The split between Facebook and Instagram should be an outcome, not an opening decision. Meta will move spend toward whichever placement converts if you let it, and constraining that early usually costs money. Still, the pattern of where budget lands is predictable enough to plan around.
Weight toward Facebook when the buyer is older or local
Home services, trades, roofing, legal, healthcare, and anything sold to an established homeowner tends to find its audience on Facebook. The user base skews older, community and local groups still drive real behaviour, and in-platform lead forms suit buyers who will not tolerate a long journey to a website.
Detail on placements, audiences, and lead forms is on the Facebook page.
Weight toward Instagram when the product is visual
Restaurants and bars, med spas, salons, event venues, retail, and anything bought partly on how it looks. The audience skews younger, discovery happens through Reels and Stories, and the creative bar is higher because the surrounding content is polished.
Formats and creative cadence are covered on the Instagram page.
Let the system decide the rest
Beyond those tendencies, splitting budget by app manually tends to underperform. Meta has more data than you do about where a given person is reachable most cheaply. Our job is giving it accurate conversion signals and enough creative variety to work with, then checking where it landed.
Structure decides whether any of it learns
The most common problem in an inherited Meta account is fragmentation: fifteen campaigns, each with a handful of ad sets, none receiving enough conversions to exit the learning phase. Everything looks organised and nothing gets good.
Consolidation usually improves results within a fortnight, and it is the first thing we look at. Fewer, better-fed campaigns beat a tidy structure every time.
Measurement is the foundation, not the report
Meta optimises toward the events you report. If your pixel fires a conversion when someone views a contact page, it will find people who view contact pages and leave. If server-side events are missing, a large share of real conversions are invisible and the system is learning from a fraction of the truth.
We fix that layer before touching budget, because every decision above it inherits its errors.