EU Cookie Consent

By clicking “Accept All ”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts.

Essential Data

The Essential Data is needed to run the Site you are visiting technically. You can not deactivate them.

Optional Data

You can choose wether to accept those non essential cookies

Loading...

From Scibids
Thought Leadership - Op-Ed

AI driven custom algorithms are essential to achieving custom KPI’s like attention or qualified traffic at scale!

In the 2nd part of this 2 part series we will be covering off why custom algorithms are a crucial part of a programmatic traders toolbox when custom KPI’s such as Attention metrics need to be achieved.
By James Whitbread - September 2022


Kicking off the 2nd part of this 2-part series with a thank you for all of your amazing feedback to part 1 of this series on AI driven custom algorithms. Amazing to hear so many brands and agencies realising the potential of this unique solution and also stepping forward for pilot opportunities.

So, after covering what you can achieve using the metadata in your programmatic bid streams to derive contextual signals which provide insights to custom bidding scripts, I now wanted to pivot and look into why AI powered custom algorithms are crucial in unifying an ad tech stack.

Locally in AUNZ, from our conversations out in the market and from the many Linkedin articles there is now a huge focus on the value of measuring attention. Furthermore, it’s great to see lots of fantastic businesses with an offering in this space including DoubleVerify, Playground XYZ, IAS, Adelaide, Amplified Intelligence and many more.

Although watching on with interest, I will leave the conversation around the true value of attention as a metric to the experts in this space. For the purposes of this article, I’d like to offer an opinion on 2 considerations for businesses who are looking to buy attention driving impressions through a DSP and of course which are measured through these 3rd party tools.

  1. “We have measured which impressions are considered valuable to our attention goals; now how do we buy more of them at scale?”
  2. “Now that we know which impressions and other variables are driving attention for this campaign/IO how do we buy more at an acceptable price?”


I have seen numerous posts including this one from the past week making the point that without some attention, reach is pointless. The focus here is on the “some” as it does not have to be a full video watched, display ad viewed to IAB standards etc to drive brand results and/or outcomes through impressions driving attention.

I loved this quote “It’s about finding the right level of attention for a given brand trying to achieve a specific outcome” as this really summarises perfectly custom algorithms value in this space, optimising per IO towards a specific KPI at scale and within budget.

AI driven custom algorithms are crucial as essentially the middleware between postbid measurement of attention and with the aim of scaling the buying of attention impressions through the DSP. Unfortunately for traders using standard algorithms or leveraging human-built algorithms, there is not a magic button which automates a DSP to buy more of what we are told via measurement is working.

AI driven custom bidding algorithms will deduce patterns using the most granular level of impression and attribution data. In fact, combining hundreds of millions of variables contained in bid-stream data with 3P measured metrics (sometimes also at Log Level) ensures Scibids can establish exactly what factors are contributing towards active attention.

There are many other reasons why self-learning and self-updating AI driven custom algorithms are able to be that crucial middleware piece, however at a macro level I cannot stress enough the importance of operating towards the KPI’s of an individual IO.

Of course, what impressions or more granularly contextual factors in bid stream metadata, drive active attention for one campaign will likely not be the same for another. Separately, changing macro-economic factors will also have an effect on consumer behaviour such as the URL’s they visit when comparing products and how often. These can have a profound impact on campaign performance and thus custom bidding models need to be refreshed often, hint hint every 3 hours is pretty good! Last week alone our AI pushed 102,337 new algorithms, something manual human models just cannot achieve.

In short AI driven custom algorithms enable an automation between measurement and DSP to buy more active attention taking out the guesswork from a trader’s perspective!

OK, so now we have solved the reach problem, what if the impressions the AI/Trader is finding are of higher “market” quality and thus are commanding a higher price? Guess it stands to reason that particularly in the tight BVOD space in Australia some premium impressions are going to be both providing active attention and are going to require higher bid prices to win.

The beauty of AI driven custom scripts like in the picture above, which Scibids push through to the likes of DV360 is that you will always be bidding to reflect the true value of each impression towards the KPIs of each IO. Those “Super impressions', yes, you may have to pay more for as you absolutely have to win them. Let’s not forget though to achieve scale there will be others driving partial active attention so let’s also bid on them but according to their value!

Attention metrics are of course going to be crucial to the likes of FMCG brands as a barometer of success as they do not have web analytics data, or online attribution data to optimise and learn directly from.

Whether you are Coca Cola who may be happy to accept “minimal attention” but at huge scale and with a success metric of a lower CPM or a Super Fund Brand looking for focused attention with less scale than Coca Cola and are happy to pay a higher CPM then custom algorithms will be an important part of the trader toolbox.

The use cases do not stop there! Custom algorithms can also help solve more complex problems for advertisers who do have the benefit of online data. This could be a retail, automotive or travel brand who are seeing customers visit and quickly leave their website, thus optimising towards qualified traffic measured via Google Analytics would be invaluable to them. Alternatively, brands who have realised the value of an independent ad serving solution like Flashtalking are able to use custom algorithms to solve discrepancy issues between DSP and Adserver. The pressure of using guesswork and the constant optimisations are taken away from traders when using AI to focus on attribution and acquiring more customers when measured by a 3rd party provider.

To summarise; Whether you are an in-housed client or agency trading desk, it is worth Investigating AI driven custom algorithms - you never know, they may just produce some results to problems you never knew were possible to solve programmatically.

Best of all, with a SAAS solution it’s easy to activate - That is the beauty of AI driven custom algorithms done right!

Contact me on james@scibids for any further information or discuss pilot opportunities.