Skip to Content

Hello Demand Generation Campaigns

August 21, 2023

Let’s talk about an important, albeit under-the-radar change to the Google Ads world: the launch of the Demand Generation campaigns (and the accompanying sunset of Discovery campaigns). 

Discovery campaigns (the new name for Gmail + Google Discover ads) have experienced relatively low adoption rates and priority within accounts since they were launched several years ago – despite some incredibly strong performance metrics (at least in accounts I’ve audited). Candidly, I’m not sure why these campaigns never took off – perhaps it was the relatively limited inventory, or the fact that most brands preferred to email potential customers/clients directly, vs. showing ads in the promotions/social tabs of gmail. 

The advantage Discovery campaigns brought to the table was simple: nothing else enabled advertisers to reach potential customers/clients earlier in the purchase journey, across gmail, YouTube (home + Watch Next) & discover properties, using Google’s 1P data (which isn’t available on all GDN sites) and without having to worry about managing an ever-increasing number of hot-trash third-party sites. For most of our clients, Discovery was consistently a top-performing ancillary campaign type, effectively augmenting our YouTube, Shopping & Paid Search, but never eclipsing them. 

Demand Generation (“DG”) aims to change that. 

DG combines everything from Discovery campaigns (gmail, Discover, YouTube: Watch Next & YouTube: Home) with YouTube Shorts (averaging ~50B daily views) and YouTube: In-Feed into a single campaign type. Basically: it’s all the non-search goodstuff bundled together.

After spending quite a bit of time in the DG Beta, there are six reasons I’m bullish on this revamped campaign type – and why I think it’s worth testing now: 

1. The Platforms:

YouTube is one of the world’s most popular and powerful video platforms – with ~3B MAUs (more than Facebook) and over 1B hours of video watched every day; Gmail is the world’s most used email platform, and Discover is a robust news, entertainment & shopping awareness engine. And, unlike other campaign types, DG does not include third-party sites on the Display Network, nor does it include search (which has always been a weird fit in PMAX). 

Instead, DG combines the three platforms that represent the most significant threat to social advertising Google has ever assembled – and that’s before we dive into some of the other features. And Google isn’t hiding their ambition – the very first sentence on the “About Demand Generation” page is the sentence, “….Demand Generation campaigns are ideal for social advertisers who want to serve visually-appealing, multi-format ads on Google’s most impactful surfaces available to any advertiser.” Yeah, it’s clear Google is coming for Social Ad budgets. And I think they might get them.

2. A True Creative Hub

One of my favorite (albeit underutilized) features of Meta Ads is the Creative Hub, which provides advertisers ability to preview and customize creative to each placement. The inability (to this point) to do this for Google’s dynamically-generated assets has been a major pain point for many clients – especially those with rich visuals and/or strict brand guidelines. There’s nothing worse than giving Google beautiful assets, only for it to butcher them into Frankenstein’s monster – then hear about it from the brand or compliance team. 

Demand Generation actually addresses this, with a robust ad preview and customization tool that allows you to visualize how your assets (images, videos, product feeds, text) will appear on each DG placement, before actually running the ads. There’s also the ability to tweak those assets + placements to be just right.

3. Hello, Lookalike Segments

I was among the many advertisers who were annoyed when Google sunset Similar Audiences (announced in November 2022; occurring now); but they’re back and (believe it or not) better than ever in DG campaigns. 

The revamped Lookalike Segments function quite comparably to Meta Ads, with a few key differences:  

  • All audiences are based on business + Google Data – customer lists (including your offline data imported into Google Ads!), website activity, app activity, and YouTube engagement. Pretty much everything you’d want, so win.
  • Some ability to toggle your Lookalike Segment Size – unlike Meta Ads, which allows you to actually select a specific percentage for your LAL, Google will offer three choices: (a) narrow = 2.5% LAL; (b) balanced = 5.0% LAL; and (c) broad = 10.0% LAL.
  • You do have the option to exclude your first-party data sources from Lookalike Segments, so you’re actually advertising to net-new customers (assuming you keep your seed lists up-to-date)
  • Because Lookalike Segments are a segment type, reporting on them is excellent: advertisers are able to see performance for their LALs vs. Google-defined Audiences vs. Custom Segments, all within the Audience Insights reporting view. That’s a HUGE win. 
  • Another HUGE difference (that really should be talked about more) is the ability to generate your Lookalike Segment from up to 10 seed lists. While this may not seem like much, if you’re a brand with multiple conversion types (for instance, website sale, store sale + app purchase), or conversion paths (demo request or free trial or lead form), this can allow you to reach the audience size minimum with much more current data. This is one of those things that has flown under the radar, but is massive in terms of its potential to unlock LALs for smaller advertisers. 

That said, it isn’t all sunshine and rainbows. At present, there are some limitations with Segments that I hope will be removed as DG campaigns roll out for all advertisers: 

  • Segments are only eligible to be used for Demand Generation campaigns; while you can view them in Audience Manager, they can’t be added to other campaigns, even for observation. The other major bummer here is that Lookalike Segments can not be edited in Audience Manager – which results in a clunky user experience for advertisers. Hopefully this gets fixed ASAP.
  • There is a substantial delay in populating your Lookalike Segments – up to 96 hours from the time the audience is created to the time it is eligible to serve. That’s much, much longer than Meta – so if you want to use Lookalike Segments, I’d highly recommend creating them about a week in advance of when you plan to run ads, to ensure they are able to deliver right away. 
  • I have not seen value rules included among the targeting / steering options yet, though it’s possible these will come along later. 

Despite these limitations, DG is a major step forward in the right direction. Where PMAX uses audience signals (no hard targeting), Demand Generation provides advertisers with actual control over who sees their ads. And while I did not mention it above, DG also includes access to all existing Google Ads Audiences + Segments, including both Google-defined Segments (in-market + affinity), along with your custom segments.

4. Creative Testing

Another sorely-needed upgrade to Google Ads is an actual creative testing mechanic that works across all inventories AND tests multiple variations of image and video assets. And…that’s exactly what Demand Gen A/B Experiments do. 

DG A/B Experiments function similarly to Google’s existing Experiments, with a similar setup process (select your dates/run time, create your experimental arms, and specify how much budget/traffic each arm should receive), and a similar reporting interface. The big difference here is that Demand Gen A/B Tests are designed exclusively for Creative Testing – they don’t work with targeting, bid or budget (at least, right now). 

A few other things to keep in mind as you begin your creative testing journey:

  • As with other experiments, Google needs ~50 optimization actions per experimental arm (so, 100 total optimization actions for an experiment with 1 control + 1 experimental variant, or 2 total arms) to surface statistically-significant results. 
  • To achieve this, Google tends to recommend that you optimize toward a higher-in-the-maze action – like add to cart or video views. This is a terrible idea that you should never, ever, ever do. Optimize for the thing that drives value to your business, or don’t run the experiment at all. It does you no good to figure out which ad is better at driving people to add things to a cart, if your ultimate goal is to drive actual, completed transactions. If you don’t, you’re going to end up with an invalid result, anyway – so what was the point?
  • One common thing I’ve seen is A/B experiments yielding inconclusive results. When this happens, it’s a great thing: it means you’re not taking big enough swings with your experimental group. If you’re wondering what that means (or how to sell it to your client/boss), check out the 10% & 10x philosophy article here
5. Additonal Bidding Options:

Demand Gen campaigns include Google’s standard, ML-powered bidding strategies (Max Conversions, tCPA, Max Conversion Value, tROAS), along with a Maximize Clicks option. I’m not a huge fan of max clicks, but for some advertisers + use cases, it’s a welcome change.

6. Reporting Upgrade:

The sixth & final feature I want to highlight for DG campaigns is the inclusion of Search Lift, Brand Lift & Conversion Lift measurements, alongside detailed insights & reporting for all individual assets & creative formats, plus product-level reporting for all Google Merchant Center Feeds. I can’t stress enough how awesome this is after spending 6+ months trying to find anything useful from PMAX. This is a welcome, welcome reversion from Google back toward transparency + actual, helpful data

7. Bonus: YouTube is about to get a lot more valuable:

I’m particularly excited about Demand Gen campaigns rolling out now for one reason: Football is back. And YouTube is the new home for Sunday Ticket – which means millions more YouTube TV subscribers (the bad version from the people-who-shall-not-be-named had ~1.5M; I wouldn’t be shocked to see YouTube double that in the first year). YouTube is about to get a LOT more valuable – and Demand Gen is one way advertisers can capitalize on it. I still believe there’s more untapped value for advertisers in YouTube-only campaigns, but DG is a welcome addition to our marketing mix – and yes, CTV is part of Demand Gen (you can see the creative specs included in Google’s documentation here).

What it all means:

The other thing I wanted to highlight is what this new campaign type means in the context of Google’s overarching plans moving forward. In my view, this represents a shift from “automate everything” to “give advertisers control”. 

It would have been easy for Google to sunset discovery + roll everything into PMAX. But Google didn’t. They actually created a new campaign type that includes all non-search, Google-owned inventory (something preferable to Meta’s Advantage+). As someone who has been highly critical of Google’s (seemingly) inexorable march toward a “link up your bank account and press go” advertising platform, this feels like a welcome shift back in the direction of giving advertisers control over their investments. 

What’s Next:

For all its benefits, DG isn’t perfect. I think it will improve over the next 3-6 months, as advertisers are able to find bugs + offer feedback. And there are promising indicators that other, much-needed features (like the ability to add Lookalike Segments to other campaigns, and to use Product Feeds from GMC with Carousel Ads) could be coming soon.

It’s clear that Demand Generation was designed to give advertisers an alternative to Meta. And based on what I’ve seen so far, Google has come very close to hitting that mark. I’d highly recommend anyone who is relying heavily on Meta (think $50k+ per month in Meta spend) apply for the Beta ASAP, and include DG as part of your BFCM strategy.

What to do:

Demand Gen campaigns are currently in beta, so they are not available for all advertisers. If you’re interested in signing up for the Beta, you can apply here.

If you have a dedicated Google Account Team, your existing Discovery campaigns will be upgraded / migrated between October 2023 and January 2024. For all other advertisers, your existing Discovery campaigns will be automatically upgraded sometime between January – April, 2024. Demand Generation campaigns should be available as a new campaign creation option sometime in Q4 (likely in October – but obviously that’s subject to change).

I’d honestly recommend any advertiser apply for the Demand Generation beta ASAP. There’s no downside at this point – it gives you better targeting, more platforms + more control, while also helping you avoid additional volatility in an already-chaotic Q4.

That’s it for this week – I’m off to go play around with Demand Gen campaigns a little more, then enjoy some time outside.