The race to AI is well and truly underway, in-fact if we all had £1 for every time we had heard “AI” just this year so far, we’d probably all be able to retire already. When bombarded with so much information, it can be difficult to know where to start on this AI journey, or even what does / doesn’t contain AI.
Highlighting this, 90% of respondents to a McKinsey survey of business leaders from a variety of verticals said they think their organisations should be using AI / Machine Learning “often” or “almost always”. This pales in comparison to the 35% who think they are currently using these technologies to the same extent.
The reasons for the difference can vary from business to business, individual to individual. Let’s consider where to start and if you’re already on this journey, understand where to go next.
Embedded AI – What is it?
Embedded AI is AI that exists as a feature or part of an existing product. An example could be something as simple as Insights within Google Analytics 4 (GA 4), which leverages modelling on the data collected by GA 4 to spot any underlying trends or changes which could highlight either an issue, or a success, for further analysis. Alternatively, it could be as impactful as Performance Max (PMax), an incredibly powerful campaign-type in ads that can reach across all of Google’s channels to optimise performance.
The challenges that embedded AI seeks to solve are;
Ensuring Quality Data
Through platforms like GA 4 and tools such as Consent Mode, Enhanced Conversions, Customer Match and more, embedded AI ensures marketers and analysts have not only a more visible dataset, but one which is more accurate as well.
Scalability
With AI features built-in to the platforms, many require no or minimal effort to enable and setup. Additionally, they can be quickly applied to live or planned campaigns, or current analysis, allowing businesses to reap the benefits quickly.
Optimising
Lastly, with higher quality data and more datapoints through scale, your brand can better understand performance: amplifying what works, and discarding what doesn’t, and providing you with better overall analysis and performance.
So, where to begin?
Foundations
Simply start at the beginning. By having a robust foundation in-place, platforms and features used later down the line will reap the benefits of more observable data, whilst remaining privacy safe.
An increasing trend we have seen over the last few years is the growing amount of modelled data within platforms. This can create issues, as modelling is only as accurate as the data entered. If the input to a model is skewed towards users who convert, then the modelling will likely assume that more people are converting than actually are. If this modelled data is then used in marketing for audiences, it reduces the efficiency of Ad Spend, with either more spent on users who won’t convert, or on a campaign which isn’t performing as well as the insights show - ultimately harming the overall performance.
Thankfully, there is a growing number of tools and features to help combat this, and as more tools leverage first-party data, it can ensure that the models perform accurately and can be of significant benefit. These are features such as:
Google Analytics 4 + Consent Mode
Enables marketers to analyse site and campaign performance in a privacy safe way, enabling maximum visibility, while respecting users’ privacy
Enhanced Conversions
Capture hashed customer data from conversion pages, to match against Google logged-in data, increasing the observable data and overall quality of conversion modelling.
Data- Driven Attribution
Leverage AI models to better understand marketing performance, and how channels can support each other to drive conversions.
Customer Match
Create a customer list with contact information, that can be used within Ads to target a customer match segment. Then, whenever those users are logged-in to their Google account, they can see your ads.
With these Foundations in place, you will likely already begin to see large increases already such as:
- Recovering 70% on average of ad-click-to-conversion journeys through Consent Mode
- 17% Increase in conversions from Enhanced Conversions
Expansion
Once you have your foundations in-place, it’s time to scale and optimise. There are over 20 Embedded AI features within GMP. To try and help navigate the array of features available, we’ve broken all of them down into three key areas; Modelling, Reach and Acquisition. We’ve listed the tools in each section, though for brevity we’ll keep the detail light.
Modelling
What has already happened? What is happening right now? What will happen next? These are the specific questions which modelling seeks to answer from your data.
We’ll start with a big one: Predictive Audiences. Within Google Analytics 4 is an exceptionally powerful modelling feature which I cannot recommend enough. Often, it’s not used when reviewing GA accounts, yet it has some of the largest documented benefits to media campaigns and insights. If you don’t have this enabled, I would strongly recommend you to begin using this, and the best part is that as long as you meet the minimum requirements (1k users who convert, and 1k who do not per month for purchase audiences, or 1k users who view and 1k users who churn for churn audiences), predictive audiences can be set up within a few minutes.
These predictive audiences can be used in custom audiences too, expanding their usefulness even further. On average, we have seen increases in conversions of 734% across a range of case studies, in addition to a 75% reduction in CPA. *Internal data from a collection of market and internal case studies
Reach
Reach is all about how to place your message in -front of potential customers, at the right time, at the right place and with the right message. To help with this, there are a range of features that can help tackle the challenge this poses either broadly, or for specific platforms. These include:;
- Audience Expansion
- Efficient Reach
- Target Frequency
- Video Reach Campaigns (VRC)
- Video Action Campaigns (VAC)
Additionally, we have the Broad Match feature. This utilises AI to match ads to relevant searches based on what users are searching for. It typically leads to an average increase in conversions of 35% in campaigns that use a target of cost per acquisition (CPA) (Your marketing, multiplied by Google AI - Think with Google).
Acquisition
Acquisition is all about turning those users who view your Ads into conversions. Similarly to the tools within Reach, they can change and adapt creative, uncover audiences and in some cases, extend across all of Google’s advertising space. These give marketers a plethora of options to shape how campaigns are ran and optimised, including tools such as:
- Automated Bidding
- Custom Bidding
- Smart Bidding
- Performance Max (PMax)
- Responsive Search Ads (RSAs)
- Value-Based Bidding
Where do I begin?
Every business is unique, and not every feature listed here may be useful. For that reason, we have been working to build a calculator. With a few details from you about your marketing metrics, our calculator can provide a guide on the kind of performance benefits these features could unlock. Through scanning all available case studies both internally and externally, we can provide a baseline of anticipated performance.
To be transparent, some of these features are new, and some are only used in-combination with others. For this reason, we know that only in certain instances will you see a 300% uplift in conversions. Therefore, we provide weightings to scale uplifts to real-world scenarios.
If you have any questions, please reach out to here where one of our experts would be more than happy to discuss how these features can be of benefit to you and your business.