Revolutionizing Paid Media: Harnessing the Power of AI and Bidding Optimization for Advertising Success

We recently sat down with a panel of paid media experts from some of Richmond's most trusted digital companies. Our panelists spoke about their strategies incorporating artificial intelligence and bidding optimization to help maximize the output of their paid media campaigns.

Among the panelists was Morgan Jarvis, Workshop Digital Director of Paid Media. She shared how she sees these cutting-edge techniques increasing engagement, boosting ROI, and creating a more efficient marketing ecosystem. Here are her answers to our questions about AI and paid media.

Note: We’re using AI and Machine Learning (ML) more or less interchangeably and recognize it’s not always accurate.

What was your "aha!" moment when you realized the potential of AI/ML and bidding optimization in the digital marketing world?

For me, it was the combination of AI bidding and closed-loop data. When you can tell the system not only what conversions to optimize towards, but also what conversions turned into new customers, you can eliminate a lot of lead quality issues.

Leveraging closed-loop data tells the system to not just drive towards every conversion, but to optimize towards the conversions and audiences that ultimately turned into customers and revenue.

There are several ways to accomplish this, but one great automated method is integrating your Salesforce data directly with Google Ads. One of my colleagues recently wrote a step-by-step guide for this on SEL: How to set up an offline conversion import from Salesforce into Google Ads

How has the integration of AI/ML and bidding optimization transformed the digital marketing landscape in recent years, and what is the most significant change you've observed?

The move from manual bidding to automated bidding was a huge shift in the day to day work of the average paid media account manager. Over the years, we’ve seen AI continue to learn and improve upon itself to the point where automated, conversion-based bidding methods are the best choice in almost every situation.

The biggest change I have observed is in the ability for the systems to optimize towards your goals on smaller data sets. We used to need to start with manual, maybe move to max clicks, and then move to a conversion-based model once we had enough conversion data. More and more we are seeing that these minimum data limits have been reduced or removed which allows us to use these tools for more S/M businesses as well.

How do you measure the success of your AI-driven bidding optimization strategies, and what key performance indicators (KPIs) do you find most valuable?

Aligning your bidding strategy with your KPI is the best way to ensure success. If you’re focused on efficiency, having a cost-per-acquisition cap is the best way to ensure you operate within that goal. If your goal is to maximize conversion volume, using a max conversions method is often your best bet. If you need to hit a certain ROAS - target ROAS. You see where I am going with this.

Can you share a specific example of how AI-driven bidding optimization has improved the performance of your paid media campaigns?

There are many, but one, in particular, would be in having the AI optimize towards converted customers.

I had a client that was performing great from our standpoint. Conversions were up, CPL was down, but they were not seeing that translated into an increase in customers and revenue. What that told us is that we had a problem with lead quality. In order to improve quality, we determined that we needed to give the system more data. To do that, we began uploading “offline conversion data” for won customers, and optimizing towards that. That allowed us to optimize towards high quality conversions and new customers, rather than just lead volume.

What are the most critical questions to consider when implementing AI and bidding optimization into your marketing strategy?

What are your client’s KPIs? Choose bidding methods appropriately based on those needs, keeping in mind what has traditionally worked to attain those types of goals.

Do you have enough data to optimize towards conversions? If not, can you restructure your account in a way that allows for thicker data sets (ie, shared budgets, combining campaigns, etc.)?

Audience identification & expansion with AI - what inputs have you found work the best for AI-driven audience creation?

We’ve seen good results with Dynamic Search Ads, but we are still experimenting to see where else we can use this technology to get even better results for clients.

Bidding Optimization with AI - How have you tapped into the strengths of bidding algorithms and learned to trust the “black box”?

Through trial and error! Step your way into it with campaigns that have larger data sets, learn what works best in each situation and for each KPI goal. It’s also important to not trust the “black box” blindly. Maintaining a healthy curiosity of what is behind the results you are seeing can ensure that you are in a position to give the AI the appropriate guardrails.

How can smaller businesses (e.g. those with less data) or those with limited budgets effectively leverage AI and bidding optimization to compete with larger brands? Are there examples where AI/ML is *not* the best solution?

Combining campaigns for thicker data sets is one tactic that can work. If that doesn’t make sense from an organizational standpoint, shared bidding and shared budget methods are a great solution.

There are always one-off situations where AI may not work the best for your business. An example is a brand campaign where you want to maintain a very low cost per click.

• Since some of the conversions from brand campaigns could be captured through the organic listings, maintaining a low cost for Brand is important to ensure that the incremental volume is worthwhile.

• Alternatively, if you have heavy competition on your brand terms, manually bidding to ensure you stay at the top of the page may be more important than increasing clicks or conversions from brand.

What challenges have you faced when incorporating AI into your digital marketing efforts, and how have you overcome them?

Small data sets can be an issue. Shared bidding and shared budgets are great workarounds for this.

Automated bidding methods have a tendency to optimize for volume versus quality. Feeding back closed-loop data on qualified leads and or new customers is a great way to train the AI to go towards both quality and volume.

How do you think AI and bidding optimization will continue to evolve in the coming years, and what new opportunities might this create for digital marketers?

I truly think that the future of digital marketing is in closed loop, 1st party data, especially as more conversion data will be sampled with increased privacy requirements. I’m sure AI can have a role in that process, with the support of new, easier to implement processes to automate the integration of 1st party data from various CRMs directly into your advertising platforms.

How important is it for digital marketers to stay up-to-date with the latest advancements in AI/ML technology, and what resources do you recommend for ongoing education?

It’s always very important to know how new technology might benefit your company, and newsletters and blogs are a great tool to stay informed on how those advancements are evolving. We have a newsletter at Workshop Digital that often speaks to upcoming and emerging trends. I’m also a fan of the SEL blog.

Can you discuss any ethical considerations or potential pitfalls that digital marketers should be aware of when using AI and bidding optimization in their campaigns?

In our privacy-first world, the ability to get accurate conversion counts with the removal of cookies on the horizon will be more difficult. Also, when using closed-loop 1st party data, particularly through manual uploads, it’s very important to make sure that you are not retaining any sensitive information.

When using the Facebook pixel and conversions API, it’s extremely important to make sure that this is in compliance. For example, we have moved to a policy of not using the Facebook pixel for any healthcare clients due to the associated risks with collecting data that we should ultimately not be collecting.

Can you share an unexpected or humorous experience you've had while implementing AI and bidding optimization in your marketing campaigns?

If you optimize towards conversions without using conversion values, you have to make sure you are only driving towards the most important conversions.

I’ve seen situations where we were tracking both form submissions and email newsletter signups, for example. If there is a higher volume of newsletter signups, the system can start to push towards those rather than the conversions that have a larger impact on the bottom line. You don’t want to turn off max conversions and end up maximizing something that isn’t super important to the client, such as a micro-conversion.

Lightning Round - “Hot Bot” or “Bad Robot”?

Google Ads Performance Max

Both! I’ve seen PMAX do well, but have also seen it lead to situations with undesirable results due to the lack of control.

There are some theories that PMAX is increasing costs in brand campaigns. And because it’s not a true search campaign, there are also theories that these searches may not show up in the auction insights reports. This can cause some unexpected issues such as rises in brand campaigns due to competitors using PMAX!

ChatGPT

BOTH! It can write code, generate ad copy ideas, and point you to the right excel formulas which are all great. It can also tell you blatantly false information with a high degree of confidence – not so great.

It will be interesting to see how Bing’s GPT search interface affects metrics in Bing Search Campaigns over time

Google’s Bard

Same issues as ChatGPT. I personally think Bard is a bit behind ChatGPT in terms of maturity.

AI-enabled Chatbots

SnapChat has one now?! It’s the bandwagon everyone is hopping on. It can be great, but the answers can’t be fully trusted, so you have to consider each specific use case.

Predictive Audiences (e.g. GA4 “likely to purchase” audiences)

This is one that will likely grow in sophistication over time. We’re keeping an eye on how this can inform strategies and tactics.

Don’t miss your chance to learn from some of the brightest digital marketers in Richmond and beyond at the next DigitalRVA event.