Squeeze out the last 5% of conversion rate with AI

In online sales and marketing, enormous amounts of data are collected that reveal a great deal about user behavior and user preferences. Making this data usable is the main goal of Acceleraid. For this purpose, we use various machine learning algorithms, from Markov models to neural networks.Today, AI, big data, and machine learning are mainly used by IT departments, business intelligence, and data science. Our focus is on making complex technology usable for everyone. Through AI-based data analysis and personalized customer engagement, we support you in your sales activities and help you increase your deals and revenue.

To do so, our software analyzes your customer and visitor data, segments target groups, automates the outreach process, and optimizes your channels based on all click and transaction data using machine learning. Our algorithms are flexible and versatile. We help you optimize them for your business case.

Contextual Bayesian Bandit

DATA-BASED CONTENT PERSONALIZATION

Not every website design or email is equally successful with every customer. We optimize how users are addressed by showing each visitor a variant tailored to them. Acceleraid offers several options for optimizing this delivery: optimization can be performed with a neural network as well as with a parameter-based Bayesian bandit.

In ordinary A/B testing scenarios of similar tools on the market, different variants are tested against each other and a preferred variant is chosen when statistically significant differences are found. Our machine learning algorithms work iteratively and converge significantly faster toward the best possible traffic distribution.

HOW DOES IT WORK? (USING THE BAYESIAN BANDIT AS AN EXAMPLE)

The underlying mathematical problem is a so-called “multi-armed bandit” problem. Here there are several triggers (~variants) with different unknown chances of winning (~conversion rates), which are to be pulled in such a way that the total gain is as large as possible.

One of our solutions for this is a further development of the Contextual Bayesian Bandit. It calculates the probability densities of the real conversion rates of different variants based on the data collected so far. A conversion rate is then drawn from these probability densities for each incoming user, and the best one is delivered.

WHAT DOES IT DELIVER?

This algorithm converges significantly faster toward an optimal traffic distribution, and it also makes it possible to optimize for a user's context, i.e., individual probability densities are estimated for each context parameter. The algorithm then independently and automatically recognizes, for example, that smartphone users from northern Germany on Mondays prefer a different website variant than desktop users on Friday, and delivers the design, product, or messaging optimized for the respective visitor accordingly.

Trends and developments in user preferences over time can also be recognized through dynamic weighting of the data in the densities and used for optimization. Our Contextual Bayesian Bandit is already in use at many companies and achieves optimization effects between 10% and 30% there.

Multi-armed Bayesian bandits enable the estimation of user behavior through probability distributions

• Requires no training data or statistical significance but learns from day 1

• Processes all parameters for estimating the conversion

• Uses all data for learning and optimizing and detects trends

In comparison, SCORING optimizes for the best variant in the text for each respective user

• Requires training data to calculate the ratings

• Focuses on parameters that are particularly relevant

• Uses training data from a predefined time window

With NEURAL NETWORKS, learning takes place by training an abstract network

• Requires training data to optimize the network

• Processes all parameters for training the network

• Uses training data from a predefined time window

Clustering with HDBSCAN

To make sure your messages also reach the right target group, we apply various automated clustering methods to your customer data. Using a density-based method based on HDBSCAN and fuzzy clustering, we generate meaningful and usable customer segments from your customer data.

All of these customer segments can be made directly usable for personalization by our system and provide insights into your customer base.

Process Automation

Intelligent process optimization for large volumes of data saves time and costs. Automatic optimization of business processes, image and text recognition, churn detection, and sentiment analysis – there are numerous fields of application for machine learning.Toward hyperautomation – we know how to train artificial intelligence: With Acceleraid's software, you can launch rule-based automated mailing campaigns as well as have them optimized automatically by the Bayesian Bandit. The Bayesian Bandit independently and continuously learns which email variant and which content achieves the highest conversion rate with which customers.