How much should your Meta Ads budget be? A realistic framework

It is the question every agency hears first: “How much should I put into Meta Ads each month?” The honest answer is that the question is slightly misframed. A budget is not a number you pick and then hope to justify; it is a consequence of the result you want and what that result costs to buy.

Total spend is the wrong headline number

Two businesses spending ₺40,000 a month can be in opposite situations: one buys orders at ₺180 against a ₺500 margin, the other at ₺600. The monthly figure tells you nothing about which is which.

What governs the decision is cost per result — per purchase, per qualified lead, per booked call. Once you know what a result is worth to you and what it currently costs, the budget stops being a matter of taste and becomes arithmetic:

1. Work out the most you can pay for a result and stay profitable.
2. Multiply it by the number of results a campaign needs each week to learn.
3. That product is the campaign’s weekly floor — not its ambition, its floor.

Your margin sets the ceiling on cost per result

Start from money you already have on the books, not from a benchmark you read somewhere. Take your average order value, subtract every variable cost — cost of goods, shipping, packaging, payment processing, your realistic return rate, marketplace commission if any — and what remains is contribution per order. That figure is your break-even cost per acquisition. Spend more than it and each order loses money, however good the campaign looks in the dashboard.

A worked hypothetical, with invented numbers purely to show the mechanics:

• Average order value: ₺1,200
• Variable costs: ₺780
• Contribution per order: ₺420 — this is break-even CPA
• Target CPA at, say, 60% of break-even: ₺250, leaving ₺170 per order

Target CPA is a business decision, not a platform setting. Push it close to break-even and you buy volume with no cushion for returns; set it far below and you may starve the campaign of delivery. Either way the number comes from your own P&L, which is why “good CPA” benchmarks quoted for a whole industry are close to useless.

Cost per result and total spend move independently

Raising the budget does not lower CPA. As spend grows you reach colder audiences and CPA usually drifts up — that is normal, not a failure. The question when scaling is never “why did CPA rise”, but “did total contribution rise even though CPA rose?”

Why a budget that is too small never finishes learning

Meta’s delivery system optimises by observing which impressions turn into the event you asked for. Until it has seen enough of them, delivery is volatile and reported costs swing week to week. That period is the learning phase, and Meta’s own guidance points at roughly 50 optimisation events per ad set per week as the volume where performance stabilises.

An underfunded campaign is not a cheap campaign, it is a slow one — and while it crawls, the money it spends does not turn into information. You pay full price for impressions and still cannot tell what went wrong.

The threshold is per ad set, not per account

This is where most small budgets are lost. Split ₺9,000 a month across three ad sets and you get three campaigns that each learn badly, plus three sets of noisy numbers you cannot compare. Behind one ad set, the same money may clear the threshold.

If you cannot fund the floor, the fix is fewer campaigns, not smaller ones. Consolidate audiences, let broad targeting do the segmenting, and accept that you are testing one thing properly instead of three things badly.

When you cannot fund purchase optimisation

Purchases are expensive to optimise for because they are rare. If 50 a week is out of reach, optimise for a cheaper event that still correlates with revenue — add to cart, initiate checkout, a qualified lead form — while you keep watching purchases as the business KPI. It is a compromise, because the event you optimise for is the behaviour you get more of, so move up once volume allows.

Sizing a first test budget from data you already have

You do not need a new campaign to estimate this. Use your existing sales.

1. Average order value. Last 90 days: revenue divided by order count, excluding VAT and shipping so it matches what you keep.
2. Contribution per order. Subtract every variable cost listed above. This is your break-even CPA.
3. Target CPA. Decide what share of that contribution you will hand to media.
4. Weekly floor. Multiply target CPA by the weekly event volume you are optimising for. Continuing the hypothetical: ₺250 × 50 = ₺12,500 a week, roughly ₺50,000 a month, for one purchase-optimised ad set.
5. Sanity check the daily budget. A common rule of thumb is at least three to five times target CPA per day, so the ad set produces more than one result a day.

If step 4 lands beyond what you can commit, that is information, not a dead end: optimise for a cheaper event, raise average order value with bundles, or start with one offer instead of the whole catalogue. Whatever you choose, give it a runway — a budget that runs ten days and is then switched off has bought you almost nothing.

Three things to fix before you raise the budget

More spend does not repair a broken system; it enlarges the loss. Before you increase anything:

• Measurement. Are the Pixel and Conversions API both firing, and are events properly deduplicated? Incomplete measurement sends the algorithm after the wrong people, and leaves you unsure whether budget or tracking failed.
• Landing page. If ads get clicks but the page does not convert, the problem is not the budget. Traffic costs the same either way, so a conversion-rate improvement buys the same effect as a budget increase, for free.
• Creative variety. One creative does not scale. As frequency climbs, performance falls, and budget does not compensate for fatigue.

How to raise the budget without resetting learning

When you find a winner, the instinct is to double it overnight. Large edits typically push the ad set back into learning, and performance often gets worse before it gets better. The controlled approach is to raise budgets in modest increments and let the campaign restabilise after each step. When vertical scaling — more budget on the same campaign — stops responding, move sideways: new audiences, new creative angles, new placements, new offers.

The order of operations

The answer to the budget question is a sequence, not a figure:

1. Verify measurement.
2. Improve conversion rate.
3. Fund the floor that lets learning finish.
4. Only then scale, in controlled steps.

Any budget increase that skips this order produces cost, not learning.

Want a second opinion on your numbers?

We can review your ad account, measurement setup and margin structure together, then send you in writing where your break-even CPA sits and what a realistic floor for your next test looks like. See how we work on the Meta side on our Meta Ads service page, or get in touch for a free audit. It is non-binding, and you do not need to hand over full access to your account for us to review it.

Frequently asked questions

What is the minimum budget for Meta Ads?

There is no universal minimum, because the floor is set by your own cost per result: multiply the CPA you can afford by the weekly conversion volume an ad set needs to stabilise. If that is unaffordable, optimise for a cheaper event or cut the number of ad sets rather than shrinking every budget.

Should I spread the budget across several campaigns?

Usually not, at small budgets. Every ad set needs its own conversion volume to exit the learning phase, so splitting spend multiplies the number of campaigns that never get there. Consolidate first, separate later.

How fast can I increase the budget?

Gradually, with a pause after each step, since large edits tend to send the ad set back into learning. Raise, let costs settle over several days, then raise again — and when vertical scaling stalls, add new audiences and creative instead of pushing the same ad set harder.

Why does my cost per result swing so much week to week?

At low conversion volume, a single unusually large or cheap result moves the weekly average dramatically. That is noise, not a trend. Judge performance over a longer window, or wait until enough conversions have accumulated to compare periods honestly.