The calculation
Three numbers give you an answer: how many customers you need, what share of enquiries become customers, and what an enquiry costs.
Say you want ten new customers a month and you close one enquiry in four. That is forty leads. At $35 a lead, that is $1,400 a month — about $47 a day.
If you do not yet know your cost per lead, estimate it from your cost per click and assume roughly 10% of clicks become enquiries on a decent landing page, or nearer 20% on a native lead form. The estimate will be wrong, but it is testable, and two weeks of real data replaces it.
The floors that actually bind
$5/day per ad set
Below roughly this, delivery becomes unreliable and the ad set struggles to gather enough events to leave the learning phase. It is a floor, not a recommendation.
50 optimisation events a week
This is the real threshold that matters. An ad set that cannot reach it stays learning limited and performs less predictably — which at a $20 cost per result means about $1,000 a week to fully exit.
One ad set, not five
Concentrating a small budget in a single ad set is almost always better than dividing it. Five ad sets on $10 a day each is five failures; one on $50 has a chance.
Start smaller than you think, for a specific reason
The first month is not really about leads. It is about finding out what your actual cost per lead is, which no benchmark can tell you.
Spend enough to get a real answer — a few hundred dollars is usually sufficient — and treat it as the cost of information rather than a failed campaign. Once you know your true cost per lead and close rate, the budget calculation above stops being guesswork.
This is also why judging a campaign in its first week is a reliable way to kill something that would have worked. The early period is the most expensive and least representative part of any campaign's life.
When to spend more
The clearest signal to increase budget is simple: you are getting leads at a cost below what a customer is worth, and you can handle more of them.
Raise gradually rather than doubling. Large budget changes restart the learning phase, so a campaign whose budget jumps sharply often gets temporarily worse — which people then misread as evidence that scaling does not work.
The ceiling is not a number Meta imposes. It is the point where cost per lead rises past what a customer is worth, or where you cannot serve the enquiries you are already generating.
Questions
Is $5 a day enough for Facebook ads?[+][-]
It is enough to deliver, and it is the practical minimum. It is rarely enough to reach the 50 weekly events an ad set needs to leave the learning phase, so expect less predictable results.
How much do small businesses typically spend?[+][-]
Widely, and typical figures are not useful — a business needing three customers a month and one needing thirty have nothing in common. Work from your own numbers instead.
Should I increase budget slowly or all at once?[+][-]
Gradually. Large changes restart the learning phase, so a sharp increase often makes performance temporarily worse before it improves.
What if I can't afford enough for it to work?[+][-]
Narrow rather than thin. A smaller geography or a single service concentrates the same money into a campaign that can actually deliver, instead of spreading it below the level where anything works.
Keep reading
EXOD does this part for you.
It plans, writes and launches the campaigns, then watches these numbers every day and adjusts them — inside your own ad account.
Start free