It’s not hard to have spreadsheets and data tables to see rows and columns full of data. Throughout all of the sorting and filtering, it can be a tedious task for the end-user. In an endless matrix of data, insights and recommendations can help zone in on the most useful information.
If they are actionable, they can bring value to personal and professional goals, and over time they can drive the adoption and retention of your product or service.
This is a progressive and straightforward framework to enhance your users’ experiences with data.
Let’s get started.
It’s essential to gain trust and build credibility before reaching the end goal of building a recommendation. Data is at the core of this framework and the perfect place to start. It’s a record of information on an action that you or your users take.
Here are a few examples that I will discuss in this article:
- Example 1: Your bank account shows you a list of transactions that you’ve taken.
- Example 2: An analytics experience shows how many people visited your webpage.
- Example 3: A streaming platform displays how many times people watched your video.
If these were unreliable or inaccurate, you’d take your business elsewhere.
Insights can influence the next action a user will take based on information in the past. You can start to correlate trends based on larger data sets. This is a great opportunity to work with your engineering or data science teams to highlight what trends can be correlated.
This may look like an average of video views over a date range, the average days it takes to close a deal, or the average opens per email campaign.
One thing you’ll notice here is that we start to think about what two or more data points look like over a snapshot of time. This will allow you to compare and contrast, and identify trends in data based on what you can determine has a positive or negative impact — what’s deemed positive or negative is different per industry and problem space, so you’ll want to make sure that you have a solid understanding of this.
To elaborate on the examples in the data section, it could look like the following:
- Example 1: Your average spend is higher this month than last.
- Example 2: Average visits this week are performing higher than this time last week.
- Example 3: You are seeing a 20% increase in streams.
Notice that there isn’t a next action we are recommending to the user, or highlighting particular causation.
As with anything, you’ll want to prioritize what types of insights and trends are the most compelling because they can be overwhelming or create blindness if you expose too many insights to your users.
This is where it all comes together. Recommendations are where we make things actionable — we have a strong hypothesis of what leads to positive outcomes we’ve learned, and users have solid confidence in our predictions. If a user wants to learn more, we have breadcrumbs to show how we’ve determined the recommendation based on data and insights/trends.
An important thing to note is that this isn’t something that UX can solve alone. You’ll want to make sure that your data science/engineering team has an established analytics/modeling platform and that there are large enough sets of data that are accurate — this doesn’t work without it, and if inaccurate, it can be costly as far as trust and adoption goes.
Now we can use our data as a foundation, supported by the trends to create that story.
- Recommendation — Reduce discretionary spending this month to put additional cash into savings.
- Insight — Your spending this month is higher than last.
- Data — 2 purchases over $500 that are in discretionary spending category. (notice these are singular data sets, category, and spend)
- Recommendation — Run a marketing campaign to drive new traffic to your website.
- Insight — You haven’t run a marketing campaign in the last month.
- Data — 30% decrease website visitors are unique traffic.
- Recommendation — Posting two additional videos a week will put you on pace to 20,000 views a month.
- Insight — Average views this week are performing higher than this time last week.
- Data — Four video uploads with 1,000 views each over the last week.
These are basic examples, but you can see how we start to build a story based on past performance.
Making recommendations based on concrete data and evidence for your users lets them have something tangible, and they will be confident with it as you have based it off of actual data.
This can be a lot of information, so it’s up to you to figure out what kind of progressive disclosure model you’ll want to use.
Let’s recap on what we covered today:
- No matter how certain you are with your recommendation, people will want to see how you came to your conclusion — You must be able to back everything up with existing data and trends. You may have a complex understanding of how the data works, but the user may not. Therefore, it is important that you show them exactly how you reached your conclusion, so they can also have the same understanding of the insights and data.
- Trust is hard to gain back once it’s lost, so you need to get it right — Try testing this framework with a limited set of users and allow people to opt-in to using this method. Ask for feedback, note down your observations, and work to make this process as cohesive and successful as possible. There is very little room for error here, as the users are relying on recommendations that have been developed from correct data. Therefore, before making this framework a standard practice, take some time to figure it out and get it right.
Here are some additional resources to that provide more insight into this topic: