DEMAND GEN BEST PRACTICES GOOGLE ADS: WHAT I ACTUALLY APPLY, NOT THE GENERIC CHECKLIST

Search "demand gen best practices google ads" and you get the same flat checklist everywhere: add more assets, use Smart Bidding, enable optimized targeting, refresh creatives. None of it is wrong. The problem is that it's presented as twenty items of equal weight, when on a live account three or four of them decide everything and the rest are rounding error. After a couple of years running these campaigns, I don't work through that list top to bottom I work in a priority order, because some best practices gate all the others and the generic version never tells you which.

So this isn't another checklist. It's the order of operations I actually follow: what I fix first, what the minimum viable build looks like, and the specific handful of checks I run before I let a campaign scale. If you want the platform fundamentals first, the primer on what Demand Gen is covers the ground; this piece assumes you know roughly what the campaign type does and want to know what a practitioner does differently from the help doc.

One framing up front, because it explains the whole priority order. Demand Gen is an automated, AI-driven campaign: you're not setting bids per keyword, you're feeding a system structure, assets and data and letting it optimize. That means the best practices that matter most are the ones that determine how well the system can learn account structure and data quality not the cosmetic ones. Get those wrong and no amount of creative polish or bid tweaking saves you. Get them right and the campaign mostly does the rest. That's the lens behind everything below.

The priority order I actually use

If I had to compress my google ads demand gen best practices 2026 into a sequence, it's four things in this order: structure the account so the AI learns at the campaign level, give every ad the minimum asset coverage so it can serve everywhere, fund the right bidding strategy, then verify data and tracking before spending. Notice what's first. It isn't creative and it isn't bidding it's structure, because the AI learns at the campaign level and a fragmented account starves every part of it before the other best practices even get a chance to matter.

The reason I order it this way is that these steps depend on each other in one direction only. Good creative on top of a fragmented account still underperforms, because the data is split too thin to optimize. A perfect bid strategy pointed at a broken conversion signal just spends faster into nothing. But fix the structure and the data, and even average creative starts to find its footing. So I treat structure and data as prerequisites, assets as the thing that lets the campaign serve, and bidding as the throttle and I never reorder them, because every time I've tried to fix creative before structure I've just made a thin account prettier.

This priority order is also why I'm wary of optimization-score-style advice that hands you a long undifferentiated list. The list is real, but flat. What separates the campaigns that scale from the ones that stall in my accounts isn't doing more items it's doing the first few correctly before touching the rest. The sections below are just that order, expanded.

Account structure: consolidate, don't slice

The single best practice I'd defend hardest is consolidation. Google AI learns at the campaign level, so the Search instinct to slice everything into tight, granular ad groups is actively harmful here every split divides the conversion data the optimizer needs. I combine similar audience themes into broader ones (the official guidance even gives the example of merging "baseball" and "football" into "sports"), keep prospecting consolidated for maximum reach, and only carve out a separate ad group when I genuinely need different creative for a distinct audience, typically remarketing versus prospecting.

My practical rule is a signal-density floor. If an ad group is generating fewer than roughly 30 conversions in 30 days, I merge it into another rather than let it limp along starved of data. Sparse ad groups don't just underperform themselves they drag the campaign's learning because the AI is trying to optimize across pockets too small to read. This is the same logic that decides whether a whole Demand Gen campaign ramps or sits flat: the algorithm can only optimize the signal you've concentrated for it, and consolidation is how you concentrate it.

I also leave the campaign running across all channels YouTube, Shorts, Discover, Gmail unless there's a hard reason not to. Restricting placements feels like control, but it cuts off inventory the optimizer uses to find cheaper conversions, and in my experience it raises CPA more often than it protects anything. The exception is a genuine brand-safety constraint, and when I apply it I treat the higher cost as a deliberate trade, not a default. Channel coverage and audience consolidation are the same idea: give the system room, don't fence it in.

The minimum asset set I won't launch below

Once the structure is right, the next thing that gates performance is whether each ad can physically serve everywhere. So I have a hard floor I won't go below: three vertical, three square and three horizontal images, plus at least one video in each of those three orientations. That's not me being maximalist it's the exact coverage that lets one ad run across YouTube, Shorts, Discover and Gmail instead of competing for a third of the inventory. Below that floor, the prettiest creative in the world still leaves most of the budget's preferred placements unserved.

This is also Google's published creative-variety benchmark, which is worth reading in full because it ties each element to a reason rather than a number to hit. Google's own best practices for high-performing Demand Gen campaigns set the benchmark at three images per ratio and a video per orientation, alongside running on every channel coverage restated as official guidance. On top of that floor I want several genuinely different headlines and descriptions, because the system finds winners by recombining text and visuals, and repeated near-identical copy gives it nothing to test.

Where I diverge from the generic "add the maximum number of assets" advice is on quality versus count. Maxing every field with thin, force-filled variants gets you an "Excellent" ad strength that doesn't actually spend. I'd rather have nine deliberate images and three real videos that each say something distinct than twenty mediocre ones padded to hit a gauge. Quality matters it just matters second, after coverage, because a beautiful asset that can only fill one ratio still can't serve where the budget wants to go.

The 3-4 checks I run before scaling a budget

This is the part the generic checklists skip entirely, and it's the part that has saved me the most money. Before I add budget to any Demand Gen campaign, I run the same short diagnostic, in order, and if any check fails I fix it before touching the budget rather than after.

One has it actually cleared learning? The campaign shows "Learning" until the algorithm has enough data, and Google's guidance is not to judge performance until it has passed roughly 50 conversions. I won't scale a campaign that hasn't, because the numbers I'd be scaling on are noise. Two is conversion tracking clean and firing? Scaling a broken or double-counted signal just buys bad data faster, so I confirm the tag is working and counting what I think it's counting before I commit more spend. Three does the asset pack have full coverage? Extra budget needs placements to flow into; if the ad is thin on ratios or video, more money just concentrates on the few placements it can fill and CPA climbs.

The fourth check is about how I scale rather than whether I should, so it bridges into the next section: am I raising the budget gradually rather than in a jump that resets learning. Those four past learning, clean tracking, full coverage, gradual increase are the entire gate. They look obvious written down, but I've watched plenty of accounts scale a campaign that failed two of them and then wonder why the cost per action doubled. Bidding and budget set the pace, but the audience signal and clean data are what the system actually learns from, which is why I check the data before I ever touch the throttle. If you want the bidding side in depth, the Demand Gen bidding strategies breakdown covers which strategy to fund and how much.

How I actually scale: slow on purpose

When all four checks pass, I scale but deliberately slowly, because the bidder is most fragile exactly when you most want to push it. At launch and through learning, the system is figuring out what works, and any large change to budget or targets forces it to relearn from a worse starting point. So my rule is single-digit weekly moves where I can and never more than about 15% at once. I'd rather flex a working budget up 5 to 10 percent, let it settle for a few days, and flex again than double it and watch the campaign destabilize.

This feels frustratingly slow the first few times, and almost every aggressive jump I've made taught me why the slow version wins: the CPA spikes while the system re-stabilizes, and the spike usually eats whatever extra volume the bigger budget bought. The same caution applies to creative when I refresh assets I add before I remove, change only a portion at a time, and only pause an underperformer after it's had a couple of weeks to ramp. Add, don't replace; flex, don't jump. The whole scaling philosophy on Demand Gen is to broaden gradually so the system never has to start over, and it's the practice I see violated most by people who treat "scale the winners" as permission to move fast.

The best practices I quietly skip

Not every item on the standard list earns its place, and part of having a real priority order is knowing what to ignore. I don't chase a perfect "Excellent" ad strength as a goal in itself I use it as a coverage gauge, and an Excellent that came from force-filling fields tells me nothing. I don't over-segment audiences in the name of precision, because on an AI campaign that just fragments signal. And I don't restrict channels or hand-pick placements unless there's a brand-safety reason, since narrowing the surface reliably raised my CPA more than it improved quality.

I'm also skeptical of constant tinkering dressed up as optimization. Demand Gen rewards patience: the 15% rule, the 50-conversion threshold, the 14-day ramp before judging an asset these all point the same way, toward changing less, more slowly. The best practice nobody lists is restraint. Most of the times I've hurt a campaign, it was by doing too much too soon to a setup that was actually fine. So my real list is short: structure it right, build the minimum coverage, fund the right bid, verify the data, then mostly leave it alone and scale in small steps.

The practical takeaway

If you take one thing from these demand gen best practices google ads: they aren't equal weight, so stop working a flat checklist and start working an order. Structure first so the AI can learn, then minimum asset coverage so the ad can serve, then the right funded bid strategy, then clean data and only after all of that, scale, gradually. The 3-4 checks before you add budget (past learning, clean tracking, full coverage, ≤15% increases) are the cheapest insurance in the whole workflow.

The version of best practices that actually holds up on live accounts is shorter and more opinionated than the generic one: get the first four things right, run the pre-scale gate every time, and resist the urge to fiddle. Do that and the campaign has room to do the part it's genuinely good at. Skip the order and chase the cosmetic items first, and you'll have a beautifully optimized account that still can't scale.