Disruptive Marketing Blog

What happens when our algorithm learns?

The information you can find when searching machine learning or intelligent algorithm is basically the same one. When you find companies like us, who work with an algorithm or machine learning, they will only show the results of their algorithm. That is fine, isn’t it? But, would you like to see an algorithm in action?

First of all, let me explain why we have created our tireless boot and our goals.

Our algorithm has two main objectives:

i) Understand why and how users are performing in your site.

Adinton can answer this question since we save and organize all the information about your traffic. Adinton saves every single click from all your ads (adwords, Facebook ads…), direct traffic, SEO, referrer… even impressions and impression viewed.

But Adinton does not just track the access clicks to your website, Adinton tracks every single page view inside your site.

Once all this information is organized, Adinton can answer different questions like:

  • How are long tail (adwords) keywords influencing your direct traffic or brand traffic?
  • How are display campaigns, affiliates or coupon sites influencing the performance of my site?
  • How are mobile devices influencing my business?
  • How much new business are we generating? Or are we just selling it to recurring clients?
  • What are the main keywords/channels that are generating more new clients? So, where should we invest for generating new business?
  • What are the channels that our recurring customers normally use for buying?
  • What channel is the most profitable even if the conversions are done by a call centre team (offline)?

ii) Boost your site performance.

But, as we mentioned before, we do not just see “the click moment”, we analyze what your users are doing inside your site, so we are also able to answer questions like:

  • What channel/keyword is generating more “add to cart”?
  • What channel or keyword is generating more page views or micro conversions…?
  • Which of them have the lowest bounce rate?

Once all the information is organized, it seems very easy to learn how to make the online marketing budget profitable, doesn’t it?

At this point, we would just need to know your goals, budget goals, CPA, conversions, ROI,… then Adinton starts to save > organise > Think > Make Conclusions > Take decisions.

Take a look at how Adinton is moving the Adwords bidding according to how the competitors, conversion rate, CPA are moving:

  1. The algorithm moves softly the CPC because we are achieving our CPA goal of $12.
    1. Look how it is decreasing CPC while trying to be closer to the $12.
    2. When Adinton detects it, we are working with CPAs lower than $12, started to increase CPC, for getting better positions > CTR > Clicks…
  2. The CPA starts to make big movements but Adinton does not change the bid, it is waiting with a low CPC, the CPA goes down. Adinton learnt that in previous days working with CPCs close to 0, 56€ can achieve conversions with $12 CPA.
  3. Once it detects the CPA it does not go down, it starts to increase the bid to those keywords that help others to convert them and also closer deal keywords.
  4. It detects the CPA go down drastically and quickly recovers the old and “comfortable” CPC, even when the CPA is moving drastically.

Let me ask you something: How many hours would you have to spend in order to optimize it like Adinton did?

We can be even more exact about the effort that we should do for at least, make the same volume of changes.

 

 

There’s day with less than 10 changes and other days with 1.200. we should keep in mind, that this charts belongs to a small Adwords Account (less than $15,000/mo, greater than $5,000/mo). You should think of depending the number of keywords, even 500 daily changes are not enough (?).

Well, we need to consider that every change is done needs a previous study about ROI optimization. So, what we really find in this chart  is how many times the tool is making calculations that can generate changes. Really Adinton is making calculations more often that the changes we see in the chart, but sometimes their studies is not triggering any change.

In other hand, we can see how the changes are more aggressively while Adinton’s algorithm is learning.

And here is, how many changes we are doing per second.

 

In other words, these are some of the results:

Navarra University (lead generation):

We have decreased their CPA by 50% the first year we started working together, and by (-) 40% the second year. On the other hand, we have increased their business by around (+) 38% and decreasing CPA by  (-) 30%.

Pet Sonic ES (ecommerce):

Each quarter we duplicate their revenue.

Petsonic FR (ecommerce):

In the last semester we have got x10 their revenue with just x3 online marketing investments

API.cat (lead generation):

We hace decreased their CPA by 50% and increase their conversions x2 yearly.

We understand that the real goal of a company which is monthly investing in online marketing is just INCREASING their BUSINESS.

When our clients contract our tool or our services, their aim is to increase revenue keeping an affordable CPA. And this is what Adinton does, to increase revenue:

  1. First of all, it is important to understand the behaviour of your traffic.
  2. Then, to clean up the sources and/or keywords that are not helping to perform.
  3. Boost those profitable keywords/sources.
  4. Adjust bids and budgets to maximize goals.

If you compare the Adinton’s optimization process with how we achieved the results commented before, you’ll find a clear correlation.

This is the Present of Online Marketing:

  1. Big Data (collect and storage)
  2. Machine Learning. Take decisions about your own data in a changeable environment.