In this guide · 7 sections
Short answer
Work out the most you can pay per lead or sale from your margins, then budget for enough results for Meta to learn: its help centre says ad sets usually exit learning after about 50 results in a week. If that is too much, optimise for a cheaper event, use one ad set and test for a week.
Step 1: work out what a result is worth to you
A "result" is whatever you ask Meta to optimise for: a purchase, a lead form, a WhatsApp conversation, a booking. Before choosing a budget, decide the most you can pay for one result and still make money.
- Average first order or job valueUse your own sales dataS$200
- Gross margin on that order (40%)S$80
- Leads that become paying customersYour own conversion rate1 in 4
- Value of one lead (S$80 / 4)S$20
Maximum you can pay per lead to break evenabout S$20
Round illustrative numbers. Repeat customers can justify paying more for the first sale; be honest about how often that happens.
If you do not know your lead-to-customer rate, track it for a month first with your current enquiries. Without it, you cannot tell whether a S$15 lead is cheap or a loss.
Step 2: budget for the learning phase
Meta's delivery system spends the first part of every new or heavily edited ad set exploring who to show it to. Meta's help article About the learning phase says ad sets usually exit learning after about 50 results in the week after the last significant edit, and that performance is less stable, with a higher cost per result, until then. If an ad set cannot get enough results, Ads Manager shows Learning limited.
| Your target cost per result | Weekly spend for about 50 results | Per day |
|---|---|---|
| S$5 (for example, a message or landing page view) | about S$250 | about S$36 |
| S$20 (for example, a lead form) | about S$1,000 | about S$143 |
| S$60 (for example, a purchase or booking) | about S$3,000 | about S$429 |
Most small businesses cannot fund 50 purchases a week from day one. That is fine; it just means you choose your set-up to suit your budget rather than hoping the system learns anyway.
Step 3: set the budget in Ads Manager
Meta's best practices for minimum budgets say minimums depend on your vertical, budget type, bid strategy, optimisation event, currency and schedule; Ads Manager warns you if a budget is below the minimum. A few rules from Meta's documentation help small budgets go further:
- Use one campaign with one ad set at first. Meta advises against high ad volumes: many ad sets each learn less than one combined ad set.
- Use a campaign budget if you do run several ad sets, so Meta moves money to the best-performing one. Meta notes the risk of under-delivery is lower this way.
- If you set a cost per result goal, Meta says your daily budget should be at least five times that goal. A S$20 goal needs at least S$100 a day.
- Do not cut a daily budget late in the day. Meta warns the system may already have spent more than the new amount.
Our guide to boosted posts versus Ads Manager explains why a proper campaign in Ads Manager gives you these controls.
Step 4: run a fair test
Fix tracking first
Make sure the Meta Pixel and Conversions API record your result once and correctly. A budget test with broken tracking tells you nothing.
Launch and leave it for at least 7 days
Meta's guidance is to wait until an ad set is out of learning before editing, because edits reset it. Check daily for errors, not for results.
Prepare several creatives up front
Load three to five ads at the start rather than adding new ones every few days, which can push the ad set back into learning.
Compare against your break-even cost
After the test week, compare the cost per result with the maximum from Step 1, and check how many of those leads became customers.
Make one planned change
Change budget, audience or creative, not all three, so you know what caused the difference.
When to stop, hold or scale
Signs to scale up
- Cost per result below your break-even for two weeks running
- Leads are turning into paying customers at your expected rate
- The ad set has exited learning
- Frequency is still low, so you are not just showing the same people the ad
Signs to stop or rethink
- Cost per result well above break-even after a full test week
- Plenty of leads but few become customers
- Learning limited for weeks with no change in results
- Comments or messages show the wrong people are responding
When scaling, raise budgets in planned steps and expect a short period of less stable results. Meta lists frequent budget changes among things that can send an ad set back into learning. If Google search ads are part of your plan, our guide to choosing Google Ads management in Singapore covers that side.
Do it yourself or get help?
The method above works for any budget; the hard parts are clean tracking and the discipline not to tinker in the first week.
Our Meta Ads Setup (S$590) creates the Business portfolio and ad account in your name, sets up the Pixel and Conversions API with key events tested, and builds one campaign with up to three ad sets and three to five creatives. If you want someone to run it week to week, Ads Management Starter (S$390 a month) covers one platform for ad spend up to S$3,000 a month, with weekly optimisation, two new creatives and one A/B test a month. Ad spend is paid by you directly to Meta. IT Sorted is a Singapore-based team of developers with over 10 years of experience; you pay upfront, send the brief in writing and follow progress on your order page.
Frequently asked questions
Work backwards from your margins. Decide the most you can pay per lead or sale, then multiply by the roughly 50 results a week Meta says an ad set usually needs to exit the learning phase. If that total is too high, optimise for a cheaper event such as messages or landing page views, and use a single ad set.
Meta sets minimums that vary by currency, optimisation event, bid strategy and other settings, and Ads Manager shows an alert if your budget is too low. Meta's guidance adds that with a cost per result goal, your daily budget should be at least five times that goal.
At least a week without significant edits. Meta's learning phase guidance says ad sets usually exit learning after about 50 results in the week after the last significant edit, and that results during learning are less stable. Editing during this period resets learning and delays optimisation.
Not on a small budget. Meta advises avoiding high ad volumes because each ad set learns less, and suggests combining similar ad sets. Start with one ad set and several creatives, or use a campaign budget so Meta moves money between ad sets automatically.
Either works. Daily budgets are simpler for ongoing campaigns; lifetime budgets suit fixed-date promotions. Meta notes that when lowering a lifetime budget, the new amount must be at least what you have spent so far plus 10% of the last two days' spend, so plan changes rather than making them mid-campaign.