The short answer: your ad budget is the number of new customers you want each month, divided by the share of visitors who become customers, times what a click costs. If you want 10 new customers, 2% of visitors become customers and a click costs €1.50, you need 500 clicks and a budget of €750. The rest of this article is about where those numbers come from and when to spend more.
Most businesses I meet set their budget in one of two ways: an amount that felt reasonable when the account was opened (€500 a month is common, purely because it's a round number), or whatever the previous agency suggested. Neither has anything to do with what the business actually wants to achieve. The model below starts with the goal instead.
What numbers do you need to work out an ad budget?
1. How many new customers do you want each month?
Start here, not with the money: 10 new customers, 25 quote requests, 40 orders. If you're unsure, start from what you can handle: 50 leads are pointless if you only have time to call 15. For the example, I'll use 10 new customers a month.
2. What is a customer worth to you?
Don't stop at the first purchase: count what a customer is worth over time. Say the average order is €350 and a typical customer buys three times a year: €1,050 in revenue. With a 40% gross margin that's €420 in gross profit per customer, and that's the number that decides what you can afford to pay for ads. If you sell services, use the value of an average contract.
3. What percentage of visitors become customers?
That's your conversion rate: how many visitors out of a hundred buy something or send an enquiry. If you already have traffic and tracking, Google Analytics will show you. Otherwise, use these rough rules of thumb: 1 to 3% for ecommerce, 3 to 5% for contact forms on search traffic with buying intent, under 1% for cold traffic from social media. I'm using 2% in the example.
4. What does a click cost?
That depends on your industry and how much competition there is. In my accounts a click on Google Search usually costs between €0.50 and €4, with some industries (law, finance, parts of B2B services) well above that. Google's Keyword Planner gives you an estimate for your keywords before you've spent a cent. I'm using €1.50 per click in the example (there's more in my guide on how much Google Ads costs).
Putting it together
The rest is simple arithmetic:
- 10 customers divided by a 2% conversion rate = 500 clicks needed per month
- 500 clicks times €1.50 = €750 monthly budget
- €750 divided by 10 customers = €75 per new customer
Compare the last line with what the customer is worth: €75 in ad spend against €420 in gross profit. That's a good deal, even before the customer sticks around for more than a year.
Now play with the numbers. If the site only converts at 1%, you need 1,000 clicks and €1,500, and a customer costs €150. If clicks cost €4 and you convert at 1%, you end up at €4,000 a month and €400 per customer. That's close to your limit, and the fix is the landing page or the choice of channel, not the budget. The model only covers ad spend, the money that goes to Google or Meta. You can see what the setup costs in the pricing calculator.
How much can you afford to pay for a customer?
Break-even is simple: a customer can cost you no more than the gross profit they bring in, €420 in the example. Pay more and you lose money on every new customer, however good the report looks.
But break-even is a limit, not a target. Set a target cost per customer that leaves room for profit and for assumptions that turn out to be wrong. My usual rule of thumb is about a third of the gross profit, so around €140 here. Below that, you can spend more with a clear conscience. Between your target and break-even you're still making money, but with no room for error.
One important point: if you measure leads rather than closed deals, you need to adjust the numbers. If you get 10 quote requests and close 3 of them, the cost per customer is roughly three times the cost per lead. People forget this surprisingly often.
How big should your test budget be, and how should you split it between channels?
You don't know your numbers until you've measured them
Conversion rate and click price are assumptions until you have your own data. The first month is a test month, and its job is to give you real numbers, not profit. I want to see at least 20 to 30 conversions before I trust a conversion rate. At 2% that means 1,000 to 1,500 clicks, so the test month may need a budget 50 to 100% higher than the model suggests, or you spread it over two months. What doesn't work is spending €180, getting 120 clicks and one conversion, and concluding that “Google Ads doesn't work for us.”
Start where the demand already is
When the budget is limited, it should go where people are already looking for what you sell. For most companies that is search: someone types “emergency plumber Dublin” and is ready to act. Google Ads on the search network is where the model is easiest to make work, because the conversion rate is highest and you only pay when someone clicks. Microsoft Ads can then pick up the same keywords at a lower price, with smaller volume.
Meta, LinkedIn and display create demand rather than capture it. The conversion rate is lower and the journey longer, so they need more budget and patience before the model adds up. For a business starting from scratch, I often put 70 to 80% into search until that channel is saturated (you're already buying every relevant search), and only then expand into social.
When should you increase the budget?
Not “when it feels right” or “when we have money left over.” You increase it when your cost per customer is below target and you're leaving demand on the table. In Google Ads you can see the second part in a metric called search lost impression share (budget): if it's above zero, there are searches you're missing because the money runs out, and every extra euro is likely to be as profitable as the ones you're already spending.
Increase in steps of 20 to 30%, and wait two to three weeks between steps. Bigger jumps throw the algorithms off, and you can't tell what happened. The same goes the other way: if the cost per customer is above break-even two months in a row, don't raise the budget to “give it a chance.” Cut back and fix the landing page, the keywords or the offer.
Why is it all guesswork without conversion tracking?
Now for the caveat. The whole model rests on a number you only get through measurement: how many of the clicks became customers. If you have no conversion tracking, or it counts every reload of the thank-you page as a purchase, the calculation is worthless. You'll probably end up spending more where you should have cut back.
The minimum before you start: a conversion that matches a real business event (a purchase, a submitted form, a booked meeting) tracked in Google Ads and Google Analytics, ideally with a value attached. If you sell by phone, you need call tracking too, otherwise half your customers vanish from your reports. And some of the gap is simply down to cookie consent. I explain why, and what you can do about it, in my article on Consent Mode and server-side tracking. If you don't know where your account stands today, take my quick analysis: five questions, two minutes.
What to do next
- Write down your monthly customer goal and what a customer is worth over a year in gross profit.
- Get a click price for your most important keywords from Google's Keyword Planner and pick a conversion rate using the rules of thumb above.
- Calculate the budget: customers divided by conversion rate times click price. Compare the cost per customer with your target cost and break-even.
- Check your conversion tracking before you spend a single euro. Make a test purchase or send a test form and check that it shows up in Google Ads.
- Run a test month with a budget big enough for at least 20 conversions, then replace the assumptions with your real numbers and recalculate.
The model isn't exact, and that isn't the point. The point is that you can answer the question “why are we spending €750 a month?” with something other than “it felt reasonable.”
