Alevli framework
The spend floor
The minimum monthly budget below which a campaign cannot gather enough conversion data to optimise, calculated as the bidding algorithm's learning threshold multiplied by a realistic cost per lead rather than guessed.
The usual way a small business sets an advertising budget is to pick a number that feels survivable, €400 or €500 a month, and ask the agency to make it work. The problem is not that small budgets buy few leads. The problem is that smart bidding is a learning system: fed too few conversions, it does not produce proportionally smaller results, it produces noise. Below a certain line, the campaign is not a smaller version of a working campaign. It is a different, broken thing.
The spend floor
Cost per lead
Learning threshold
The floor
Under it?
Use it when
- A prospect proposes a budget chosen by comfort instead of arithmetic
- Deciding whether an account should use smart bidding at all
- A campaign has run for months and "never seems to get going"
Do not use it when
- Lead quality is not yet measured; a floor computed on junk conversions optimises the account toward junk
- Brand or reach campaigns, which are not conversion-fed and have no learning threshold to feed
The steps
- Establish a realistic cost per lead. From the account's own history if it exists; otherwise from auction CPCs and a conservative conversion rate. Never from the client's hope.
- Multiply by the learning threshold. Google's guidance for target CPA bidding has long pointed to roughly 30 conversions per 30 days as the level where the algorithm has signal to work with. Treat 30 as the planning figure and verify against the account's own stability.
- Compare the floor to the proposed budget. Floor = 30 × cost per lead. A €35 lead makes the floor €1,050 a month. A proposed €500 is not "half the results". It is below the line where results become predictable at all.
- If the budget is under the floor, shrink the scope, not the standard. One service, one campaign, tightest match types. This is not a consolation prize. It is the same arithmetic run again: a tighter scope lowers the cost per lead, and since the floor is 30 × cost per lead, a cheaper lead lowers the floor itself. If even the shrunken floor is out of reach, switch to manual CPC or to organic and Google Business Profile first. "Spend €500 on smart bidding anyway" is the one recommendation the arithmetic forbids.
Worked example
A physiotherapy practice proposing €500/month, illustrative numbers:
| Input | Value |
|---|---|
| Realistic CPC (auction data) | €1.75 |
| Conservative conversion rate | 5% |
| Implied cost per lead | €35 |
| Learning threshold | 30 conversions / month |
| Spend floor, full scope | €1,050 / month |
| Cost per lead, tightened scope (one service, exact match) | €22 |
| Spend floor, tightened scope | €660 / month |
| Proposed budget | €500 → ~14 conversions |
At €500 the campaign would feed the algorithm about 14 conversions a month, less than half the threshold, so every "optimisation" would be a reaction to statistical noise. Run the arithmetic again at a tighter scope: one service, one campaign, exact match only, with an expected cost per lead around €22. The floor drops to 30 × €22 = €660, close enough that the honest recommendation becomes "€660, one service" rather than "€500, everything." If the client cannot stretch to the shrunken floor, start at €500 on manual CPC and revisit smart bidding when the floor is fundable.
Why it works
The spend floor converts a taste-based negotiation into arithmetic the client can check. It also protects both sides of the relationship: the client is saved from a budget that was always going to "prove ads don't work," and the agency is saved from being judged on an account that was unwinnable at signing. A budget framework that sometimes concludes "don't advertise yet" is the only kind a sceptical owner has reason to believe. Whether the business can afford that floor in the first place is the other half of the question, and the margin-back budget works it out from the client's own prices and margins.
Who this is for
The same method, read three ways.
01
Running it
Cost per lead times thirty, against the budget on the table. Under the floor, shrink the scope and run the arithmetic again; the standard does not move, the scope does.
02
Teaching it
A clean case of a learning-system constraint wearing a budget question's clothes. It shows a room that a platform's minimum data volume is a strategic limit, not a vendor detail.
03
Setting a budget
Before you sign anything, ask what one lead is worth to you and multiply by thirty. If that number is out of reach, the honest options are a tighter scope or manual bidding, not optimism.