This week’s chat featured guest Ginny Marvin of Google Ads and we asked her a ton of questions after Google Marketing Live the previous week.
There were a couple of questions asked where chatters answered:
Q1: What was the most important or exciting thing announced at Google Marketing Live (GML) last week in your opinion?
Q2: What was the most concerning or disappointing thing announced at Google Marketing Live (GML) last week in your opinion
Q3: We have had a bunch of questions submitted for Ginny already, but what are your burning questions after GML last week?
Read the full chat recap here.
Listen to the podcast reap here.
Listen to a recording of the audio version of the chat with guest Ginny Marvin:
Questions for Ginny
Q: What is an agile budget? It was talked about by Jerry Dischler in his segment and then never mentioned again.
A: Agile budgeting offers flexibility to respond to market & business dynamics versus fixed budgeting based on a set timeline, specific channel etc. Jerry cited agile budgeting across sales, marketing, operations as an example of how businesses are using AI to optimize results. It requires close coordination w/ marketing & finance. In the example Jerry noted, the CMO & CFO worked in tight partnership on using agile budgeting to respond to franchisee needs & fluctuating demand in real time. You can read a bit more about agile budgeting here and here.
Agile budgeting enabled by AI is also the idea behind cross-channel campaign types. In PMax, for example, rather than allocating to a specific channel, your budget is applied wherever there is demonstrated customer demand to find incremental conversions/conversion value at your target.
When Jerry said the LLM was “tuned from Ads data” what does this mean? Is the Google Ads AI training on all advertisers content and then using that to make recommendations? Is my client’s ad copy training the AI/LLM to build similar ads for competitors?
A: Generative AI helps create new assets based on your own inputs, such as website content, existing ads and your prompts in the conversational experience. Advertiser data is used in the LLMs in a highly aggregated way only. Automatically created assets (ACA) and conversational experience are both opt-in features.
Q: When will these features be available in Europe?
A: Automatically Created Assets (ACA), for example, is available in Europe, and conversational experience in Google Ads testing will begin globally in July, in English to start. We’re working hard on language expansion to ensure each language meets a high bar for quality. We’ll be testing and learning – and will be looking for advertiser feedback.
Q: Why is Google Ads support so shockingly bad and are there any plans to make it even worse by using Bard powered chatbots?
A: I get the sentiment, but will also say that now being here, I also have more empathy for the fact that great support at scale is really hard. We’re making active investments in several areas to help customers get answers & resolutions to run effective campaigns. We’ve also seen that the fantastic marketers in the Google Ads Help Community can help answer many questions & challenges often better than our support staff who don’t actively manage campaigns. We believe AI can play a role as well.
Q: Will Google Ads deploy Bard as an avenue for disapproval resolution? It’d be pretty neat if it could escalate on our behalves!
A: Thanks for the suggestion, I’ll pass it along!
Q: I’d like to learn more about “Demand Gen” to see if there might be some opportunity there for SMBs and / or those of us generating leads for small businesses.
A: We will have more to share on this in the coming weeks, stay tuned.
Q: If my competitors and I are all using generative AI to create our creatives, how does Google ensure we won’t get similar copies?
A: All of the AI-generated assets are grounded in your inputs like your website and what you input in the conversational experience. The more unique details you use to guide and inform the AI, the more likely the generated ad assets will be unique.
Further, you’ll be able to review and revise suggested assets created in the conversational experience & new asset creation flow for PMax to ensure they reflect your brand voice. Advertisers can view reporting on ACA and remove assets they no longer want to serve.
Q: Any product updates that would be beneficial to B2B marketers? The content was very retail/D2C focused.
A: I know many of the case studies were retail focused but we see these new tools that use GenAI working for all types of marketers, B2B included. We will be continuing to experiment and take in feedback from B2B marketers to help ensure they provide value.
Some examples include:
- Conversational experience in Google Ads UI: This will start testing in July (it will be opt in)
- New PMax asset & image creation workflow: Enter your URL to create customized, relevant assets. It’ll also be able to suggest images created just for your business, and you can ask it to generate more images based on what your vision and brand..
- More data integrations are coming to make it easier for businesses of all sizes to connect their consented first-party data sources to Google in a privacy-centric way. More to come on this soon.
- And there are video capabilities to make video assets accessible to more businesses with the help of AI. It can take your existing image and text assets as well as your product feed or app store listings, if applicable, to automatically create video ads in multiple orientations, including vertical for YouTube Shorts.
Q: It wasn’t always clear to me if/when we were seeing conceptual demos vs real technical demos. The Product Studio demo for example — very tidy. Is that where the product is or where the product wants to be?
A: When Product Studio launches later this year, it will have three key features as outlined in the demo to help you generate scenes, increase image resolution and edit backgrounds.
Q: Are there mechanisms in place to prevent my autogenerated ad copy from using trademarked or “similar to previously used” copy from my competitors?
A: We have long-standing policies that apply to all ads and advertisers, regardless of how that ad was created. When assets are created by ACA using generative AI, we have guardrails in place, including grounding content to advertisers’ offering – primarily using the LP as source of truth. And all assets go through the same policy checks as advertiser created assets. This is still early days, especially when it comes to image generation, and we intend to engage broadly across the ecosystem as we build out these capabilities and experiment.
Generative AI helps create new assets based on your own inputs, such as website content, existing ads and your prompts in the conversational experience. Advertiser data is used in the LLMs in a highly aggregated way only. The more unique details you use to guide and inform the AI, the more likely the generated ad assets will be unique. Note that both ACA & conversational experience in Google Ads are opt-in features.
Q: Will advertisers still be able to use feeds once GMC Next rolls out? Many advertisers don’t put all their data in their stores/sites due to scale.
A: When you switch to Merchant Center Next all existing setups & settings will be retained, feed uploads and ad campaigns will keep running. Merchants can still use feeds, and multi-feed support is coming soon. You can expect more features & improvements to come.
Q: Will they be able to share insights on how to identify when an ad appears in a SERP with a conversational AI and when one appears without one?
A: Ads in Search Generative Experience (SGE) will trigger using the same principles that Search & Shopping ads do today: considering whether users will find them useful and relevant. Metrics will be reported as usual during the experiment. More here.
Q: How will queries be reported in SGE? Would all search terms involved in the SGE count as a separate impression and how would a clicks conversions be attributed across the different queries within a single Search Generative Experience?
A: The search queries throughout the journey in SGE will appear in the Search Terms Report, as long as they meet the current privacy thresholds. Some more details here.
Q: With Google presumably seeing Generative AI as the next step for Google Ads. What training will Google Reps be given to provide value to their accounts and agencies other than telling people to use Generative AI for “OptiScore reasons”?
A: Best practices reviews and refreshes as well as Internal training are always ongoing. A quick note about Optimization score and why it’s often a starting point of reference for reps: OptiScore estimates how well your account is set to perform, and we see a *correlation* between higher optimization scores and conversions. Advertisers that increased their account-level Optiscores by 10 points saw a median 14% lift in conversions.
Applying relevant recommendations or dismissing recommendations that aren’t relevant to your campaign will change the overall optimization score of your account. It’s a tool for understanding potential optimization opportunities but it doesn’t directly impact performance. More on optiscore here, including how it’s calculated and applying and dismissing recommendations here.
Q: How are ads going to be integrated into the new conversational AI SERP, if at all?
A: At I/O we showed how Shopping ads can appear in SGE and at GML we showed how text ads can show in the experiment. Text and Shopping ads can show above generative results in a new mode in Search for conversations when users’ ask follow up questions. In SGE, context is carried over from question to question. Both organic results and ads will reflect the context of the conversation as it evolves. Ads will always be clearly labeled as “Sponsored”. If it’s not labeled it’s not an ad.
You can sign up for Search Labs at g.co/Labs
Q: Weren’t the alleged PMax reporting improvements/added transparency already there? The ranges for search provided have always been very broad…
A: There were several new updates to Insights announced. More functionality is coming to Search terms insights, including the ability to download reporting, set custom date ranges and access reporting via the Ads API. And more Asset insights are coming to help you understand what’s resonating with certain audiences & more. Lastly, we announced the Experiments beta coming later this year to A/B test bid strategies.
Q: What does the future of agency support look like? Getting rid of agency reps for Google premier partners and just assigning reps for each client is REALLY unwieldy for those of us with a lot of clients – and doesn’t help Google win new biz.
A: Thanks for the feedback. We’re working to improve this experience.
Q: Where can we see details about “privacy” as it specifically relates to masked search queries?
At times, it feels like they slip away unannounced. We advertisers understand (+ have to accept) it conceptually. Pointing to [HERE] for customers would help.
A: In order to protect user anonymity, we do not provide search term data when there’s very limited search volume. I understand this is frustrating, but the privacy and regulatory landscape has evolved significantly, and we use the same threshold regardless of query to ensure we protect user anonymity. You can point to this Help Center article which notes that search terms are reported when “a significant number” of people have used them in a couple of places.
We have also worked to enable more searches such as those with 0 clicks to appear in SQR – again, only when there’s significant overall search volume – in 2021. The groupings in search terms Insights can provide more visibility into what’s working. And as announced at GML, more capabilities, including downloading and custom date ranges are coming to search terms Insights soon.
Q: Will we be able to set custom compare date ranges in search terms Insights as well?
A: Yes, you’ll be able to set custom comparison date ranges, as long as that historical data is available.
Q: GA4 – anything advertisers should know about it relative to G Ads?
A: We announced at GML that you’ll be able to build audiences from GA4 directly in your Google Ads accounts, including predictive audiences.
A few highlighted tweets
— AdsLiaison (@adsliaison) May 30, 2023
— Julie F Bacchini (@NeptuneMoon) May 30, 2023
Q3. Once Google auto pulls SKU data into GMC Next, are we still able to modify/change data via supplemental feeds, rules and other tools?@adsliaison
— Duane Brown 🇨🇦🏳️🌈 4x Vax’ed! (@duanebrown) May 30, 2023
There was this question too – it was confusing during the preso. Advertisers want to drop brand from PMax campaigns, not a tool to turn broad match into brand campaigns. And what’s to stop an advertiser from making sure they show up on competitor brands? #ppcchat https://t.co/oX1GYX4QS6
— Melissa L Mackey (@beyondthepaid) May 30, 2023
All of the AI-generated is grounded in your inputs like your website & what you input in the conversational experience.
The more unique details you use to guide and inform the AI, the more likely the generated ad assets will be unique. #ppcchat
— AdsLiaison (@adsliaison) May 30, 2023