{"id":26614,"date":"2021-11-09T10:00:14","date_gmt":"2021-11-09T10:00:14","guid":{"rendered":"https:\/\/www.intercom.com\/blog\/?p=26614"},"modified":"2021-11-17T21:27:29","modified_gmt":"2021-11-17T21:27:29","slug":"proactive-support-experiment","status":"publish","type":"post","link":"https:\/\/www.intercom.com\/blog\/proactive-support-experiment\/","title":{"rendered":"The experiment that reveals how proactive support directly affects your bottom line"},"content":{"rendered":"

Proactive support<\/a> is a lifesaver for busy teams looking to provide exceptional support at scale.<\/p>\n

By allowing support teams to get ahead of known problems and provide help to customers before they need to reach out, proactive support has enabled major companies to scale their support<\/a> without increasing headcount, drive down inbound conversation volumes (in some cases by up to 80%<\/a>), and maintain incredible customer satisfaction ratings.<\/p>\n

What is proactive support?<\/strong>
\nProactive support is a strategic approach to anticipating and delivering on your customers\u2019 support needs, where they are, before they ask for help.<\/p><\/blockquote>\n

And while these results already speak volumes about the impact of proactive support<\/a>, we wanted to dig deeper and get a better understanding of the long-term benefits of proactive support, too.<\/p>\n

We believe that proactive support is about so much more than just reducing inbound conversation volumes: it can also directly impact activation, retention, and ultimately, your business\u2019s revenue. So we designed an experiment that would show us the long-term effects of proactive support on activation \u2013 and subsequently on ARR.<\/p>\n

Designing the experiment<\/h2>\n

To do this, we decided to A\/B test our in-product help to get a greater understanding of the impact it has on user activation.<\/p>\n

What is in-product help, you may ask? In-product help is all of the links featured throughout the product \u2013 you\u2019ll probably recognize them as being blue with an icon \u2013 that lead to extra guidance. This guidance might take the form of Product Tours, Articles, videos, Academy courses, or anything else that gives a user a little more information and context.<\/p>\n

Our in-product help links were the most clear-cut example of proactive support that we could test: they\u2019re tailored around specific challenges customers might face and are presented proactively, so customers don\u2019t need to find them.<\/p>\n

\"An

An example of our in-product help<\/p><\/div>\n

In-product help is there to give new customers a jumping off point to get them started. So why would we want to A\/B test it?<\/p>\n

\u201cWe\u2019ve found that people who use our content are approximately 2.5x more likely to activate with our visual campaign builder Series than those who don\u2019t\u201d<\/p><\/blockquote>\n

Well, there\u2019s a really strong correlation between people who engage with in-product help and people going on to activate with Intercom and become active users of the features they\u2019ve been learning about. For example, we\u2019ve found that people who use our content are approximately 2.5x more likely to activate with our visual campaign builder Series than those who don\u2019t.<\/p>\n

That\u2019s a really strong signal \u2013 but it\u2019s correlation, and we wanted to prove causation. We could just as easily hypothesize that a customer who is super keen to activate with Series and who\u2019s really engaged with the product is more likely to want to read an article or watch a video in the first place. So we needed to test it to see if the content was having a positive impact on the way that people activate with our product.<\/p>\n

Running the experiment<\/h2>\n

The experiment worked like this: for 28 days, 50% of new self-serve workspaces had no links to help content within Intercom. The other 50% had all the help links per usual, as the control. We were super strict on getting a statistically significant result, and this passed the test without any p<\/em>-hacking<\/a>.<\/p>\n

We had two major hypotheses.<\/p>\n