Читать книгу The Customer Education Playbook - Daniel Quick - Страница 27

If you're a large enterprise, you may have the traffic and infrastructure available to you to distribute various types of content and test out different scenarios. For example, you might randomly distribute your customers into two groups and then run an experiment where one group watches a “Getting Started” video and the other does not. Then, you could track consumption rates and adoption rates over the first 30 days and get meaningful information on a causal relationship between your content and user behavior.

However, in the vast majority of cases, this probably isn't realistic. Instead, here are two other ways you can measure the success of your content on product adoption.

Trained/Not Trained Cohort Analysis In this analysis, you simply compare the behavior of two different groups: one trained, one not. First, you'll need to define trained. To do this, create content that users need to opt into, like video content that you have to register for and watch, a guidebook that needs to be downloaded, or anything else you can actually track. You'll then compare the people who interacted with the content (“trained”) to those who didn't (“not trained”) and measure product adoption rates between the two groups. Although this isn't a causal analysis, it'll still give you a great understanding of the relationship between consuming that content and product adoption rates.

Pre/Post Content Cohort Analysis With this approach, you push the customer education content to all of your customers by, for example, creating a three-minute video that new users watch the first time they sign into your product environment. You then track the data for product adoption after this content has been viewed, measuring the difference between the metrics you had before you launched the video. If you can get the support of your business leaders to push content to all of your users, this can be a great replacement for a true A/B experiment. While the pre/post analysis is also correlative and not causal, it more or less controls for confounding variables.