July 3, 2020

What Budget Should I Set for Online Advertising?

Once you have made the plunge and decide to test online advertising, the next question you are likely to have is, “What budget should I spend on online advertising?”



Once you have made the plunge and decide to test online advertising, the next question you are likely to have is, “What budget should I spend on online advertising?” Many clients that we talk to start with extremely conservative budget numbers and are too quick to cast judgment based on how they think their campaigns should perform. To make a proper judgment call on online advertising performance, especially for first time advertisers, it is best to follow a few rules of thumb:

1). Start at the bottom of the advertising funnel and move up as you see success.
2). Do a sufficiently long test so that you can get feedback and make performance improvements over the course of the test.
3). Use your cost-per-goal (or cost-per-product) to develop your total budget.
4). Set reasonable benchmarks.
We have already written an article that outlines the first rule of thumb in detail, so I will be using this article to describe rules 2, 3, and 4.

Do a Sufficiently Long Test

Typically, when we launch a new client, we do a 3-month minimum contract, as we find that it can take up to this long to determine if we are able to find success for a client. Many of our clients are startup CEO’s or CMO’s, a group not particularly known for being patient, so a common response we get is, “What if we increased the budget significantly - could we speed up the testing process?” While a larger budget does make it possible to test things faster, we normally say that a test should ideally be 90 days or more. If you want to go faster and you have the budget for it, you could do a minimum of 60 days (or an absolute minimum of 45 days) for your test.

This is because even if you are able to get more performance data faster with a larger budget, it still takes time to interpret the results and then act on them to improve performance. If you had $1 million to spend on online ads, but only three days to spend it, there are only so many updates or tweaks you would be able to make based on the data you get back (there is typically a 12-24 hour reporting delay between when your budget is spent and when you can see and act on the results).

You might be able to make 2-3 major updates (e.g. shifting around budget, changing tactics, adjusting your creative strategy) during that time frame, but very likely, you would end up wasting a huge portion of your budget on poor performing tactics and you would end up with subpar performance. On the other hand, if you had 90 days in which to spend $1 million, you would be able to get regular feedback on what is working and make updates consistently. By the end of your test, you might have made as many as 20-30 major changes to your strategy, and as a result, you would end up with much stronger performance than in your 3-day test. This is illustrated by the chart below, which shows the CPA of a 3-day test of $1 million (where 2 optimizations were made) vs. the CPA of a 90 day test with a budget of $1 million (where close to 30 optimizations were made):

Use Your Cost-Per-Goal to Determine Your Total Budget

It is important to keep in mind that there is no single ‘correct budget’ number to use for your online ad tests, and that the number you end up with should factor in your cost-per-goal. At a minimum, we would recommend a budget of cost-per-goal * 50 (or an absolute minimum of cost-per-goal * 30). If you don’t currently have a goal or a cost-per-goal, use the cost of your product instead (e.g. if you sell a $50 product, use $50 as your cost-per-goal). So, with a $50 cost-per-goal, you ideally have a minimum test budget of $2.5k over the course of your test. If your product is $100, that’s a minimum of a $5k budget over the course of your test.

It is important to note that there is no single correct budget number to use, but rather, the larger the budget you select, the greater the likelihood that your results are accurate and not skewed by chance. For example, if you had a $100 product, ran $300 in ad spend, and made $400 in revenue, you might use this as proof that your online advertising strategy was effective. Realistically, this is such a small sample size, that if you were to try to scale out your campaign, you might quickly find that you got lucky initially and that your current strategy is significantly unprofitable. Alternatively, had you spent $30k in ad spend as a test, whatever results you got would likely be very accurate in terms of predicting how you will perform long-term.

Set Reasonable Benchmarks

Clients that we have worked with in the past have all sorts of ideas as to what performance they should expect from their online ads campaigns, some justified, some not. These expectations have ranged from a 10x return-on-ad-spend to clients expressing serious doubt as to whether online advertising would produce break-even results for them. There are a few ways that you can set benchmarks for your test:

1). Past Performance: If you have run online advertising before, and are working with a new partner, it is often best to use past performance as your benchmark and to see if you are able to beat it with this new test. Alternatively, if you are testing a new channel, it can often make sense to judge performance off the previous channel (e.g. if you have only run FB ads in the past, and want to do a Google Ads test, use your FB ads performance as your performance benchmark).
2). Business Profitability/ROI: A great benchmark to use if you have the numbers at your disposal is to use business profitability as your benchmark. For example, if your product is $50, and you know that your profit margin is 75%, then you could set your goal for the advertising test as a cost-per-sale of $37.50 or better.
3). No Goal or Benchmark (Continuous Improvement): If you don’t have good data for #1 or #2, a decent approach is to simply start your test, see where performance is after the first week, and aim to consistently lower the cost-per-goal by the end of the test. This approach often isn’t quite as effective as #1 or #2, but it’s not a terrible alternative.

One final thing to mention: if, after your test, you haven’t achieved profitability or you fell short of your goal, this doesn’t necessarily mean that you won’t see long-term success. If you are within striking distance of your goal, it is usually a sign that if given a little more time, you will hit your target. A good rule of thumb is that after your test, if you break-even in terms of profitability (or hit your goal), then you can expect to be successful long-term. If you are within 10% (or even 15%) of your goal (e.g. you wanted to hit a $100 CPA, you hit $110), the odds are still likely that you will hit your target with some extra work. If you miss your goal by more than 15%-20%, it is likely that your approach is not a winner, and that you should try a new strategy or channel to achieve your business objectives.

If you’re looking for this type of partner, Stackmatix could be your solution. From pre-seed to Series C, we aim to build integrated technology stacks that create consolidated data sets and analytics across all sales and marketing activities to maximize revenue and marketing return. Kick off an email thread at for a free growth consultation to explore how we can help you to zero in your measurement and scale your business.