Demand Gen Budget: Where It Goes and How to Control It

A practical reference on how Demand Gen allocates spend across YouTube, Discover, and Gmail, and the levers that actually move it.

· QueryCut

The short version

  • Demand Gen has no per-surface budget; the YouTube, Discover, and Gmail split is an output of your bid strategy and creative formats.
  • Creative format is your strongest lever: the aspect ratios you upload determine which surfaces you can serve on.
  • Daily spend can hit 2x budget on a given day; judge pacing on month-to-date spend versus expected pace, not single days.
  • Compare accounts on spend-versus-pace percentage, not raw dollars, so hot and cold accounts surface instantly.
  • Check the surface breakdown before cutting budget; the money is often on Discover and Gmail, not YouTube.

Demand Gen campaigns spend your budget across YouTube (in-feed, Shorts, in-stream), Discover, and Gmail, and Google decides the split automatically based on where it predicts your bid strategy will hit its goal most cheaply. You cannot set a per-surface budget. The real controls are your bid strategy, the creative formats you upload, and how you split traffic into separate campaigns. If spend is going somewhere you dislike, you change those inputs, not a slider.

Where the money actually goes

A single Demand Gen campaign is eligible to serve across every surface Google has mapped to it: YouTube in-feed and in-stream placements, YouTube Shorts, the Discover feed, and Gmail promotions and social tabs. Your daily budget is a shared pool, not a set of earmarked buckets. The system allocates in real time toward whatever placements its model expects to deliver conversions (or clicks, depending on your goal) at or below your target.

This is the first thing to internalize: there is no UI field where you say 'put 40 percent on YouTube.' The split is an output of the auction and the bid model, not an input you control directly. Two campaigns with identical budgets and targeting can end up with wildly different surface mixes because their creative assets and conversion histories push the model in different directions.

The practical consequence is that the surface mix drifts. A campaign that spent mostly on Discover last month can shift toward Shorts this month if the model finds cheaper conversions there, or if you added a vertical video asset that unlocked more Shorts inventory. Treat the mix as something you influence, not something you dial in.

The three levers that actually move allocation

Creative format is the strongest lever, because format determines which surfaces you are even eligible to serve on. Upload only square and portrait images and you will skew toward Discover and Gmail. Add a 9:16 vertical video and you open Shorts and vertical in-stream. Add a 16:9 landscape video and you open standard YouTube in-stream. If you want more or less of a surface, add or remove the assets that qualify for it. This is the closest thing to a direct control you have.

Bid strategy shapes where the money flows within your eligible surfaces. Maximize Conversions with a tCPA tells the model to chase cheap conversions wherever it finds them, which often concentrates spend on whatever surface currently has the best measured conversion rate. Maximize Clicks spreads differently, usually toward cheaper-click placements like Discover. Switching from a volume goal to an efficiency goal (or setting a tighter tROAS) will visibly reshuffle the surface mix within a week or two of relearning.

Campaign segmentation is the blunt but reliable lever. If you need hard separation (say, you want YouTube spend walled off from Gmail spend), the only clean way to get it is separate campaigns with different creative sets, because you cannot exclude an individual Demand Gen surface inside one campaign the way you can exclude placements in a standard Display or Video campaign. Splitting gives you separate budgets and separate reporting at the cost of thinner data per campaign and slower learning.

Reading where it went and pacing across accounts

To see the actual split, go to the campaign, then use the segment and breakdown tools: Campaigns > the Demand Gen campaign > look at the format and placement breakdowns available under the campaign's reporting. YouTube, Discover, and Gmail performance is reported separately enough that you can see which surface ate the budget and what it returned. Check this before you touch any setting, because the fix depends entirely on which surface is overspending relative to its results.

For budget pacing, the number that matters is spend against daily budget over a trailing window, not a single day. Demand Gen, like other Google campaigns, is allowed to spend up to twice your daily budget on a high-opportunity day and will balance it against low days so the monthly total lands near your daily budget times the days in the month. A single day at 180 percent of budget is normal and not a sign of a problem.

People managing multiple accounts (the 'comparing budget pacing across accounts' question) should pull pacing at the account or MCC level rather than eyeballing campaigns one at a time. Build a view that shows month-to-date spend versus the expected linear pace (budget times days elapsed) for each account, so an account running hot or cold stands out immediately. The failure here is comparing raw spend instead of spend-versus-pace: a big account and a small account will always differ in dollars, but both should sit near 100 percent of their own expected pace.

Decision rules for tightening or redirecting spend

If one surface is spending heavily with poor returns, do not reach for a budget cut first. Diagnose the cause. If Discover is burning budget at a weak conversion rate, the usual culprit is cheap, low-intent clicks that your bid strategy is chasing because you are on a volume goal. Move to a tCPA or tROAS target and let it relearn before judging the result.

When you do adjust the target, move it gradually and give the model time to settle between changes. Large overnight swings in tCPA or budget trigger a fresh learning period and can produce erratic spend for days, which defeats the purpose. Smaller, spaced adjustments are a practitioner heuristic for keeping the campaign out of constant relearning, not a hard system rule: the tradeoff is slower correction in exchange for stabler delivery.

Give any new campaign or major change enough volume before you trust an efficiency read. On a brand-new Demand Gen campaign with a handful of clicks and no conversions, the surface mix and the apparent CPA are both noise. Wait until you have accumulated meaningful click and conversion volume before concluding a surface is bad. How much is 'enough' is a judgment call that depends on your conversion rate: a campaign that converts at 10 percent needs far fewer clicks to produce a trustworthy read than one converting at half a percent.

The traps that waste budget

The most common mistake is assuming the budget is sitting on YouTube because Demand Gen feels like a video product. Many accounts discover their spend is dominated by Discover and Gmail, where clicks are cheaper and the model finds easy volume. If you optimized your creative and messaging for a YouTube audience but most impressions are served in Gmail, your numbers will disappoint for reasons that have nothing to do with the budget level. Check the breakdown before you change the budget.

A second trap: trying to control surface mix by pausing assets reactively. Removing your best-performing video to 'force' spend onto Discover usually just lowers overall efficiency, because you have removed an asset the model was winning with. Add assets to open surfaces; be cautious about removing assets to close them.

The honest limit: Google does not expose a per-surface budget control, does not let you exclude a single surface within a Demand Gen campaign, and does not fully explain its allocation logic. You are working with influence, not direct control. The reliable move is to manage the inputs you can see (creative formats, bid targets, campaign structure), measure the surface breakdown weekly, and accept that some of the mix is the system's call and will shift on you.

Questions people also ask

Can I set a budget just for YouTube in a Demand Gen campaign?

No. Demand Gen uses a single shared budget across YouTube, Discover, and Gmail, and you cannot earmark an amount per surface or exclude an individual surface. If you need hard separation, the only reliable method is building separate campaigns with different creative sets, accepting thinner data and slower learning in each.

Why is my Demand Gen campaign spending mostly on Discover instead of YouTube?

The bid model allocates toward whichever surface it predicts will hit your goal most cheaply, and Discover and Gmail often have cheaper clicks and easy volume. Your creative formats also matter: if you uploaded mostly static images and no vertical video, you are barely eligible for Shorts and vertical in-stream, so spend concentrates on feed surfaces.

Is it normal for daily spend to exceed my daily budget?

Yes. Google can spend up to twice your daily budget on a high-opportunity day and balances it against lower days so the monthly total stays near your daily budget times the number of days in the month. Judge pacing on a trailing window, not a single day, before concluding something is wrong.

How do I compare budget pacing across multiple accounts?

Work at the MCC or account level and compare month-to-date spend against expected linear pace (daily budget times days elapsed) as a percentage, rather than comparing raw dollar spend. That normalizes for account size, so an account running at 130 percent of its own pace stands out next to one sitting at 90 percent regardless of total budget.

How quickly should I adjust bids or budget if Demand Gen is underperforming?

Make gradual, spaced changes rather than large overnight swings, because big jumps in tCPA or budget trigger a new learning period and destabilize delivery for days. Also give a new campaign enough click and conversion volume to produce a trustworthy read before acting; with a handful of clicks and no conversions, both the CPA and the surface mix are noise.

More from the blog

Landing Page Experience: The Changes That Actually Move ItImpression Share Lost to Rank vs Budget: What Each One MeansHow Many Campaigns Is Too Many for the Budget You Have?Primary vs Secondary Conversion Actions: What Each One Actually Changes