Setting up a Demand Gen campaign takes about 20 minutes. Getting it to actually work takes weeks of testing.
The setup itself is straightforward. The hard part is what comes after: finding the right audience and the right ad format for your products. Those two things determine whether Demand Gen makes you money or wastes your budget.
This guide walks through the technical setup step by step, then covers the testing strategy that actually matters.
Prerequisites: Get These Right Before Starting
efore you create your first Demand Gen campaign, make sure you have four things ready.
- Conversion tracking. You need purchase tracking set up in Google Ads. Either GA4 linked to Google Ads with purchase events, or Google Ads conversion tracking directly. If this isn’t working, fix it first. Nothing else matters without it.
- Product feed. Your product catalog needs to be in Google Merchant Center with current prices, images, and availability. If you’re already running Shopping ads, you’re good.
- Creative assets. At minimum: 3-5 images (landscape 1200×628, square 1080×1080) and 2-3 videos (15-60 seconds, vertical 9:16 preferred). More variety gives the algorithm more to test.
- Budget. At least $100 per day, sustainable for 4-6 weeks. That’s $2,800+ per month committed to this campaign. If you can’t do that, wait.
Know your target ROAS or CPA before you start. Calculate it from your actual margins, don’t guess.
Step 1: Create Your Campaign
Log into Google Ads and click the blue “+ Campaign” button.
[image: Google Ads “New Campaign” button location]
You’ll see a list of campaign types. Select “Demand Gen.”
Next, you’ll choose your campaign objective. For ecommerce, select “Sales.” This tells Google you want to optimize for conversions and purchases.
The other options (Leads, Website Traffic) are available but “Sales” is the right choice for ecommerce stores.
Click Continue.
Step 2: Campaign Settings
This is where you configure the core campaign parameters.
Campaign Name
Name it something descriptive. Examples: “DG Lookalike Audience Shoes” or “DG Retargeting Customers Winter Sale.”
This helps you identify the campaign later when you’re looking at performance data.
Locations & Language
Select your target locations (countries, states, or cities). Start with your primary market. You can expand later once you see performance.
For language, select the language(s) your customers use.
Pro tip: If you only sell to the US, target “United States” only. Don’t select “All countries.” It wastes budget on irrelevant traffic.
Bidding Strategy
This is critical. Google will recommend “Maximize Conversions” with a target ROAS. Here’s what I recommend for ecommerce.
If you have 50+ conversions in the past 30 days: Use “Maximize Conversions” with target ROAS. Set your target ROAS to your realistic requirement, like 3.0 if you need 3.1 ROAS. Learn about Smart Bidding strategies.
If you have 20-50 conversions in the past 30 days: Use “Maximize Conversions” with target CPA instead. Set your target CPA to your affordable acquisition cost.
If you have fewer than 20 conversions in the past 30 days: Use “Maximize Conversions” without targets. Let Google optimize for volume first. Then you’ll have enough data to add targets later.
[image: Bidding strategy dropdown menu]
Important: Don’t use “Maximize Click-Through Rate” or “Manual CPC.” These aren’t designed for conversion optimization. You’ll waste budget on clicks that don’t convert.
Daily Budget
Set your daily budget. If you’re committing $100 per day, enter 100. If $200 per day, enter 200.
Google may spend up to 20% more on good days when there’s high-intent traffic. So a $100 budget might spend $120 some days. This is normal.
Important setting: Make sure “Accelerated” delivery is OFF. You want “Standard” delivery, which spreads your budget throughout the day. Accelerated delivery burns budget early in the day and can deplete it before the algorithm optimizes.
Campaign Start & End Dates
Leave start date as today. For end date, I’d recommend 6 weeks out. This gives the algorithm time to learn. That’s 4 weeks learning period plus 2 weeks of optimized performance.
You can always extend the end date later if performance is good.
Campaign Scheduling
Optional: If you want to exclude certain hours or days, use Campaign Scheduling. For most ecommerce stores, 24 per 7 is fine. Let the algorithm spend budget whenever it finds good opportunities.
Click Save and Continue.
Step 3: Connect Your Product Feed
Now Google asks about your product feed. This is important.
Select “Yes, use product feed” and then select your Merchant Center account and product feed.
[image: Product feed selection screen]
If you have multiple Merchant Center accounts, make sure you select the one with your actual products. You can only select one feed per campaign.
What this does: When someone sees your ad on YouTube or Discover, Google can show them specific products from your feed with current prices and availability. This is way better than generic ads.
If you don’t have a Merchant Center feed: You can still run Demand Gen without it, but performance will be weaker. The algorithm can’t show specific products. Set up Merchant Center first, focusing on product title optimization and proper categorization.
Select your feed and continue.
Step 4: Audience Targeting
This is where you define who sees your ads. This is critical. Bad audience targeting equals wasted budget.
Audience Signals
Google asks you to add “audience signals.” These are hints about your target audience. You can add multiple signals, and Google expands from there.
Option 1: Lookalike Audiences (Recommended)
Create a lookalike audience based on your existing customers. Go to Audience Manager, then Audience Sources, then Customer Match. Upload your customer email list.
Once you have a customer list, you can create a lookalike audience. Go to Audience Manager, then Audiences, then Lookalike Audience. Select your customer list as the source. Set reach to 500% or 1000% to find more similar people.
Add this lookalike audience as your primary audience signal in Demand Gen.
Option 2: Website Visitor Audiences
If you don’t have a customer list yet, use your website visitor audience, also called “Website Traffic” audience. This targets people who’ve visited your website but haven’t bought.
Google automatically creates this if you have Google Analytics or conversion tracking on your site.
Option 3: Interests
You can target people interested in specific categories. Examples: “Fitness & Wellness,” “Fashion,” “Outdoor Gear.” This is less precise than lookalike audiences but works if you don’t have customer data yet. Learn about in-market audiences and affinity audiences for more targeting options.
Best practice: Start with Option 1 (lookalike audiences) if you have customer data. It typically outperforms broad interest targeting.
Detailed Demographics
Optional: You can exclude or target specific age ranges, genders, or parental status. For most ecommerce stores, I’d leave this on the default with no restrictions.
Only narrow demographics if you have a product that’s clearly gender-specific or age-specific. Examples: children’s clothing or senior health products.
Optimized Targeting
Google enables “Optimized Targeting” by default. This allows Google to expand your audience beyond your specified signals.
My take: Leave it ON. Optimized targeting helps Google find more potential customers similar to your audience. The downside is less control, but in Demand Gen, you’re already accepting automation.
If you’re worried about control, you can turn it OFF, but expect slightly weaker reach.
Exclusions (Important)
You can exclude certain audiences.
Exclude your existing customers: Create an exclusion audience list of your customer emails. Don’t waste budget retargeting people who’ve already bought.
To do this: Go to Audience Manager, then Audience Sources, then Customer Match. Upload your customer list. In campaign settings, add this list as an exclusion audience. You can also use dynamic remarketing audiences for exclusion.
This is critical. Without excluding customers, you’re running an acquisition campaign that keeps trying to sell to people who already own your product.
Click Save and Continue.
Step 5: Channel Selection
Now you choose which Google properties to include: YouTube, Discover, Gmail, Google Display Network (GDN), Maps, etc.
My recommendation for ecommerce stores: Start ONLY with YouTube and Discover. Uncheck Gmail, GDN, and Maps.
Here’s why.
YouTube + Discover: These are high-intent, visual placements. They convert decently for ecommerce. Learn more about YouTube Ads and YouTube Shorts performance.
Gmail: The Promotions tab gets low engagement. Start here only after YouTube and Discover prove successful.
GDN (Google Display Network): Lots of low-quality placements. Ad fraud is common. Once you understand performance, add it back carefully.
Maps: Most people don’t click ads on Maps. Skip it.
After 2-3 weeks: If YouTube and Discover are converting profitably, add Gmail and GDN to expand reach.
Check ONLY YouTube and Discover. Uncheck everything else. Click Save and Continue.
Step 6: Create Your Ads
This is where you build your ads. Before uploading anything, understand the three ad formats that matter for ecommerce Demand Gen.
The 3 Ad Format Types
1. Video + Product Feed. A video ad (typically 15-30 seconds) paired with product cards from your Merchant Center feed. The video plays while your products display alongside it. This is the strongest format for ecommerce because you get the emotional impact of video plus the specificity of actual products with prices.
This is what replaced Video Action Campaigns. If you were running VAC before, this is your format.
2. Image + Product Feed. A static image ad paired with product cards from your feed. Works well on Discover and Gmail where people are scrolling. Lower production cost than video. Good starting point if you don’t have video assets yet.
3. Products Only (Feed-Based). Your product feed does all the work. Google pulls product images, titles, and prices directly from Merchant Center and creates the ad. No custom creative needed.
This is the easiest to set up. But it’s also the least differentiated. Your ad looks like every other product ad.
My recommendation: test all three. Video + feed typically wins, but image + feed can surprise you depending on your products and audience.
Uploading Your Assets
- Images: Upload 3-5 images. Landscape (1200×628), square (1080×1080), and vertical. Lifestyle photos showing product in use outperform product-on-white.
- Videos: Upload 2-3 videos. Vertical (9:16) videos perform better on Shorts and Discover. Keep them 15-30 seconds. Hook in the first 3 seconds or people skip.
- Headlines: Add 3-5 headlines (30 characters max). Emphasize benefits or urgency. “Shop Summer Collection Now.” “Free Shipping Over $50.”
- Descriptions: Add 1-2 descriptions (90 characters max). Optional but helpful for context.
- CTA: Select “Shop Now” for ecommerce. It’s the standard.
- Product Feed: Select “Yes” to show product ads. This is important. Product ads with prices convert better than generic images.
Review the preview across placements before launching. Make sure nothing looks blurry or cut off.
Step 7: Review & Launch
You’re now on the final review page. Google summarizes everything.
Campaign name, locations, bidding strategy, budget, audiences, channels, and creative.
Double-check these key points.
Bidding: Is your target ROAS or CPA realistic. If you set 5.1 ROAS but only need 2.1 profit, Google will underbid.
Budget: Is $100+ per day sustainable for 4-6 weeks. If not, reduce the campaign scope or wait.
Audiences: Did you add lookalike audiences and exclude existing customers.
Channels: Are only YouTube and Discover selected.
Creative: Do you have at least 3 images and 2 videos.
Product Feed: Is it connected.
If everything looks good, click “Create Campaign.”
Your campaign is now live.
What Actually Makes Demand Gen Work
The setup is the easy part. Getting Demand Gen to actually perform comes down to two things: audience testing and ad testing.
Everything else is secondary.
Audience Testing
Getting Demand Gen to work is about narrowing down on one or more audiences that convert for your products. This takes intensive testing.
Start with your best audience signal (usually a customer lookalike) and give it 2-3 weeks of data. If it works, great. If it doesn’t, try a different source.
Test these one at a time:
Customer email lookalike at different reach levels (500%, 1000%, 2500%). Wider reach means bigger audience but less similarity.
Website visitor lookalike based on your highest-value visitors (purchasers, cart adders, not just all traffic).
In-market audiences for your product category. Google identifies people actively researching products like yours.
Custom segments based on search terms your customers use or URLs they visit.
The mistake most people make: testing too many audiences at once. The algorithm can’t learn when you spread budget across five audience signals.
Test one. Give it time. If it doesn’t work, swap it. If it works, scale it and test a second alongside it.
Ad Format Testing
The three ad formats (video + feed, image + feed, products only) perform very differently depending on your products and your audience.
You won’t know which works until you test.
Run each format as a separate ad group within the same campaign. This lets you see which format drives conversions at the best cost.
Video + feed tends to win for products that need demonstration or lifestyle context. Fashion, beauty, home goods.
Image + feed can win for products where one strong image says everything. Jewelry, electronics, simple accessories.
Products only works best for catalog-heavy stores with hundreds of SKUs where you can’t create custom creative for everything.
After 2-3 weeks, you’ll see clear winners. Pause the underperformers. Double down on what works.
Then start testing variations within the winning format. Different videos. Different images. Different product selections.
This cycle of testing and narrowing is how Demand Gen goes from “meh” to profitable.
First 2 Weeks: What to Expect
Your campaign is running. Here’s what happens.
Week 1: Learning Period Starts
Google is collecting data on which placements, creative, and audiences drive conversions. Performance data will be sparse or erratic.
Don’t panic if you see.
High cost per click (CPC) in Week 1 as Google tests placements.
Few or zero conversions in days 1-3. It takes time for conversions to be attributed.
Inconsistent daily spend. Some days $80, some days $120. This is within 20% variance.
What you should do: Check the campaign daily, but don’t make changes. Let it run. The algorithm is still learning.
Week 2: Initial Data Emerges
By week 2, you’ll have 50-200+ conversions, depending on your budget and conversion rate. The algorithm is learning patterns.
You might see:
CTR (click-through rate) stabilizing.
CPC dropping as Google gets smarter about placements.
Cost per conversion becoming predictable.
Still don’t make changes. The “learning period” is typically 2-4 weeks. If you tweak things during this time, you reset the learning period.
Weeks 3-4: Optimization Phase
After 2 weeks, assuming 50+ conversions, the algorithm shifts into optimization mode. Google applies what it learned and optimizes more aggressively.
This is when you should start monitoring.
Is ROAS at or above your target.
Is CPA at or below your target.
Which creative (images/videos) are generating clicks.
Google Ads interface tip: Go to Assets, then Image/Video Performance to see which creative is driving clicks. Use this data to inform future creative decisions. You can also check Google Ads experiments for comparison data.
[image: Assets performance report in Google Ads]
Weeks 5-6: Scale or Pause
By week 6, you have enough data to make decisions.
If ROAS is above target: Increase daily budget by 20-30%. More budget equals more conversions at same ROAS.
If ROAS is below target: Reduce budget or pause the campaign. Something’s not working. Consider:
Is your lookalike audience the right fit.
Are your creatives compelling.
Is your product feed up to date.
If ROAS is close to target but not there: Give it another 2 weeks. Changes take time to propagate through the algorithm.
Common Setup Mistakes to Avoid
Mistake 1: Budget Too Low
Launching with $30-50 per day and expecting great results. The algorithm needs volume to learn. You’ll see high CPAs and erratic performance. Commit to $100+ per day minimum.
Mistake 2: Too Many Audiences
Adding 5+ lookalike audiences, multiple customer lists, and broad interests all at once. This confuses the algorithm. Start with ONE primary audience signal, your main lookalike audience. Once it stabilizes, test a second audience.
Mistake 3: No Customer Exclusion
Forgetting to exclude existing customers. Your campaign spends budget trying to sell to people who already own your product. Always exclude customer lists.
Mistake 4: All Channels Enabled
Launching with YouTube plus Discover plus Gmail plus GDN plus Maps all at once. This spreads budget across low-performing placements. Start narrow with YouTube and Discover, then expand.
Mistake 5: Poor Product Feed
Linking a product feed with outdated prices, wrong images, or missing data. Google can’t show products properly. Update your feed before launching the campaign. Remove out-of-stock items or set them to 0 quantity.
Mistake 6: Unrealistic ROAS Target
Setting target ROAS to 5.1 when your business only needs 2.5.1 to be profitable. Google will underbid to hit your target, limiting conversions. Use realistic numbers based on actual business metrics.
Mistake 7: Changing Settings During Learning Period
Tweaking audiences, bidding, or budgets in Week 1-2. Every change resets the learning period. Let it run for 2 weeks before optimizing.
Mistake 8: Not Using Vertical Video
Only uploading landscape videos. YouTube Shorts and Discover feed show vertical video. Vertical videos get better CTR and conversion rates. Create vertical versions of your ads with 9.16 aspect ratio.
Mistake 9: Ignoring Creative Performance Data
Launching with 10 images and 5 videos, never checking which ones drive performance. After week 2, look at Assets, then Image Performance. Pause underperforming creative and refresh with new angles.
Mistake 10: Running Demand Gen + PMax Simultaneously
As I mentioned in our complete Demand Gen guide, running both campaigns means they compete for conversions. Unless you have a specific audience segmentation strategy, pick one. See our detailed comparison at Demand Gen vs Performance Max.
One more thing. You’ll sometimes hear that you shouldn’t run Demand Gen and Performance Max at the same time. That’s wrong.
They target different parts of the funnel. PMax captures bottom-funnel demand (Search, Shopping). Demand Gen builds top-of-funnel awareness (YouTube, Discover). They’re complementary, not competing.
Run both if your budget allows.
Next Steps After Launch
After 4 weeks of data, you have enough to make informed decisions.
If it’s working: Increase budget by 20-30%. Refresh creative with new images and videos. Expand to additional channels like Gmail and GDN if core channels are profitable. Test lookalike audiences at different reach levels.
If it’s not working: Pause the campaign. Analyze: Is the creative engaging. Is the audience the right fit. Is the product feed current. Fix the issue, then relaunch with revised strategy.
Ongoing optimization: Review performance weekly. Pause underperforming creative. Test new creative quarterly. Monitor ROAS versus target. Scale profitably or pause underperforming campaigns.
Compare your performance against Demand Gen benchmarks to see where you stand in your industry. Also review broader Google Ads cost data and Google Ads benchmarks for context.
Demand Gen is a long game. Give it time, measure honestly, and optimize based on real data. Not Google’s recommendations.
