Peter Drucker famously said, “What gets measured, gets managed.” If your goal is to rank in the top 10 in search results – and it should be – you need to implement measures to get there and stay there.
Most of our measuring is done after the fact. It’s time consuming to test everything before implementation, and it’s easier (and more important than ever before) to adjust your strategy on the fly. But none of this means that market research is dead.
Thankfully, today’s digital world offers tools that make testing easier and more affordable than ever before.
AdWords as Market Research
Few small business owners recognize the power of PPC, not just as a marketing tool, but as a research tool. Tim Ferriss, author of The Four Hour Work Week, chose the name of his book based on a Google Adwords campaign.
- He wasn’t using AdWords to promote his book.
- He didn’t even have a website up.
The sole purpose of the campaign was to test book titles and pick the one with the highest clickthrough rate. It worked. It stayed on the New York Times bestseller list for over two years.
When it Makes Sense
- Anytime attention is a factor. Use the click-through rate to gauge how much attention your slogan, title, or catch phrase is capable of grabbing.
- To test user behavior. This can be a great way to test how users engage with different versions of an article, software tool, or video before it goes public.
Guidelines
Use a relatively popular but low competition keyword. The purpose of these experiments isn’t to drive conversions or directly make money, so there’s no sense spending more than you need to on a high competition keyword. Be sure to use the exact same keyword throughout the test for each alternative.
Think relative. The only purpose of these experiments is to find out which of your ideas is best, so absolute values aren’t the goal. Relevancy isn’t incredibly important either, since you’re only paying for the users that click on the ads. In fact, the ideas that grab attention even when relevancy is low are the ones you absolutely can’t ignore.
Quality score is a factor. You can’t completely abandon the laws of good marketing or your quality score will suffer, and the price of the ads will go up. Drop the worst ads as quickly as possible, and considering running the campaign on a completely separate AdWords account.
Surveys
Using AdWords is great because it allows you to test how users actually behave, instead of asking them how they think they would behave, which are sometimes two completely different things. Of course, there are some circumstances where a survey is the only option.
When it Makes Sense
- If user behavior can’t be measured directly. As much as we’d like it to be this way, not everything can be measured directly. Analytics can’t measure why a user chooses to leave a page, for example.
- To test against competitors. Some competitor data is available to you through online tools such as KeywordSpy, or AdGooRoo. Analyzing competitors keywords, traffic ranking, linking domains, etc. is integral part of the HubSpot marketing software system. As helpful as those tools are, they don’t provide the all-important “why”.
- Use surveys to test what users think of competitor articles, products, or tools compared to our own.
Guidelines
- Use a tool like Smartsheet to simplify the data collection process.
- Recruit the survey participants through a service like Mechanical Turk. Relying on voluntary surveys is a flawed method because these people tend to already be your customers. Mechanical Turk isn’t perfect, but it’s good enough for many scientific researchers.
- Randomly include a few “test” questions to make sure the participants paid attention to the survey. These can be questions about how the tools worked or obvious information from the articles. They can also be ridiculous questions like, “Have you ever spontaneously combusted?” Disregard any results that get the test questions wrong.
- Pay above average. It may be tempting to buy survey responses on the cheap, but even test questions are relatively easy to spot. If you aren’t getting any patterns in your data, it could be because you aren’t paying enough and users are giving you junk answers.
While this type of testing shouldn’t be used in every situation, it’s important to use it in cases where not all data can be gathered after the fact, where everything is riding on the initial release, or when the investments are big enough to warrant testing.
What are you measuring, and how are you testing? What other tools are you using?