Using AI for Onboarding

Three key changes that increased conversion by over 10%

Onboarding customers can be a painful process and often contributes to a loss in users at the last part of the funnel.  By making three key changes we were able to increase our overall funnel conversion rate by just over 10%.  All three of these changes were around customer onboarding.

For those that are unfamiliar, onboarding is the process of familiarizing your new customer with your software and setting up their account for the first time.  This often involves filling out personal and company information along with other data.

Without further ado, here are the three changes:


1. Prune Unnecessary Data

Think about the data you’re collecting during your onboarding process.  Is all of it really necessary?  Are you using every bit of data to either maximize or customize your user’s experience?  Do you need all of it up front?

Tim Ferris is a huge proponent of ‘just in time’ data over ‘just in case’ data.  This mantra can apply to many parts of your life including your SaaS company’s setup flow.  Are you collecting data for some feature that you might decide to implement down the road, or is all of the data necessary and currently being used by your application?  If it’s the former, you should consider pruning out those input fields and just add them when it’s actually applicable.

Getting every bit of data in one fell swoop can sound appealing, but in reality it can pose a big barrier to entry for your new customer/user.  If they find themselves asking, “why does this company need this information?”, or they begin thinking to themselves, “I don’t have time to fill out all of this information right now”, you likely just lost a customer.


2. Formalize the Setup Process

Initially, we just sent the user to their ‘settings’ screen and displayed a modal pointing out what information should be filled out.  We gave the option to skip or navigate away from the setup.

Screen Shot 2017-09-02 at 9.47.51 PM.png
Old, Unsuccessful Setup Flow

The results? People navigated away or skipped.  If you give people the easy way out, they’ll take it.  Although the modal displayed ‘Step 1 of 3’, some people felt unclear how long this would take or where the “setup journey” would take them.

To address this, we created a separate page just for the onboarding setup.  The steps are now very defined, it requires some effort to skip, and there are few distractions.


Screen Shot 2017-09-02 at 9.55.58 PM.png
New, Defined Setup Flow

The whole journey is defined before the customer even starts the process, the navigation is familiar and they can go back and change information, and it’s separated into chewable, bite-sized chunks.


3. Use Artificial Intelligence to Assist

Once you’ve collected some data from the user, use that data to simplify the rest of the onboarding process.  At Cloud Campaign we noticed that customers really enjoyed the idea of triggers, but had trouble coming up with what terms to trigger upon.  For a bit of context, Cloud Campaign allows you to trigger social media posts on trends and weather.  So when I say terms to trigger upon, I mean words for the trends trigger.

To combat this, we ask for the customer’s company website.  We then grab the meta tags from their website and extract the ‘keywords’ attribute, which is traditionally used for SEO and should contain terms relevant to the company.  This is a good first step, but for the trend trigger to work, we’ll likely need more than just a few keywords.

Next, we take those keywords and look for backlinks on Wikipedia.  We make the assumption that if another Wikipedia article is linking back to this Wikipedia article, they must be related.  This produces some really cool connections that otherwise wouldn’t be made by a simple synonym lookup.  For example, our keyword recommendations for ‘rock climbing’ include ‘Alex Honnold’ and ‘Yosemite’.

If the customer didn’t supply a website or their website doesn’t contain the ‘keyword’ metatag, we recommend our most popular schedules.

Screen Shot 2017-09-02 at 10.31.04 PM.png
Using AI to Recommend Triggers


We frequently blog about our findings, successes, and failures.  Subscribe to avoid pitfalls we’ve made and learn from our positive experiences that actually move the needle.

If you have unique knowledge or an experience you’d like to share with the community, feel free to fill in your name and email below, and we can discuss the possibility of a guest post.

– Ryan

Curious about automating your company’s social media campaign? Get started with a free month trial of Cloud Campaign.


All graphics designed by FreePik

Leave a Reply

Your email address will not be published. Required fields are marked *