Pay per click is an advertising laboratory.
Every new account, every adgroup, every keyword, every ad is a test. There are no guarantees when you start – which keywords will work, which ads, which search networks, and so on. Everything’s a test in the beginning. And no matter how long your account has been running, the way to improvement is always to keep testing- keep what works, throw out what doesn’t, and try to learn from that in the process.
Some of the big mistakes people make in PPC in regards to testing are:
- Not testing at all
- Not having conversion tracking in place so you can measure test results by the right metric
- Running too many tests for your budget
- Running too many tests and lowering account performance
- Not padding tests to reduce the risk of performance decreases
Before I amplify those, let me define a few acronyms that trip some people up:
- CPL is cost per lead, which is the same as cost per conversion in lead generation accounts.
- ROAS is return on ad spend, an ROI metric. ROAS = revenue / ad spend.
- ROI is return on investment. I use it generally. Technically it’s calculated differently from ROAS but I only use it to mean “your return”.
- CTR is click through rate. Percentage of ad viewers who click on your ad.
- CR is conversion rate. Percentage of ad clickers who become a lead or sale.
1. Not testing at all
Some newbies just put in some keywords (at worst just guesses off the top of their heads, not using keywords tools) and one ad and wait. It’s the rare account that’s mismanaged this way and actually produces a good ROI.
Not only does the account need to be organized in a granular way… you also need to run at least two ads per keyword set (adgroup). Thus, you’re split-testing one ad against the other. Once you have enough stats, you get rid of the less effective one and write a new ad. Every test increases ROI.
2. Not having conversion tracking in place so you can measure test results by the right metric
If you’re measuring results by CTR, you have no idea what your ROI is. Sure, using analytics, you might know overall the results you get for PPC compared to the overall spend or clicks, but you won’t be able to see WHICH ads and keywords work and which don’t.
Until you can see the ROI for each keyword and each ad, you can’t really optimize. Granular optimization leads to overall account ROI improvement. Without conversion tracking for individual keywords and ads, your PPC account is a black box, a guessing game, and you lose a lot of potential revenue.
The problem with looking at CTR only is that there’s a certain CTR sweet spot, and once you exceed it, you get unqualified clicks, possibly from people you’ve overhyped or overpromised to. Every ad produces a different CTR and CR, so unless you look at both (as well as revenue per sale), you can’t judge one ad’s ROI relative to another’s. That’s why the best metrics are CPL for leads and ROAS for sales.
3. Running too many tests for your budget
No test can be judged complete until you have significant statistics. You need a certain number of clicks in the same way that medical research requires a certain number of test patients. You can use splittester.com to judge when you’ve reached that point in CTR terms, but not in conversion terms.
At the simplest level, you need 100 clicks to judge a keyword or ad. If you run two ads per adgroup, you need 200 clicks to judge those ads. If those clicks are $1.50 each, it’ll cost you $300 to test those two ads. It’s a good idea to reach that level at least monthly, if not sooner.
So do your adgroup’s keywords provide 200 clicks per month, and are you budgeting enough for that?
Say you have a $3000 monthly ad spend, and avg CPC of $2. You can get 1500 clicks- and that means you can run 7 adgroups with 2 ads each.
If you don’t keep this in mind, you might end up without enough data to optimize even monthly- and the longer every test takes, the longer it takes to make your account more profitable.
4. Running too many tests for desired account performance
The fact is: tests produce lower ROI than optimized adgroups. So you should spend a specific portion of your overall budget on tests. Don’t test so many things at once that you destroy your overall ROI.
For example, if based on the strength of previous test results your account is producing a 500% ROAS and you want to increase that, you have to run more tests. These new tests will temporarily decrease the overall account ROAS. So think about what the tolerable lower limit to your ROAS should be during the testing period. If it’s 400%, then spend 20% of your budget on tests. You’ll probably get more than 0% return on the tests, so the overall account will stay over 400% ROAS.
5. Not padding tests to reduce the risk of performance decreases
If you have an ad producing 1000% ROAS, and you test another one against it- what if it produces only 200%- that brings the entire adgroup down to 600% ROAS. A 40% drop in ROI. Very bad.
But if you duplicate the good ad twice, you have three copies of it running against one new ad. Then your ROAS overall would only be 800%, only a 20% drop. It takes longer to get the clicks needed to see how the new ad performs, but it keeps your ROI up.
Brian Carter is the Director of Search Engine Marketing for Fuel Interactive, an interactive marketing agency in Myrtle Beach, South Carolina. He is responsible for the SEO, PPC, SMM, and ORM programs at Fuel and its partner traditional agency Brandon Advertising & PR.