Looking back on the summer of 2018, it's hard to ignore the optimism that reigns in the air. Sunny weather? Check. Football triumph in England? Almost! AI as the next big thing in digital marketing? Try to count how many articles, blog posts, and sound clips you've come across in the past month that hype-tastically cite AI.
Now we are all for a little reasoned optimism, and there is no doubt that AI is an extremely powerful toolkit that will positively impact all kinds of socio-economic activities. But we're not so sure about the true value of AI in the context of digital marketing , and especially for international pay media.
Back to Basics
Cutting through the hype, let's start by looking at exactly whatt how AI and machine learning work in the context of international media compensation. For example, at the keyword level, how much and what kind of data is needed for the AI to make a good decision?
Well, Google's machine learning product Smart Auctions say they "allow you to tailor bids based on each user's context." Smart bidding includes important signals like device listings, location and remarketing for better automation and better performance ”.This implies that the signals required by the algorithm can be extracted from the sum of user behavior and that its "capabilities to learning quickly maximize the accuracy of your auction models to improve the way you optimitate long tail models [by evaluating] your campaign structure, landing pages , ad tex t, product information, keyword phrases and many others to identify more relevant similarities between auction items in order to effectively borrow lessons from them ”.
This suggests that the “go to” data source is our own campaign. But what are these patterns, how long does "fast " mean, and how can landing page data help manage auctions ?
Staying with auctions as an example, we think it works like this:
- Primary data: the algorithm goes back to the historical direct interactions with a keyword within a customer campaign and makes a cost / position decision based on pre-defined goals such as ROI or CTR , and enough data.
- One way to resolve a possible data volume problem would be to look back way. But that would ignore seasonality, promotions, and changes in consumer behavior over time.
- Secondary data - the algorithm does not have sufficient data to make a `` good '' decision on the primary basis, therefore uses corroborative data (performance indicators of others campaigns that have similar characteristics (eg, same vertical, same language) to make decisions.
Do we even have enough data?
The question is whether, aside from very high volume campaigns of large categories (think auto insurance, credit cards), there is enough primary data to fuel a effective decision-making in AI. AI needs a huge amount of data to be effective.IBM's ep Blue learned from failures, for example, the developer relied on 5 million datasets. Most industry experts believe AI's biggest limitation will be accessing high-quality data at a sufficient scale.
We have no idea what a 'good "data volume is. This is even more unlikely for international PPC, where campaigns are often very granular, multilingual and designed to include many long tail keywords (which by definition don't have a lot of volume).
When it comes to secondary data, in How far is the corroborative data? For maximum relevance, taking CLIENT X as an example, we should assume that the algorithm quickly assimilates data from CLIENT X's direct competitors and uses it to better inform the auction management strategy .
Surely this type of cross data would feedthe bidding tactics of all auction players, creating a loop where no player has an advantage?
Resources
Analytics The Ultimate Guide to Forum Link Building in 2020
Analytics Top 5 SEO Mistakes
Development Improving SEO & Developer relationships
Analytics How to do an Online Competitor Analysis
If competitor data is not used, so whattype of secondary data is relevant enough to make good AI decisions. It would be easier if we definitely knew how the rules of the algorithms were built, but of course we never will.
It's time to verify reality
To recap, if we knew that 10, 100 or even 1000 interactions were enough to deliver superior efficiency through the 'IA, we would love to. Campaigns could be planned and executed to use the optimal combination of AI and human capabilities, with the best results for ad platforms, agencies and clients. AI could focus on brand and category level interactions, with human oversight and detailed long tail management.
It seems unlikely that adequate transparency about how AI actually works, the amount of data is needed, asnt the `` rules '' work k, will be forthcoming unless significant changes in business models or practices occur.
Instead, AI is optimistically overstated as the next big thing in digital while blithely ignoring basic AI premises and current practicalities both nationally and nationally. international paid digital media
Find out more about:
Aucun commentaire:
Enregistrer un commentaire