How much do you need to spend on LinkedIn ads to determine whether they work? Use the Plan and Peek approach to get the right budget for testing LinkedIn Ads. It’s based on solid statistics, not inaccurate and misleading rules-of-thumb like: “Test budgets should be ___ percent of your total budget.”
The Plan and Peek approach helps you avoid two very common mistakes: 1) Spending too little on your test and wrongly concluding that LinkedIn ads don’t work and 2) spending too much on campaigns that will likely never work and waiting too long to make optimizations or ending the test.
Jumping ahead, here’s the simplified version of the Plan and Peek approach for determining the right test budget:
Plan on spending about 20x your cost-per-conversion goal, but peek at the results when you’ve spent about 2.5x your cost-per-conversion goal. Then, use this calculator to decide whether to continue or abandon the test. You can find more precise figures in the charts below, but that’s the gist. No PhD in statistics required.
Avoid baseless rules-of-thumb, and, instead use Plan and Peek – an efficient and statistically-valid approach to testing LinkedIn ads.
Keep reading for details…
Plan and Peek Overview
The goal of your test budget is to reveal the “truth” about whether LinkedIn ads will work for your business over the long run. Any ad campaign can get lucky or accidently bomb in the short run. We need to gather enough data to see beyond the short-term noise and, instead, reveal the long-term truth. Just like when you judge a person, you want enough data (experience with them) to make a reliable assessment – one that’s likely true over the long run.
When early test results looks great, it’s easy to decide to keep going. When early results are disappointing, however, the decision whether to continue is harder. The ideal test budget gives you just enough data to confidently decide whether to continue buying ads or change course.
Plan and Peek gives you a worst-case scenario for a test budget (the largest you might need), but also provides valid, early check-in points. An early check-in could reveal results that are so bad that you should end your test. Alternatively, it could show that the initially-poor results could be due to bad luck, are not necessarily representative of the the long-term truth, and you should continue the test.
Follow the Plan and Peek steps below to get the right test budget for your LinkedIn ads and make valid decisions on whether they are working for your business.
1) Estimate your Cost-per-Click and Cost-per-Conversion Goal
To determine the right test budget for LinkedIn ads, you just need: 1) the average cost-per-click for your LinkedIn ads and 2) your cost-per-conversion goal.
Cost-per-Click estimate: You can base this on the maximum cost-per-click bid that you set in your LinkedIn ad campaigns. Alternatively, you could just pick $2, which will give you a reasonably, conservative suggestion for your test budget. This works well for LinkedIn Text Ads because their minimum cost-per-click is $2. The cost-per-click for Sponsored Message ads could be less than $2, but the recommended test budget doesn’t change much at lower cost-per-click estimates (e.g., $1 or $0.50). The cost-per-click for Sponsored Content ads will likely be higher (e.g., $4-$10), resulting in a smaller, recommended test budget.
Cost-per-Conversion goal: This threshold should reflect what’s profitable for your business and how you’re using LinkedIn ads. For example, a conversion could be when someone becomes a lead (gives you their contact info) or a customer (makes a purchase). When you use LinkedIn to get leads, pick a cost-per-lead goal that’s profitable, given your ability to turn leads into sales. For example, if you need to acquire customers at a cost that is <= $1,000 per new customer and you think you can convert 5% of leads into sales, you should set your cost-per-lead (conversion) goal to $50.
2) Plan for the largest test budget that you might need
Use the Plan chart below to find the largest test budget that you might need to adequately test LinkedIn ads.
If you end up spending this amount and your LinkedIn ads delivered a cost-per-conversion that is 20% higher than you need, you can confidently decide that your LinkedIn ads aren’t working for your business. Try optimizing your ads or move on.
Example of the "Plan" step
Let’s say you need LinkedIn ads to deliver a $150 cost-per-conversion and you estimate that your cost-per-click for LinkedIn ads will average $4. You should plan on spending as much as $3,117 on testing LinkedIn ads.
3) Determine when you have enough data to peek at the results
Use the Peek chart below to find the amount you need to spend on your LinkedIn ads before you check your results using this calculator.
If, after spending the recommended amount, you have zero conversions, you should consider changing your ads or abandoning LinkedIn ads.
If you have at least one conversion, keep running your ads, but check your results in the calculator each time you spend the amount in the Peek chart.
Example of the "Peek" step
Continuing our example… given our $150 cost-per-conversion goal and estimated $4 cost-per-click on LinkedIn ads, we should check our results in this calculator when we’ve spent at least $344.
Each time we spend another $344, we’ll check the results in the calculator. If the likelihood of the results is more than 10%, we’ll continue the test until we spend $3,117. At that point, we’ll check the results in the calculator and we’ll see how close we are to our cost-per-conversion goal. If the results are way off, we can confidently walk away from LinkedIn ads. Otherwise, we’ll continue testing and optimizing our ads.
If you feel that a 10% chance that the results were due to bad luck isn’t the right cutoff for deciding whether to continue testing your LinkedIn ads, pick a cutoff that your comfortable with. For example, you could decide that a 20% likelihood is the right cutoff for stopping and you’ll end your test if the probability of your results is below that. Whatever cutoff you choose, you’ll at least have an objective criterion that is based on well-grounded statistics.
The math behind the behind the Plan and Peek approach
Let’s start with an example of how someone should explain this to their team or boss:
“We are going test whether LinkedIn ads could work for our business. We know that a cost-per-lead of $150 or less is a profitable benchmark for us. Therefore, our goal is to assess whether LinkedIn ads can reliably achieve that. We’ll plan an adequately-sized test budget, but we’ll peek at the data along the way and stop early if the initial results are really bad. Here’s our Plan and Peek approach:
- First, we’ll plan for the biggest test budget that we might need (worst case) in order to be 80% confident that the cost-per-lead from LinkedIn ads is $180 or less. A $180 cutoff is 20% higher than our benchmark of $150, but that buffer allows us to run a valid test that requires less time and money. For example, if we used only a 10% buffer (i.e., tested for a $165 or lower CPL) our test budget would need to be about 4x bigger. We don’t need that kind of precision at the beginning. To start, we just want to know if LinkedIn will even be in the ballpark of the results we need. Using the 20% buffer allows us to spend a reasonable amount and learn quickly. We estimate that, at most, we’ll need to spend $3,152 on this test.
- Once we start running ads, we’ll peek at the results when we’ve spent at least $344. If the initial results are horrible, we’ll stop the test early. However, we’ll continue the test, despite bad results, if they could have been due to bad luck (i.e., >10% chance of randomly getting bad results). Each time we spend another $344, we’ll peek at the results again. We’ll keep the test going as long as the initial results have a reasonable chance of achieving $150 CPL in the long run.”
Now let’s get into the statistics behind the Plan and Peek method…
Testing LinkedIn ads to see what cost-per-conversion you can get is the same as seeing what proportion of clicks turn into conversions (leads or sales). This kind of experiment is analogous to sampling from a binomial distribution that has an unknown probability of success. Therefore, the math behind the Plan and Peek approach is based on assumptions related to binomial distribution and the Wilson Score Interval for estimating sample sizes.
Getting very specific, the assumed probability of success for the binomial distributions are set to the probability that would achieve the cost-per-conversion goal (= estimated cost-per-click / cost-per-conversion goal). The recommended budget sizes in the Plan chart are based on the Wilson Score Interval for these binomial distributions. The confidence level was set to 80% for a one-sided test and the upper bound of the confidence interval was 20% higher than the cost-per-lead goal. So, the Plan chart answers the question: What sample size do we need in order to calculate an 80% confidence interval for the “true” cost-per-conversion that is, at most, 20% higher than our goal?
The figures in the Peek chart are based on the same binomial distributions used in the Plan chart estimates. The Peek chart, however, answers a different question: If we assume that the “true” cost-per-conversion from LinkedIn is equal to our goal, what’s the smallest test we can run where there is only a 10% chance of getting zero conversions? If we see zero conversions at this point in our test, we should be worried that the “true” cost-per-conversion is higher than our goal + 20%, since there is only a 10% we’d see this result due to bad luck. And we should think about making changes to our ad campaigns or abandoning LinkedIn ads.