Your audience data holds the clues to dramatically improving ROAS (return on ad spend).
Being able to harness data into an AI Audience Model™ is a critical first step to discovering insights on an audience; these insights help improve ad creative and campaign results.
An AI Audience Model™ is a simulation of an actual audience since it’s based on 1st party data. It delivers insights into what drives an audience to engage.
But what happens when there are multiple products, differing locations, competitor focused campaigns, or campaigns related to special events?
Mixing all this data together might not provide the richest and most relevant insights needed to drive the targeted improvements for a campaign. This makes sense—a department store would not find highly relevant and helpful predictions for improving menswear conversions based on homewares data!
The problem for most marketers is 3-fold:
– It can be hard to segment data on Google and Facebook platforms to derive insights
– It can be time-consuming to have a data scientist collect, format, and generate queries to discover insights (as well as limited by time, cost, and human assumptions)
– AI analytics require expensive and time-consuming engineering and coding
Good News for a Data Focused Marketer
With Junction AI’s latest release, we tackled these issues head-on. We delivered the sector’s only user-driven audience model building tool, and we’ve made it easy and fast! Marketers simply point and click on the data to include in their audience model.
And the best part? It’s accessible to anyone! No data science needed, nor AI engineering or coding skills required.
What’s an Audience Model?
A Junction AI Audience Model™ is a model of a marketer’s 1st party data (more AI definitions marketers should know here). This data is generated from advertising and available on Google, Facebook, and other advertising platforms. These are the metrics (CPA, CTR, CPC) along with all the activity data, demographics, financial, and other targeting data.
The data is used to train and build an AI audience model. Utilizing the model, marketers can simulate the impact of ad-creative on their audience. This provides predictions on performance and insights to improve and optimize ad-creative.
With an audience model, marketers can surface critical insights for demand generation – which includes what an audience wants to see: products, features, and benefits. Insights generated from models can inform marketers on very specific concepts, for example, insights on the creative for a Facebook single image ad. Insights might identify including a sofa in the image to improve engagement, or changing copy, like using the word “boost” instead of “increase”. It can also provide insights on more general concepts like age of people, product color to feature, background image concepts, and the overall story flow of the ad.
Because the data used to train the AI is 1st party data it is highly relevant to a brand, as it is the actual data from the audience engaged by the ad. This makes the data the ultimate “simulator” to deliver that critical OBS (optimization before spend) return on investment.
Modeling with All-Data
When we first released our point-and-click AI audience modeling, we offered marketers the ability to generate their own audience model using all data from all campaigns on their Google and Facebook ad accounts. For many, it immediately provided rapid ability to quickly surface important relationships on their audience vis-a-vis the ad creative.
The AI identified insights representing new and useful concepts to incorporate into ad creative. Without the AI it could take marketers months of live experimentation, and lots of ad spend, to discover the same information.
This all-data approach was a great starting point. Having the ability to rapidly understand the audience at a global level helps define the critical next steps – segmenting data to generate more specific and targeted insights.
Segmenting an Audience Model
As marketers built all-data audience models, many wanted to test ad creative on audience segments. This meant giving marketers the power to select the campaigns, ad groups, and even specific ads to include or remove from the source data.
With our latest Junction AI release marketers can now specify the data. Our Audience Model wizard allows marketers to choose “all data” or select one or a mix of campaigns, or one or a mix of ad groups, to generate a segmented audience model.
The process works the same. Marketers define their model training based on generating more clicks or conversions and then select the data to include. Within only hours marketers have trained the AI on a segmented model so they can start surfacing critical insights on their choosen audience segment.
Options with a Segmented Model
Since rolling out our modeling options for both Facebook and Google Search ad data, brands and marketing agencies have taken up the option with gusto. Marketers are generating segmented models based on:
– location segmentation, for example, national (US vs Canada), regional (northeast US vs southwest US), or even city level
– product or service segmented models using data only on a single product
– product or service feature based segmented models, zeroing in on a specific feature or benefit of a product
– special event-based models for campaigns related to sales, conferences, or seasonal events like holidays, back to school, etc.
– campaign based models covering a reporting period to identify insights for reporting and campaign planning briefs
The possibilities are endless.
We recommend marketers take advantage of segmenting audience data. “All data” models provide valuable insights that are shared across audience(s), validating important customer value propositions. Segmented modeling enriches the analysis with rich and relevant insights. Together marketers are dramatically improving ad results, all while saving time and money on long discovery cycles and live experimentation.
Ask us for help! It only takes a couple of minutes to choose data, and within hours, a new audience model is ready for testing. This makes it possible to generate an audience model, test creative on our ad performance simulator and deploy to market highly optimized ads ready to perform at their peak, all on day 1.
Contact our customer success team for more information.