Maxia AI is in closed beta. We're looking for pilot users for our new Segment integration. You can request an invitation here.
Need a down-to-earth explanation on AI? We'll try to answer some of the basic questions that should help you get started with our AI tools.
What is an AI model?
You can think of an AI model as a special system that is designed to solve a complex mathematical problem, by finding patterns in datasets. At Maxia, we are designing different kinds of AI models that use this math to help solve real-world marketing and business problems.
What are datasets?
A dataset is an organized collection of data. AI models need to be specially designed to work with specific datasets, and datasets are organized so an AI can understand the data. A dataset is a collection of data samples, and each sample is a collection of different features and labels. Usually, more data features and samples in your dataset can lead to a better AI model.
What are samples?
Data samples, sometimes called records or rows, are the collection of data objects in your dataset. Maxia AI's are person-based, so each sample in a dataset represents a different visitor, user, or customer of your business (e.g. User A, User B, User C, ... ).
What are features?
Data features, sometimes called fields, or columns, are the different attributes of the data objects in your dataset. Maxia's AI's are person-based, so each feature represents a different piece of information about each visitor, user, or customer of your business (e.g. User Country, User Device Type, etc.). Features have a data type, which means that the AI model knows how to interpret the data in a specific way, for example numbers need to be handled different than text.
What are labels?
Labels, sometimes called targets, are a special feature in your dataset. The label is the attribute that we try make the AI predict, so the AI model tries to learn about all of the relationships between the label, and all of the other features in your dataset. In order for your AI to make predictions, it needs to learn from a large enougn number of samples provided in the dataset.
What is a CSV File?
A comma-separated values (CSV) file is a text file that uses a comma to separate values. A CSV file stores a table of data, where each line of the file is a row of the table, and each row consists of values separated by commas. It is a very common format that most spreadsheet and data-related software can export to.
What are trait-based features?
Trait-based features are the descriptive attributes of each of your samples. Maxia's AI currently supports categorical and boolean traits. Numeric trait support is coming soon.
What are categorical trait features?
A categorical feature is a data type has many possible distinct values. It is a very general data type will represent most of the data in your dataset. For Maxia's person-based AI's, consider what categorical traits you know about your visitors, users, or customers:
- What country are they from?
- What kind of device do they use (iPhone, iPad, Android, etc)?
- How did they hear about your business?
- How did they first come to visit your website or app?
To represent categorical traits in your Maxia AI dataset files, we recommend using non-numerical text.
What are boolean trait features?
A boolean feature is a unique categorical data type that has only two possible values. Some common ways of thinking about boolean data are:
- Yes or No
- True or False
- 1 or 0
A boolean feature cannot have any empty (null) values, as that would be a third possible value. For Maxia's person-based AI's, consider what boolean traits you know about your visitors, users, or customers:
- Did they sign up for your email newsletter?
- Are they using your mobile apps?
- Are they signed up for a free or paid account?
To represent boolean traits in your Maxia AI dataset files, we
What are numeric trait features?
A numeric feature is a unique data type meant especially for numbers. Unlike categorical data, where it can difficult to relate different values (for example a country feature with values like Canada, United States, and Australia, or a device features with values like iPhone, iPad, or Galaxy S20), numeric values are all related to each other - we know how to order them from biggest to smallest. Maxia AI support for numeric features is coming soon.
What are behavior-based features?
Behavior-based features are the distinct actions of each of your samples. Maxia's AI does not currently support behavior-based features. Date and Datetime behavior support is coming soon.
What are date behavior features?
A date or datetime feature is a unique datatype meant to represent the moment a particular action was taken by a visitor, user, or customer. Maxia AI support for date and datetime features is coming soon.
What is AI model training?
AI model training is the process of learning how the labels in your dataset relate to your features. Maxia's person-based AI's need historic visitor, user, or customer data for their models. We use your most historic (i.e. least recent) data from the dataset to train your AI.
What is AI model evaluation?
AI model evaluation is the process of determining how well the model is able to relate the labels in your dataset to your features. Maxia's person-based AI's need historic visitor, user, or customer data for their models. We use your semi-historic data from the dataset to evaluate your AI.
What is AI model prediction?
AI model prediction is the process of creating new labels based on the features in your dataset. Maxia's person-based AI's need current visitor, user, or customer data for their models. We use your current (i.e. most recent) data from the dataset to train your AI.
Maxia Conversion AI
Learn more about how our conversion AI works, and about the business and marketing problems it helps solve.
What is conversion?
Conversion is how you measure the actions taken by a group of customers, users, or visitors, of your business, website or app. In your business, there are key actions taken by your customers. Some of these may include:
- A new customer signing up for your business
- A new customer making their first purchase with your business
- A returning customer making an additional purchase with your business
- A new customer subscribing to your business
- An existing subscription customer upgrading with your business
The most basic measure of conversion is conversion rate. It is calculated as the total number of customers that converted, divided by the number of potential customers.
Conversion Rate = Converted Customers / Potential Customers
For example, if 100 potential customers visited your website, and 15 of those made a purchase, your conversion rate is 15/100 = 15%. Once you are tracking and analyzing conversion for your business, you can take decisive actions to increase conversions, this is called conversion rate optimization.
How does Maxia measure conversion?
Maxia's person-based AI analyzes the customer dataset
you upload to determine the conversion rate of your customers. Using two of the required conversion AI
converted_at, Maxia analyzes how many of your customers
converted, and how long it took each of them to convert (in days). Maxia's AI provided you with a
conversion analysis, so you understand how long it takes your customers to convert overall. Maxia AI uses
the 90th percentile conversion time when building your AI. By ignoring the customers that took much longer
to convert, our AI focuses on analyzing the more typical customer.
Maxia Conversion Analysis Sample Screenshot
What insights does Maxia's conversion AI give me?
Maxia's AI automatically analyzes your conversion data and provides you an AI report about your different groups of users. By analyzing impact, frequency, and importance, you can identify groups of users that should be actioned upon:
- What do my best converters look like? What can I do to grow the size of my best converting groups?
- What do my worst converters look like? What can I do to shrink the size of my worst converting groups? What can I do to increase their conversion rates?
Maxia Conversion Insights Sample Video
What is Impact?
Impact is the difference in the conversion rate between people that have a given feature value and the average person in your dataset. Positive impact means that this group converts higher than average, and negative impact means that they convert lower than average.
What is Frequency?
Frequency is how often this feature value is found in your dataset. When frequency is high, that means many people have that feature value. When frequency is low, that means few people have that feature value.
What is Importance?
Importance is how valuable of a signal is this feature value to our AI. When importance is high, that means that the feature value is very helpful to making better conversion predictions. When importance is low, that means that the feature value isn't as helpful for predicting conversion.
What predictions does Maxia's AI give me?
Maxia's AI analyzes your dataset for users that have not yet converted, and are still within the 90th percentile conversion time. Maxia's predicts the probability of each of those users converting, and lets you know how long into their conversion window each user is. Top predictions are displayed in the app, and all predictions can be downloaded in CSV format.
Maxia Conversion Predictions Sample Screenshot