There are two ways new advertisers tend to get hurt in the first month, and they pull in opposite directions. The first is spending too much before anything is proven: pouring money into a campaign on the hope it will work, then watching it disappear with little to show. The second is expecting results too fast: switching things off after a week because the numbers look disappointing, when in truth the campaign never had a chance to settle. Both are common, both are painful, and both are almost entirely avoidable. The fix is not a bigger budget or a cleverer ad. It is a plan made up front for three things: how much you will spend, what you expect to get back, and how long you will wait before you decide.
The plan, made up front
Three numbers to decide before you spend
How much you can afford to lose while learning, what a healthy result actually looks like, and how long to wait before you judge it. Work through them in order.
Big enough to learn from, small enough to lose
Number one · Budget
How much to start with
A good starting budget lives in the tension between two rules that argue with each other. The first rule is comfort: start with an amount you could afford to lose entirely and still sleep at night, because early on you are paying for learning, not results. Treat the first month as tuition. The second rule is meaningfulness: your budget has to be big enough to gather real data. If your spend only produces one conversion a fortnight, you will never learn anything useful, because two or three results tell you nothing about whether a campaign works. You want to be generating a handful of conversions a week, not a trickle.
So the honest answer to "how much should I start with" is not a single magic number. It is a calculation. Work back from your cost per result and how many results you need before the numbers mean something. If it typically costs you around $40 to get an enquiry, and you want roughly 10 enquiries a week to judge anything with confidence, you are looking at about $400 a week, or $1,600 a month, as a floor for that channel. If that is more than you can comfortably lose while learning, you have three choices: narrow your targeting so each dollar works harder, pick a cheaper channel, or accept that learning will take longer. What you should not do is spread a tiny budget so thin that it never gathers enough evidence to teach you anything.
Work back from what one customer is worth
Number two · Cost per result
Working back from what a customer is worth
Every sensible budget starts with one question: what is a customer actually worth to you? Not their sale price, but the profit you keep after costs. Once you know that, the rest of the maths falls into place, because it tells you the most you can afford to pay to win one.
The chain works like this. If a customer is worth $X in profit, and you convert Y% of the clicks that land on your site, then you can afford roughly $Z per click before the campaign stops making money. Turn that around and you get your acceptable cost per result, often called CPA (cost per acquisition): the most you are willing to pay for one enquiry, sale, or booking. For a store, the same idea is usually expressed as a target ROAS (return on ad spend): the revenue you need back for every dollar spent to stay profitable.
You do not need this to be precise to the cent. A rough, honest version of these figures is enough to set a ceiling, and a ceiling is what stops a campaign bleeding money without you noticing.
- Customer value, the profit you keep after costs, not the sale price
- Conversion rate, the share of clicks that become a result
- Acceptable cost per result (CPA), or a target ROAS for a store
The 20x screenshot is the exception, dressed up as the rule
Number three · Expectations
The "20x ROAS" trap
Somewhere online there is a screenshot of a campaign returning $20 for every $1 spent, and it has quietly ruined a lot of first months. Those screenshots are real, but they are the exception dressed up as the rule. They tend to come from mature campaigns with years of data behind them, an established brand people already search for, or a one-off product with unusually fat margins. Held up as a baseline, they make a perfectly healthy result look like a failure.
Most healthy small-business campaigns look far more modest, and that is especially true early on. A new campaign that returns $2 to $4 for every $1 spent, or brings enquiries in at or under your acceptable cost per result, is doing its job. The eye-watering numbers, when they come at all, arrive later, once the platform has learned who your buyers are and you have trimmed away the spend that was not working. Judge your first month against your own maths (the CPA and ROAS you worked out above), not against someone else's highlight reel.
Two to four weeks before the algorithm settles
Number three, continued · Timeline
Timelines and the learning phase
Ad platforms do not perform at their best on day one, and this is by design, not a fault. When you launch a campaign, the platform's algorithm knows almost nothing about who responds to your ads. It spends the first stretch experimenting: showing your ads to different people, watching who converts, and gradually steering spend toward the audiences that work. This is called the learning phase, and it typically takes two to four weeks and a steady flow of conversions to settle down.
The most common self-inflicted wound is interfering with that process. Every time you make a significant change (a new budget, a rewritten ad, a different audience) or switch a campaign off and on, you can reset the learning and send the algorithm back to the start. Judging the numbers daily and reacting to every wobble does the same thing: it keeps the campaign in a permanent state of relearning, so it never gets the chance to stabilise. Early performance is genuinely noisy, and a bad Tuesday is not a trend.
So the discipline is patience with a deadline. Leave a new campaign broadly alone through the learning phase. Let it gather its conversions, resist the urge to tinker, and set a review date a few weeks out. When that date arrives, judge it properly against your cost per result and target ROAS. If it is working, scale it, but scale gradually: raising a budget by 20% or so at a time lets the platform absorb the change, whereas doubling spend overnight can throw it straight back into learning and undo the very performance you were trying to grow.
A worked example
The same sale, at two different click prices
Watch what a single number, the cost per click, does to whether a campaign makes money. Everything else stays the same.
Say the average order in your shop earns you $60 in profit. Your website turns 3% of visitors into buyers, so on average it takes about 33 clicks to make one sale. If each click costs you $1.20, that sale cost roughly $40 in ad spend to win, against $60 of profit. That works: your cost per result is $40, comfortably under the $60 the customer is worth, and there is margin to spare. If clicks crept up to $2.20 each, that same sale would cost about $73 to win, more than the $60 it earns, and the campaign would quietly lose money on every order. Same ad, same product; the only thing that changed was the cost per result, which is exactly why it is the number to watch.
The first six weeks
What each stage looks like
A new campaign moves through four rough stages. Knowing which one you are in tells you whether to leave it alone or act.
| Stage | Roughly when | What to expect |
|---|---|---|
| Learning | Weeks 1 to 2 | Noisy, uneven results. Cost per result often high. Leave it alone; do not judge yet. |
| Stabilising | Weeks 3 to 4 | Performance begins to settle as the algorithm finds your buyers. Make only small, deliberate changes. |
| Judge and decide | Weeks 4 to 6 | Enough data to compare against your acceptable cost per result and target ROAS. Keep, adjust, or stop. |
| Scale | Once cost per result is proven | Raise budget gradually (around 20% at a time) so the platform absorbs the change without relearning. |
Before you switch on spend
Write down three things
Decide them up front, in writing, and the two common first-month mistakes largely take care of themselves.
-
1
Set the budget you can afford to lose.
An amount you could lose entirely while learning and still sleep at night, but big enough to gather a handful of conversions a week.
-
2
Set the acceptable cost per result.
The most you will pay for one enquiry, sale, or booking, worked back from what a customer is worth. For a store, set it as a target ROAS instead.
-
3
Set the review date.
A few weeks out, past the learning phase. Hold your nerve until then, and judge on evidence rather than a bad Tuesday.
Common questions
Frequently asked
How much should I start with?
There is no single magic number. Work back from your cost per result and how many results you need before the numbers mean something. If it costs you around $40 to get an enquiry and you want roughly 10 a week to judge with confidence, that is about $400 a week, or $1,600 a month, as a floor for that channel. Start with an amount you could afford to lose while learning, but big enough to gather a handful of conversions a week.
How long before I see results?
Ad platforms go through a learning phase that typically takes two to four weeks and a steady flow of conversions to settle. Weeks 1 to 2 are noisy and often expensive, weeks 3 to 4 begin to stabilise, and weeks 4 to 6 give you enough data to judge properly. Leave the campaign broadly alone through the learning phase and set a review date a few weeks out.
What is a good ROAS or cost per result to aim for?
Judge it against your own maths, not someone else's screenshot. A new campaign that returns $2 to $4 for every $1 spent, or brings enquiries in at or under your acceptable cost per result, is doing its job. The eye-watering 20x returns you see online come from mature campaigns, established brands, or unusually fat margins, and are the exception dressed up as the rule.
Should I change things if it is not working after a week?
No. A week is inside the learning phase, when results are genuinely noisy and a bad Tuesday is not a trend. Every significant change, or switching a campaign off and on, can reset the learning and send the algorithm back to the start, so reacting to every wobble keeps the campaign in a permanent state of relearning. Leave it broadly alone until your review date, then judge it against your cost per result and target ROAS.
Should I pay an agency?
You can run a first channel yourself if you plan the three numbers up front: the budget you can afford to lose, the cost per result you will accept, and the review date. An agency can be worth it once you are ready to run several channels, or when your time is better spent elsewhere, but it does not remove the need for these numbers. Whoever runs the campaign should be judging it against the same cost per result and target ROAS you set.
Before you spend a dollar, write down three things: the budget you can afford to lose while learning, the cost per result you will accept, and the date you will review it. This is where a clear report earns its keep. Clearly keeps your cost per result and ROAS visible against the target you set, so when the review date comes you are judging on evidence rather than nerves. The next chapter is about exactly that: measuring success, and knowing which numbers actually tell you whether it is working.