelevant for machine learning.

Our answer: We provide data insights about fill rate of a certain attribute to give insights into identifying areas to improve data quality and fill rate of the same.

Providing this kind of information we give marketers the tool at hand to systematically increase accuracy and availability of their contact data.

What makes thalox different from other solutions out there?

Our main mission is to democratize machine learning techniques and make them easily available for marketers.

Our customers don’t need a team of data engineers and machine learning specialists to leverage the power of the same.

Our algorithm automatically detects the right data and generates a custom machine learning model based on the customer specific data and refines it over time.

Provide insights into our customers’ data to identify areas to improve their data quality. And last but not least continuous learning and improvement of the algorithm to take the latest data into consideration and automatically adapting the model whenever necessary.

That sounds exciting!

It is, and with every customer we learn and get better.

One final question to close the loop again. That means the more campaigns a client does, the better it is?

No, absolutely not. The marketer must create targeted campaigns based on the segments identified by thalox for marketers in order to improve the quality of the data and to develop the engagement of the contact base with the brand. Running targeted communication also prevents the database from being burnt because contacts are unsubscribing if they are getting spammed.

You can find more information about thalox for marketers here.

Or feel free to do a 30 day trial

Was this article helpful?

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *