Where are you in your analytics journey?
The path from data to insights is rarely a straight line. Here are a few tips to help your organization from getting derailed on those first projects.
Business transformation is enabled when people can easily engage with your data, ask questions, find answers, and ultimately generate insights and recommendations that help you evolve. Deciding that you want your organization to be more data-driven is the easy part - getting there is hard work, and in my experience, never a straight line from start to finish.
Transforming data into strategic advantage is a non-linear and upending journey. Put another way, turning data into insight is a creative activity – driven by curiosity and a thirst to learn and understand. In nearly every case, what you end up with as the best solution is not where you thought you were going at the start.
Why does this happen? You have databases - you likely have technical talent on staff - you may even have tools like Tableau that help you visualize and explore your data. Isn't that all you need?
The answer is no, because humans are involved in both the generation and the consumption of your data. When you start getting into it, you're likely to find:
Metrics like revenue, leads, prospects or number of interactions have a different meaning across touch points or business units
Data may have been collected from one system for a period of time and from a second system for a later period - and the rules on how the data gets there and what it represents may be in conflict
You may have basic data quality issues, like missing records, incomplete data and corrupted data
Business objectives change all the time, meaning the rules for how certain metrics or segments are calculated may change in the middle of your efforts (and if not, they certainly will change after you first deliver your analysis)
There's a commonly shared notion that up to 80% of an analyst's time is spent just getting data ready to analyze. Think about that - most of the effort is spent just getting to the point where you can actually start developing insights.
So as a data leader, what can you do to set your team up for success?
Identify the owner of each data set and engage them up front. Secure their involvement to identify and resolve data challenges.
Your data owner is likely a businessperson who is responsible for a particular function. Keep in mind that this person may not even realize they are a data owner. For a sales team, this may be the head of sales, and for a traffic system, the production or operations manager.
Be aware that as business leaders themselves, they will often designate experts in the team that are responsible for administrating each system and process - these individuals are good resources to engage, because they are versed in how things really work and are handled in each system.
Tip: Make sure that discrepancies between business policy, directives and "how it really works" get surfaced early.
Engage by asking business questions before you start building reports.
It's tempting to jump right into the data and start cranking out dashboards. Data visualization tools make it easy to explore and surface what's in the system, but data discovery should be a step in the process, not the end result.
Start instead by asking your peers and other leaders, "what is the biggest problem you face in meeting your goals?" or "what's the one question you can't answer that would make all the difference?" Follow up with, "what data do we have that can inform this topic?" This will engage your end customers and also help you understand where to have your team start their explorations.
Tip: Identify the KPIs or key metrics that are valued by the business, and look for gaps and areas which are already acknowledged as needing improvement. This gives you a point at which to engage, and through your discussions, you will start to understand the underlying challenges that can be addressed with data and analytics.
Pick a few simple and well-understood processes to focus on first.
Trying to understand the value of each possible permutation of your customer journey may simply be too much to tackle at the start. Use your discussions on business needs and challenges to identify a good beach head and start there.
Tip: Often it's not a metric itself that is the most interesting - it's the attributes that add color to the metric that tell a story. For example, I have an age, a social security number and spent an amount of money buying groceries last week. But what music I like, my favorite foods, where I grew up, and my hobbies make me interesting.
It's the same with data. In addition to sales, look to enhance that data point with information on customer type, lifetime value, acquisition source and preference data that can reveal a pattern or story.
Map out the business process that generates each data set.
Begin with a marker and whiteboard, drawing out each process with simple boxes and arrows. It's a great way to engage the data team with how the business works. Then, your team's analysts can ask questions and start mapping what's in your systems to these diagrams.
Tip: When mapping business processes, focus on human activity and interactions. It's easy to go right to the details of data entry or how a piece of software works. That will come – but for those first conversations, focus on what the people do!
Follow where the data leads you, even if it's uncomfortable.
Sometimes you end up finding out that data that is supposed to be in order isn't, or that upon further examination, some generally accepted notion is not true. That's actually quite common. And this is where things the ride can get bumpy, because data may reveal issues with business processes and training, and inconsistencies in management or implementation.
Tip: Be clear with your data owner that it's their data and their business process needs to produce information that is accurate, if you're expected to provide reports and analyses. You might be tempted to "fix it with code" but in general, it's a bad idea to embed logic in your data processing to compensate for process issues because it will limit your options later.
Be flexible and adopt an iterative approach.
There are lots of ways you can run your project...just be aware that many of the approaches to software development like agile, rapid application development, extreme programming, etc., may need to be adapted to work well for data discovery and analytics. Schedule regular, frequent meetings with key contributors where face-to-face interactions can happen and decisions can be made. You can change styles later – for example, transition to a more structured process after all of the details are known and you need to engage a group that needs more detailed direction and oversight.
Tip: Recognize that your team has different audiences when setting up meetings and communications. Discussions that happen within the team when they are troubleshooting a data processing issue may not be appropriate for a meeting with all of your data owners and stakeholders. Keep separate schedules for meeting with these different audiences, and invite only those people who need to be present for the discussion at hand.