> ## Documentation Index > Fetch the complete documentation index at: https://dogoodthings.co.nz/docs/llms.txt > Use this file to discover all available pages before exploring further. # Paid ads > Media buying run against the growth sheet, not platform-reported ROAS. Media buying is where the system's economics play out. We run it against the [growth sheet](/system/data-attribution#the-growth-sheet), not platform-reported ROAS, with an account structure built around one distinction: creating demand is a different job from capturing it. ## Push and pull: creating demand versus capturing it Meta and the social platforms are push marketing, the modern TV. Nobody there is searching for you. The ad steps into the feed and creates the desire, building awareness and making new people want the thing in the first place. For consumer brands, this is where most of the budget should go, because it's the channel that grows the pool of people who want what you sell. Google Ads, Bing Ads, and increasingly ChatGPT and the other AI assistants work the other way: they capture. They catch the people who are already searching, often because push marketing put the idea there. Capture channels are essential, and we keep them in proportion. Let them grow past the demand being created and they start absorbing budget to convert people who were coming anyway. The account looks great; the business grows slower. Within the push channels, prospecting and remarketing do different jobs too. Prospecting finds new people: it's where creative variety lives, and frequency should sit under about 2 over a month. Remarketing repeats the ask to people who've leaned in, where frequency above 2 is the point. Exclusions keep the two separate: site visitors, engagers, and existing customers move down the funnel instead of being counted as new people forever. ## Give the algorithm something to work with Modern delivery algorithms have absorbed most of what audience settings used to do. Running paid well is now less about media buying mechanics and more about giving the platforms enough creative diversity to optimise against. Every distinct angle, hook, and format is another doorway to a new pocket of people. The sharp implication: if all your ads say the same thing, broad targeting will find the same kind of people forever. Changing who you reach starts with changing what the ads say, which is why creative strategy and media buying are one discipline here, not two departments. ## How an account runs month to month ### Structure that's built to learn Prospecting and remarketing separated, exclusions in place, and overlapping campaigns unpicked so they stop bidding against each other. It's done incrementally, staged around what's already working, and never as a rip-out. Long-running winners keep running. ### Make sure the platforms optimise on the right signals Qualified outcomes flow back to the platform, as covered in [data & attribution](/system/data-attribution), so optimisation chases your best customers instead of your cheapest form-fills. Platform automation features get adopted on their merits, not by default. Some help; some can wreck a working account. It's worth knowing which before the switch gets flicked. ### Stay ahead of saturation This isn't watching dials. It's keeping a finger on the pulse and moving before problems land. Frequency creeping up in prospecting means creative is heading toward saturation, so new angles go into production before the burn-out, not after it. Learning Limited means the structure or the signal needs work. Read weekly, problems show up as trends we've already acted on instead of a surprise bad month. ### Use platform signals, trust your own numbers Platform metrics are useful leading signals, and we use them to optimise inside each channel. We don't take them as gospel. Budget moves when the growth sheet says the economics hold, never because a platform dashboard flattered itself. Increases are incremental: scale in steps, watch the growth sheet, keep going while it holds. When spend deserves cutting, we cut fast and explain why. ## How we split budget across channels * **Meta and social: demand creation.** Usually the biggest share of budget. This is the channel that builds desire and brand awareness, the one that makes new people want the thing. * **Google and Bing: demand capture, kept in proportion.** Search and PMax capture the demand that social and brand create, so we typically keep capture spend at roughly 20 to 25% of the social budget, and let it grow only as the demand does. * **Where your audience actually lives.** Channel picks follow the audience, not fashion. Older demographics sit natively in Microsoft's ecosystem, for example, so Bing reaches the over-45s with hardly any competition. Niche channels get tested when [the audit](/the-audit) says the audience and the economics are real. * **AI-era placements.** Visibility and ads inside AI assistants as they open up. Early days; covered in [AI Search/SEO](/system/ai-search-seo). **Don't break what works** If an ad has paid its way for 18 months, it has earned its budget. We build the new structure around it and retire it only when a tested replacement beats it. Plenty of working accounts have been optimised to death in week one of an agency takeover. We'd much rather inherit your momentum than restart it. ## Spend and access Ad spend is paid by you, directly to the platforms, in accounts you own. We never mark up media and we take no spend-based commission, so when we say raise, hold, or cut, the recommendation has nothing to do with our fee.