The hardest part of most analytics project is the first part – getting started. Almost without fail, you have questions, you have data, and you have some tools to help you analyze data in some fashion or form. With the necessities at your disposal, how can you get a project off the ground that will ultimately deliver results that matter to your company?
Below are four tips to help get you started. These are based both on my experience managing projects and what I’ve seen while engaged with Teradata partners and customers across a variety of verticals such as telecommunications, financial services, healthcare and insurance.
1. Start with an objective that matters to your company.
If you want your analysis to matter to your boss at the end of the day, tie it to a corporate or departmental objective. If your organization’s objective for the year is to increase the ROI of retention and loyalty marketing, don’t waste your time trying to understand correlations between weather events and on-time deliveries. Maybe you will develop a great insight, but good luck figuring out who can act on it, getting them to adjust their operations, and getting any credit for yourself.
2. Ask questions that your data will support.
You may be curious about the ROI of social media marketing. But if you aren’t already tracking sales driven by social media, save that battle for another day, or at least fight it in the background of more pressing concerns. Such projects are important but will take longer to deliver business value than analyses where you already have relevant data. Perhaps you have access to website data. Consider performing a cart abandonment analysis here to identify opportunities to improve your website to increase sales (assuming one of your corporate goals is to increase revenue ).
3. Encourage business people to ask hard questions.
Data scientists receive (and deserve) a lot of the credit for pushing advanced analytics projects into the mainstream at their companies. Despite the media and marketing hype, there are no solutions that can turn business analysts into data scientists, and there is no such thing as a “data scientist in a box.” The data science skill set is unique.
But business users are becoming smarter about data, and their tool kits are adapting and expanding. For example, BI tools let business people take advantage of more complex analytics than they once did. In the case of something near and dear to me at Teradata, we’ve developed guided interfaces for Aster Analytics so a less technical business user can apply extremely advanced analytic techniques to solve some of their toughest business questions without writing a line of code.
Encourage business users to ask tougher questions of your data. Both business people and data scientists will benefit from this in the long run.
4. Have a plan to operationalize your insights.
Data is great, and analytics show you know what to do with that data. But to really deliver business value, you need a plan to operationalize the results of your analyses. Let’s consider the example of a marketer at an e-commerce website. Maybe your objective is boost online sales. You have access to your web logs and customer purchase history. You use a tool like Aster’s Product Recommender solution to understand which products are frequently purchased together and which customers are similar. Then you try to understand paths to purchase – perhaps even purchases of specific products or categories – with Aster’s Path Analysis Guided Analytics Interface. Now, to operationalize your results, you propose recommending specific products on specific pages within your website. With the Product Recommender solution, you were able to perform “people like you also bought…” analyses, and you can apply these results in ongoing email or social campaigns. Without writing a single line of code, you have moved from a corporate objective to the operationalization of your analyses via multiple channels.
5. Bonus: Apply the right visualizations to communicate your insights.
The catalog of visualizations available to help you explain your analytic insights is really quite remarkable. While some of your most data-centric colleagues may be happy with a table view, and the PowerPoint crowd is loaded up on bar charts, you might want to explore other options that can help explain what’s happening in your underlying data. Consider a sigma graph to show the strength of relationships (how tightly products are correlated, in the Product Recommender example above) or tree charts, Sankey and sunburst diagrams to show how people flow to or from events of interest. Any of these are sure to help you socialize your analyses. Your palette as an analytics expert is growing.
Follow these rules and your next analytic project will be a home run. Best of luck!
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.