Why fewer campaigns beat clever ones
Over-segmentation is the most expensive habit I see in DTC accounts, and almost nobody thinks of it as a cost. You inherit an account with 30 campaigns, tightly themed ad groups, three remarketing stacks, and a bidding setup someone was clearly proud of in 2018. It looks like control. It performs like a starved algorithm.
The reasoning is simple once you accept how Google actually works now. Conversion data is the fuel that trains the system, and when you split spend into too many little buckets you turn one useful pool of signal into a set of tiny puddles that never fill up. None of them learn quickly. All of them look mediocre. So you add more structure to fix the mediocrity, and you make it worse.
The shape I push for is deliberately boring. Roughly 3-5 campaigns total, and often even simpler than that. One Performance Max campaign carrying the commerce load, a couple of Search campaigns doing specific jobs, and you are done.
People hear that and assume it is laziness dressed up as strategy. It is the opposite. A consolidated account learns faster, stabilises sooner, and holds performance better as you push budget through it. Every extra campaign you add should have to earn its existence with a real reason, like a genuine margin gap or an audience that behaves completely differently.
My default position is to start consolidated and split only when the evidence forces me to, never the other way round. Most accounts I audit are carrying five campaigns of pure ego and paying for it in learning speed. Strip that out and the machine finally gets enough data to do the thing you hired it to do.
Your feed is doing the selling now
Rename one product title and you can watch the whole strategy fall into place. Take a SKU sitting in Merchant Centre with an internal name like "MW-BLK-42-V2" and rewrite it as the thing a customer would actually type into Google. That single edit changes what the system understands you sell, which queries it matches you to, and whether it decides to serve the product at all.
That is why I keep saying your feed is your account. For ecommerce, Merchant Centre is the layer Google reads to work out what you stock, how to prioritise it, and where to show it across its surfaces. If the feed is vague, the system quietly decides it does not understand the product and never serves it. You never see the loss. It just shows up as underperformance you cannot explain.
I treat feed work with the same seriousness as landing page work, because commercially it does the same job. Keyword-rich titles written for how people search, clean and legible images, correct GTINs, accurate Google product categories and product types, descriptions in real customer language, and custom labels so you can segment by things like margin tiers, hero SKUs, or price bands later.
None of it is glamorous. It is the unsexiest quarter of work a growth team can sign up for, and it is also the highest-leverage. A strong feed lets PMax find demand efficiently and lets you scale with far less manual scaffolding.
If a brand asked me for one thing to fix before touching a single bid, I would send them straight back to the taxonomy and the titles.
PMax quietly ate your retargeting stack
Put Performance Max next to the old Google Display Network remarketing setup and the comparison is almost embarrassing. One is a consolidated engine that reads your feed, ingests your creative, and distributes across Shopping, YouTube-style inventory, and everything in between. The other is a bolt-on retargeting stack that most brands are keeping alive out of habit.
For the majority of ecommerce accounts I would let PMax be the centre of gravity. It is the main shopping and scale vehicle, and I run it as a single consolidated campaign with asset groups mapped to major categories or hero lines. The temptation is to slice it into ten neat asset groups because it feels tidier. Resist it. You are back to draining the signal you worked to pool.
PMax is only as good as what you feed it, and this is where most brands fall down. It needs plenty of images and, more importantly, video. Simple, native-feeling video usually outperforms the overproduced stuff, and it is what unlocks the YouTube-adjacent placements sitting inside the system. Brands with weak video supply are the ones who quietly cap their own results.
Standalone GDN retargeting is largely a dead end now. I would let PMax handle retargeting once it has enough assets to work with, or move that job to Demand Gen if I wanted cleaner control. Keeping a legacy remarketing campaign running alongside all this is just paying rent on a channel Google has already folded into something better.
The brands winning here are not the ones with the most sophisticated targeting logic. They are the ones with a fat library of decent creative pointed at a clean feed.
Splitting brand off from everything else
Here is the one place I actively want a separation, which might sound odd after two sections arguing against them.
Branded Search gets its own campaign, and it gets it for two specific reasons. It is usually your most efficient traffic by a distance, and if you let it bleed into your prospecting campaigns it will flatter numbers that deserve scrutiny. A healthy blended ROAS can hide genuinely weak non-brand performance, and you will not notice until the brand demand plateaus and the whole account sags.
Into the branded campaign goes the brand name, brand plus product combinations, and common misspellings. That is roughly it. This is defensive work. You are protecting demand you already generated and making sure a competitor does not sit on your name and tax your conversions on the way through.
I want to be honest about what this campaign is not, because founders sometimes point at a beautiful branded ROAS and call it growth. It captures demand. It does not create it. Reading branded performance as proof your acquisition is working is one of the more common self-inflicted wounds I see.
Non-brand Search sits in its own campaign chasing high intent that you have to earn, and keeping the two apart is what lets you read each honestly. When brand is isolated, you can finally see whether your prospecting stands on its own or whether it has been quietly leaning on people who already knew you.
That clarity is worth the extra campaign, which is the whole test I apply before splitting anything.
Broad match stopped being the villain
Most people still flinch at broad match because they remember what it did to their budget a few years ago. Broad match used to be the setting you turned off first, the one that dragged you into every loosely related query and lit money on fire.
That reputation is out of date. Intent understanding has improved a lot, and searches themselves have grown longer and more conversational, which plays directly to broad match's strengths. Run the modern way, it is now materially better than the tight, sculpted keyword lists people still cling to.
There is one hard rule I will not bend on. Broad match with manual CPC is a disaster, full stop. Broad match only works when it is paired with Smart Bidding, because the whole point is to hand the machine freedom and then give it a clear objective to optimise towards.
For ecommerce that usually means Target ROAS once you have stable conversion-value tracking, or Maximise conversion value earlier on, later constrained with a ROAS target as the data firms up. The label matters less than the principle. The system needs room to move and reliable feedback about what a good outcome looks like.
What trips brands up is trying to have it both ways, running broad match while still "managing" it like a 2018 account with negative keywords and manual bids everywhere. You get the loose matching without the intelligence that is supposed to steer it. That is the actual disaster, not broad match itself. Give the machine the keywords and the bidding strategy it was built for, and it will do work your spreadsheet never could.
When Search simply runs out of demand
What do you do when PMax is humming, Search is dialled in, and the account still will not grow? At some point you hit a ceiling that has nothing to do with your setup and everything to do with how many people are actively searching for what you sell.
This is where Demand Gen earns its place. Search and PMax are largely capturing demand that already exists. Demand Gen creates it, running creative-led prospecting across strong placements including Discover-style experiences, which is a far more modern and higher-quality version of the old prospecting display idea.
I do not reach for it early, and I would not add it to an account that has not sorted its inputs first. I add Demand Gen when PMax is genuinely working, when there is strong creative in the tank, especially video, and when the brand is ready to invest in top and mid-funnel demand creation rather than just harvesting whatever is already there. Bolt it on without those things and you are just spending into an audience that has never heard of you with creative that cannot carry the conversation.
There is a second use I like. If you want explicit control over retargeting, with sequencing, exclusions, and creative rotation, Demand Gen retargeting is cleaner than legacy GDN remarketing ever was. That is the option for brands that do not want to hand every remarketing decision to PMax.
Either way, Demand Gen is the campaign you graduate into once the fundamentals are stable, not the shortcut you use to paper over a thin feed or a demand problem you have not admitted to yet.
The comfort of thirty campaigns is a trap
A good structure fails without a good cadence, and the cadence I run is deliberately calm. The instinct with automated accounts is to fiddle, and fiddling is exactly what wrecks them.
Measurement comes first because everything downstream depends on it. Purchase and conversion value have to be tracked cleanly, attribution settings need to stay consistent over time, and you have to be disciplined about what actually counts as a primary conversion. You are not tracking to fill a report. You are training the optimiser, and a noisy or delayed signal teaches it the wrong lesson.
Budget scaling follows the same logic. Increase gradually, avoid constant resets, and let learning stabilise before you judge anything. Every big structural change and every bid-strategy swap wipes part of what the system has learned, so the more you restructure the more you reset yourself to zero.
Most of my weekly effort goes into the feed and the creative, not the campaign structure. Feed optimisation is the relevance engine, creative volume and quality is the distribution engine, and account structure is just the container that keeps the signal pooled. So the work skews towards better titles, images and descriptions, fresh assets in PMax and Demand Gen, and a catalogue and tracking that stay healthy.
Founders and investors often love the look of 30 tidy campaigns because granular reporting feels like control. If that structure is harming learning, you are paying for the illusion of control with real margin. The better question is whether you are compounding performance by pooling signal and improving inputs. Answer that honestly and most of the tidy little buckets stop looking so comforting.
Where I think this all goes next
Stand back from the individual campaigns and the whole 2026 playbook is really one shift, from micromanaging an account to training a machine. The 3-5 campaign shape, the feed obsession, the asset-heavy PMax, the branded and non-brand split, the broad match rule, the calm operating cadence. They all serve the same idea, which is that Google Ads is now an optimisation system that rewards pooled data and strong inputs and punishes clever structure.
What I keep coming back to is that the real competitive advantage has moved. It is no longer account complexity or some bidding trick you found. It is the quality of your inputs and the speed of your iteration. Feed hygiene, creative volume, clean conversion tracking. The unglamorous inputs the system depends on.
My honest read on where this heads next is that the gap between winners and everyone else widens on exactly those inputs. Google will keep automating the structural layer, which means the parts humans still control shrink down to feed, creative, and measurement discipline. Brands with weak video supply, messy taxonomy, or unstable tracking are going to struggle hardest, because there will be fewer and fewer levers left to hide behind. The teams that treat their feed and their creative as core marketing, run a handful of consolidated campaigns, and leave the machine alone long enough to learn are the ones I would bet on to compound. Everyone still sculpting keywords at midnight is optimising for a version of Google Ads that has already gone.
