Under The Hood: A Deep Dive Into What Makes Bidalgo AI So Effective

By Reenah Nahum

Bidalgo AI is a sophisticated, end-to-end ad automation platform for app install ads on Facebook, Instagram, Google, Snapchat, Pinterest and Apple Ads.

It uses advanced algorithms to manage and optimize the entire media buying cycle, from ad creation (Bidalgo AI is the only ad automation platform that can create new ads automatically) to campaign optimization, revenue optimization, and more.

We thought we’d take this opportunity to share some insights with you about how Bidalgo AI actually works.

There are three pillars of Bidalgo AI’s success: inputs, our various AI methodologies, and the results that our system delivers for each campaign. Understanding each pillar, and how they work together, yields fascinating insights into the future of online advertising.


The inputs, are three main pieces of information or parameters that our platform requires to begin delivering results. These include the following:

  • Your business goals – As you know, any marketing activity worth doing should serve your business goals. These are usually the business KPIs you want to achieve, and your budget. These parameters establish the foundation for Bidalgo AI’s work on your campaigns.
  • Your ad creative – Great creative matters, certainly in mobile app user acquisition. You can provide a range of ad creative such as images (including carousels and slideshows), videos, and a variety of other multimedia. Later, I’ll show you how our AI can actually improve your creative and help you design better performing ad creative.
  • Audience descriptions – The last key input is a description of your intended audience. This can factor in Custom Audience lists (whether you want to include or exclude these lists in your targeting), as well as specific geos you may, similarly, want to include or exclude.

Bidalgo AI can conduct campaigns without audience demographics and your preferred geographic targets, but, in our experience, this data usually improves campaign performance noticeably.

AI Methodologies

As you’ll see below, our solution uses various methods to deliver high quality results. This is possible because Bidalgo AI selects the most effective method based on your data and your specific objectives.

Using these parameters, it selects:

  • Time frames for your campaigns;
  • The most relevant variables to factor into its analysis; and
  • The method (or methods) that will consistently produce the most accurate results.

In addition, Bidalgo AI constantly assesses its selections and modifies them in order to ensure that it continues to deliver consistent and accurate results over time.

One of the keys to achieving this is a high degree of flexibility when choosing the methodologies the system employs, to better suit our customers’ changing requirements.

Prediction methods

The magic really starts when we apply our AI methodologies to your inputs. A significant part of our AI relies on advanced modeling. We use a variety of machine learning methods in two ways, in particular:

  • Prediction, such as predicting ad success rates, best performing campaign elements, and so on;
  • Classification and clustering, for example clustering similar clients for industry benchmarking.

Bidalgo AI also relies on association rule learning, mainly for daily trading and repetitive actions such as bidding, ad creation, pausing, and so on.

Generally speaking, our machine learning and statistical models consider many variables from which it selects a few, highly relevant variables for the final prediction model. These variables are selected using dimensionality reduction and regularization methods.

Bidalgo AI’s prediction methods are central to its success. The primary reason for this is that these models tend to predict causation, not just correlation.

Industry benchmarking

In addition, the platform also employs advanced industry benchmarking based on a variety of segmentation parameters.

While there are many use cases for prediction modeling, they aren’t appropriate for every customer’s requirements (for example, where customers have less complex datasets). This may be because applying these advanced models to data-sets not suited for them yields inconsistent or inaccurate results.

More conventional methods

Not every dataset produces sufficiently consistent and accurate results when analyzed using advanced models. Accordingly, Bidalgo AI’s toolkit also includes other, more familiar approaches.

One such method is a variation of traditional A/B testing. Most of our testing protocols are automatically structured by the AI. Typically, the next course of action is frequently determined by the test results of the previous analysis.

The testing is very robust when applied to similar entities (for example, ads and ad creative) at similar time intervals, while isolating their outcomes. As you may expect, this also applies to multi-variant testing where multiple variables are assessed and modified. We use a number of sampling techniques to choose the best test samples. At the same time, statistical inference methods help us form correct and unbiased conclusions.

A dynamic, iterative approach

There are instances when it isn’t possible to reliably determine causation based on the data-sets the platform has been supplied with. In these cases, Bidalgo AI analyzes correlations instead, using trends analyses in order to produce insights into markets and audience behaviors.

These layered and varied methodologies may seem time-consuming given their complexity but this is far from the truth. Bidalgo AI considers, applies, and iterates these methodologies (and more) in milliseconds, in real-time. It constantly analyzes, learns, and delivering improved results at an astonishing scale.


Using our various AI methodologies enables to deliver amazing results to our customers with end-to-end campaign optimization.

The ad creation process, itself, is greatly simplified as a result. With just one click you can upload top performing ads for your campaign. Bidalgo AI automatically determines the best targeting, bid, and launch pace combinations to produce top performing campaigns.

One of the platform’s best capabilities is the ability to determine the ideal times to launch your campaigns.

When it comes to delivery optimization, campaign management is similarly greatly simplified – our AI automatically pauses, reactivates ad sets, and, in rare instances, changes bids. Revenue optimization is one of our most important goals and achieving it is made possible using optimal budget allocation between the different ads based on predictions of ad sets’ success rates.

Bidalgo AI offers a powerful combination of:

  • Automated ad creation and delivery; and
  • Revenue optimization informed by Bidalgo AI’s strategic and creative recommendations and insights.

So far it has delivered 49% higher ROAS and 50% lower CPIs to our clients, on average. It has helped clients like Playstudios, Booksy and dozens more to achieve better performance and efficiency.

Another interesting metric we are following is our AI quality score. It measures the percentage of AI managed campaigns that outperformed or did as well as manually managed campaigns. Our current AI score is at 87% and increasing.

Sophistication, packaged simply

Drawing on your key, minimal inputs, Bidalgo AI leverages advanced machine learning methodologies to fully automate your media buying, ad creation and delivery optimization, and revenue optimization.

It frees up your time to focus on strategy. Bidalgo AI’s powerful strategic recommendations and insights empower you to gain a significant competitive advantage by scaling your marketing activities smartly and profitably.

Published on November 6, 2017
Written by
Reenah Nahum

Leading Data Science at Bidalgo. Reenah holds 2 master's degrees in Statistics and Information System Management. Previously a data scientist at eBay.

Get Started With Bidalgo Today