Quick answer: how much should you spend on AI ads?
Work it out from three numbers. Your break-even cost per purchase sets the most you can pay per order. The learning floor, about 50 purchases a week per ad set on Meta, sets the minimum that lets the AI learn. Start one campaign at that floor, then scale by 15 to 20% a week while profit holds.
The three numbers that set your budget
Most budget advice starts with a percentage of revenue. That works for planning, but AI ads add a second question: is the budget big enough for the platform to learn? For the full operating plan, see our complete guide to AI ads.
1. Break-even cost per purchase
This is your margin per order before ad cost: price minus product cost, shipping, payment fees, GST and expected returns. Pay more than this for an order and you lose money on it. Our break-even calculator works it out.
2. Expected cost per purchase
Use your own recent data, or, if you have none, run a small test for two weeks. Industry averages are a poor guide because offers and prices differ so much.
3. The learning floor
Meta says an ad set needs about 50 results in the week after its last significant edit to leave the learning phase. Multiply 50 by your expected cost per purchase to get a weekly floor for each ad set.
A worked budget calculation
Worked example: a skincare brand
Margin and break-even
Average order ₹1,200. Product cost ₹300, shipping ₹90, payment fees and GST ₹210, returns ₹120. Margin before ad cost is ₹480, so the break-even cost per purchase is ₹480.
The learning floor
Recent cost per purchase is ₹400, comfortably under ₹480.
Weekly learning floor = 50 × cost per purchase
50 × ₹400 = ₹20,000 a week per ad set, about ₹86,000 a month (₹20,000 × 4.3).
With ₹1 lakh a month for Meta, this brand can run one purchase-optimised ad set properly, not five. With ₹3 lakh, about three. Spreading ₹1 lakh across five ad sets would leave each one with about 10 purchases a week, stuck in learning.
When the floor is too high
If your budget cannot reach 50 purchases a week, you have three options: run a single campaign with one ad set, optimise for an earlier event such as add to cart or checkout started, or improve the offer so cost per purchase falls.
What different budget levels can support
The table uses a cost per purchase of ₹400 to show how far a monthly Meta budget goes. Swap in your own number.
| Monthly Meta budget | Purchases a week at ₹400 | What it supports |
|---|---|---|
| ₹50,000 | About 29 | One ad set; consider optimising for add to cart |
| ₹1 lakh | About 58 | One purchase-optimised ad set |
| ₹3 lakh | About 174 | Prospecting, a customer-capped sales campaign and retargeting |
| ₹10 lakh | About 581 | Several campaigns plus steady creative testing |
Splitting between Meta and Google
Start Google with brand and top product searches, which are usually cheap and high-intent, then add Shopping or Performance Max once tracking and the product feed are ready. Many stores end up with roughly a quarter to a third of budget on Google, but let your cost per new customer decide. For cost ranges, try our Google Ads cost calculator and Meta ads cost calculator.
How daily budgets really spend
Google treats your daily budget as an average. A campaign might spend up to twice the average daily budget on a given day, but over a month you will not be charged more than 30.4 times it. So a ₹2,000 daily budget can spend up to ₹4,000 on a strong day, with the month capped at about ₹60,800.
Meta
Meta also lets delivery vary day to day around your daily budget to catch better opportunities. Judge spend by the week, not by a single day.
What this means in practice
Do not panic at one high-spend day or cut the budget the next morning. Each cut is a change the AI has to adjust to. Look at weekly spend and weekly results. The same applies to results: one cheap day or one expensive day says little; a week of data says much more about where the campaign is heading.
Decide the most you will spend this month and divide by about 30 for the daily budget. That keeps the platforms' daily swings inside a number you are comfortable with.
How fast to scale a winning campaign
- Confirm the win on profit. Cost per purchase under break-even for at least two weeks, judged on completed orders.
- Raise budget by 15 to 20% on the winning campaign, once a week.
- Add creative as you scale. More budget reaches more people, so give the AI new angles to show them.
- Watch cost per new customer. If it rises above break-even for two weeks, hold or step back.
- Avoid big jumps. Doubling a budget overnight is a significant edit and can restart learning.
For larger accounts, our guide for brands spending ₹10 lakh+ a month covers structure and team rules.
Setting a test budget for a new product or market
A test is about learning cheaply, not about profit in week one.
Size it to a decision
Decide in advance what result would make you continue, for example cost per purchase under ₹500 after 30 purchases. Budget enough to reach that number of purchases at a realistic cost: 30 × ₹600 = ₹18,000 if you expect early costs to be higher.
Keep the test clean
One campaign, broad targeting, three to five different creative angles, no edits for seven days. Then decide: stop, fix the offer, or move to the learning floor.
Budget for creative too
A test with one weak creative tells you about the creative, not the product. Set aside a small creative budget alongside the media budget, enough for three to five genuinely different ads, whether you make them in-house, with creators or with AI tools. Cheap creative variety is often the best money you spend in a test.
Use free tools first
Check the maths with our ROAS calculator before spending, so you know which ROAS you need to break even. Our comparison of AI ads and manual campaigns shows how to run a fair test.
What most budget advice tends to get wrong
It gives one number for everyone
"Spend ₹1,000 a day" ignores your margin and your cost per purchase. Start from your own numbers.
It spreads small budgets thin
Five small ad sets learn slower than one larger one. Consolidate until each ad set can reach the learning floor.
It judges by platform ROAS alone
Budget decisions should follow profit after ad spend on completed orders, not the platform's attributed ROAS.
Meta's delivery varies daily around your budget. Use weekly totals for every budget decision.
Read next in this series
Key takeaways
- Your break-even cost per purchase equals your margin per order before ad cost.
- The learning floor is roughly 50 conversions a week per ad set, times your cost per conversion.
- If you cannot afford the floor for purchases, optimise for an earlier event or run fewer ad sets.
- Google may spend up to twice your average daily budget on a given day, but no more than 30.4 times it in a month.
- Scale in steps of about 15 to 20% a week, and only while profit after ad spend holds.
Ai ads: frequently asked questions
What is a good starting budget for AI ads?
How do I work out my break-even cost per purchase?
Can Google spend more than my daily budget?
How fast can I increase my ad budget?
Should I optimise for purchases or add to cart on a small budget?
How should I split budget between Meta and Google?
Sources
- Google Ads Help: About overdelivery and your average daily budget
- Meta Business Help: About the learning phase
- Google Ads Help: About Performance Max campaigns
Let ScaleWin read your ad data every day
ScaleWin checks your Meta and Google Ads daily, judges every ad on profit and tells you exactly what to change.
Start free, no card



