More Than Just Reporting: Uncovering Actionable Insights from Data

welocalize September 1, 2020

The power of data is limitless – it’s a huge area of untapped value (often enabled by AI). What’s important is for global brands to tap into that value – taking inspiration from the large tech giants – by tackling their unstructured data sets and turning them into information and assets. Experts estimate that 80-90% of any data in organizations is unstructured. The potential is real.

In this article, Welocalize Conversion and Analytics Director, Ryan Webb looks at how organizations can derive wisdom from data and shares some basic principles of data literacy.Actionable Insights from Data

Let’s be honest, many businesses have now got access to more data than they know what to do with. The result? Internal teams are drowning in their data lakes and getting lost in their data warehouses. It becomes harder and harder to take a step back and remind yourself “Why are we doing this?!”

Remember, data only has any real value in the first place if you can use it to uncover actionable insights. If you’re not careful, you’re making the mistake that Andrew Lang highlighted over 100 years ago…

“Many people use statistics as a drunken man uses lamp-posts—for support rather than illumination.”

There is a reason many businesses fall into these bad habits. It’s much easier to simply share ‘data’ in a report than it is to provide ‘insight.’ You’ve got to put some effort in to turn the data into information, then that information becomes more valuable knowledge, eventually you uncover insight and ultimately, you reach a level of wisdom!

Actionable Insights from Data

We’ve pulled together our suggestions on how to take a step back and examine your data from a different perspective. With a bit of luck, you’ll put some of these into practice and the actionable insights you uncover will lead to measurable business success.

PREPARE. Plan what you’re hoping to get out of the data.

You’re going to struggle to uncover any insight if you haven’t got a clear idea of what it is the business is trying to achieve. Think of the steps in this stage as being similar to calculate your route on a map before you set off on a journey.

  1. Make someone the data insights champion: As with many things, you’ll get much more done if you make someone ultimately responsible for it. It might be one person, or it might be several people dotted around the business. Rather than being a data genius, these people can promote a positive data culture and ensure some consistency in the approach.
  2. Ask a business question: The best way to plan your journey is to know the destination ? The problem with data analysis is that the ‘destination’ is often unclear. A great way to help with this is to ask yourself a business question; it helps you to remain focused on the objective and not get lost in the numbers.

“Asking a business question and then turning to your data for insights will consistently lead you to more success than just clicking around reports, waiting for something to pop out at you.”

Yehoshua Coren, The Analytics Ninja

To get you started with those questions, here are some examples: 

“Only part of our website is translated into French, how can I prove we should translate more?”

“Is it worth translating content in the Philippines & India where English is widely spoken?”

“Why does the Spanish version of my website perform better in Mexico than Spain?

  1. Agree on the metrics that really matter: What are the key performance indicators (KPIs) that help you to measure success? Turnaround time and linguistic accuracy are important localization metrics to monitor, but they aren’t the objective of your business. Examples of the more critical “performance metrics” are sales revenue or new business leads. These KPIs are the metrics that really matter to your business. Agree these with your stakeholders and regularly remind yourself of them.
  2. Keep things simple: Complexity confuses people. Many data projects have failed over the years because things have been made over-complicated. Trying to achieve far too much that the project becomes overwhelming, unrealistic and simply undeliverable. It takes confidence to just stay simple, but it’s usually in the simple stuff that the most powerful, reliable insights can be discovered.

“Truth is ever to be found in simplicity, and not in the multiplicity and confusion of things.” Isaac Newton

  1. Visualize the output: The visualization of data is proven to be a more effective way to communicate information, so before you start, do a quick sketch of the graph or chart you want to see. You don’t need a work of art, just a simple sketch will force you to think of the end result, rather than the data itself. To help you consider what’s possible, your reporting tool should have a gallery of great examples (Microsoft Power BI, Google Data Studio, Tableau, Amazon Quicksight). 

IDENTIFY. Focus in on the data that’s important.  

Once you’re confident in your planning, then you start digging into the data. Even with all that planning it can be overwhelming. The key thing to keep in mind is that you don’t want to be looking at ALL the data. You want to find ways to focus in on the data that is important to what you’re trying to identify.

The following principles will hopefully help you to remove all the data that is distracting and focus in on what you need.

  1. Filter: Don’t be afraid to cut out the noise and start with a smaller data set. You can always come back to the bigger picture later.
  2. Sort & Rank: Make sure you know how to sort data in your reporting tool so you can look at the performance at the extremes. Also, rank and prioritize the data. Use “weighted averages” so you remove data that appears to rank highly but isn’t actually common enough to be significant.
  3. Segment: Cut up the data into segments and then compare them. ‘Language’ is an obvious segment, but think of other ways to segment data to gain understanding. What about country, content type, content format, product, content length, human vs machine translated content. You might find at the top level that “French” seems to perform poorly, but it’s only if you drill down into another segment that you find “Marketing content” in French performs relatively well, but “Product Content” in French performs so badly it’s driving the overall score down.
  4. Trends: Snapshots of data at a particular time only tell you so much. Trends over time and comparisons of date ranges tell you so much more. If our linguistic accuracy score for German in April was 97%, that sounds pretty good. However, if you then see that for the previous 12 months, it was consistently higher than 99%, April might not be so great after all.
  5. Be sceptical: Data can be misleading. Not only that, but reporting tools are so simple to use now, it’s easy to produce a report very quickly and hence make a simple mistake when doing so. If you uncover something that looks significant, don’t just accept it, question it. Run the report again from another angle and see if it tells you the same thing.
  6. Use automation: Make use of automated insights contained within many reporting tools (increasingly being referred to as ‘no code predictive analytics’, because you don’t need to understand a programming language, the tool extracts the insight from the data for you). Machine learning and AI is incredibly powerful and it’s quite easy for these tools to spot interesting trends in the data and then highlight them. Remember, the machine will tell you something looks atypical, but you then have to interpret this correctly and dig around further to uncover the truly actionable insight.

PERSPECTIVE. Before you draw too many conclusions, take a step back and look again. 

Once you’ve started to pull out some insights from the data, it’s important to make sure you try to remain objective.

Data Analysis can take time and effort. It’s quite easy to get excited about a particular finding and get drawn in one direction as you hunt down the insight. In many cases that’s fine, but in others you may get a little blinkered and be too eager to see what you want to see. This final set of tips will help you to remain objective and make sure you can have confidence your insight is valid.

  1. Fresh pair of eyes: Don’t be afraid to ask others to take a look at the data too. It’s great to get a different set of eyes offering a different view.
  2. Correlation is not causation: If you spot that two data trends align with each other, remember that one may still not be causing the other. They might be indirectly related, or the trend may be coincidence. Just be careful how you draw conclusions and make sure you monitor the correlation over time. [You can see some amusing examples of totally unrelated correlating data here].
  3. So what?! Before you share your findings with others, ask yourself “So what?!” Imagine you discover that the Spanish language version of your website performs better in Mexico than it does in Spain. Don’t just report on that; keep digging! Rather than being a language challenge, maybe it’s because people view the website on different devices in those countries and your action becomes a device related one. If you ask yourself “so what?!”, you’re more likely to uncover actionable insights.
  4. Remember people, not just data: Quantitative Data Analysis tells you what is happening, but it doesn’t tell you why. Qualitative (behavioural) research usually helps you uncover the reasons behind the changing data. There are companies like CSA Research who’ve grown to do this on a large scale, but you can start small. You could run a small workshop with internal stakeholders, send out a short customer survey or even set up some testing with end users.
  5. Beware bias. Humans are great at recognising patterns, but we sometimes spot patterns when there are none. More specifically, when acting on your data analysis, confirmation bias can lead you to believe your change resulted in the outcome you wanted to see. The best way to be objective is to a/b test your change (test one version of your changed content against another version of the original content). Furthermore, by using a statistical significance calculator, you can ensure the difference is 100% valid.

These tips only touch the surface of data, but they should act as a reasonable framework to give you the confidence to take the plunge and start to uncover insight. If you’d like more help, many of the websites linked to within this article have training content and tutorials available. There is also a great LinkedIn learning course on Data Fluency.

The team here at Welocalize would love to talk through your data challenges and seek ways we can help you move from data to actionable insights. Get in touch with us here to find out how we can help you leverage data to increase the performance of your global digital content campaigns.