Stay marketing and tech-savvy. Get the latest in martech - subscribe to MarTech Today.
Jaywing launches neural net-powered Archetype predictive engine for non-technical marketers
The UK-based marketing firm says the new engine has greater accuracy and flexibility than standard prediction tools.
It’s getting difficult to remember back to the time when marketers didn’t have prediction engines that could forecast if this person or that company would become a customer.
Nowadays, predictive analytics of many kinds are standard tools. But they could be better, says UK-based, data science-based marketing agency Jaywing, which has released a new neural net prediction engine that is designed for non-technical marketers.
It’s called Archetype, and Jaywing says that it has demonstrated a 10 percent greater accuracy in predicting financial services fraud — such as fraudulent credit card applications — than standard models. The engine is designed for marketing, credit risk and fraud detection.
Head of Product Development Martin Smith told me that marketing use cases should see a similar kind of increase in accuracy. Additionally, he said, Archetype allows a user to constrain the model with certain business rules, such as stipulating that there’s a greater likelihood a customer will buy if her salary increases.
While there are “definite advantages” if a user has a bit of experience in data modeling, Smith said, it isn’t necessary, and Archetype is designed for use by non-technical marketers.
Acknowledging that neural nets have been around for a while, he said they have previously been limited by the cost of computing power, plus there have been some recent “math breakthroughs.”
In Archetype, a user uploads data, such as in a spreadsheet. There might be browsing histories, purchase histories and demographics on customers who bought a product and those who declined, in addition to browsing histories, purchase histories and demographics on customers who have not yet been presented with the product.
After the data is uploaded, the user tells the system what kind of outcome is being modeled — that is, the kinds of predictions desired. The interface allows the user to indicate what the columns of data mean, and the system determines which variables are the most relevant. The results are then tested against known results and refined until it’s ready to run.
At the moment, Archetype is not yet integrated into other marketing tools, but Smith said he expects an API and export functions to be available within the next two months.