6 Steps to Take Control of Your Data Architecture

By March 15, 2018Analytics

Without the right processes and tools, it’s easy for a digital analyst to spend more time pulling and organizing data than reporting their findings and delivering meaningful analyses.

What can organizations do to ensure the right processes and tools are in place? 

With the relatively recent introduction and adoption of the data warehouse, a central repository for the collection of business data, it is getting easier to cut down on the manual manipulation and processing we’ve historically needed to integrate data from disparate sources. Successfully moving to a more integrated system, such as a data warehouse, will save time and resources, allowing you to dedicate more time to actually analyzing the data you collect.

If you are having trouble getting started with this process, breaking the transition down into manageable pieces can be helpful. My guidance for breaking down this process is as follows:

Step 1 –  Assess tools and systems and how they work together

First, take stock of all the tools and systems that your organization uses currently and how they relate to each other. Groom this list if necessary. Talk to the stakeholders involved with each system and find out what is working well and where there are pain points. How can these systems be better integrated with one another?

Step 2 –  Develop an overall plan for data structure

Is your company using a data warehouse? If so, there should be clear and up-to-date documentation on what you are capturing in the data warehouse in the form of a data dictionary. If there are data sources that are not included in the data warehouse, find out why. Would it be worth adding these ad-hoc data sources to the data warehouse? Adding the sources may not be necessary if you can easily integrate them using a reporting tool.

If your company is not currently using a data warehouse; are you joining data sources manually, or using an automated tool like Tableau or Domo? Connecting data sources manually is excruciatingly time-consuming. Tableau or Domo can substitute for a data warehouse, and will cut back on manual joins.

Examples of data that you might want to include in your warehouse include:

– Digital data, like page views, users, and visits (from Google or Adobe Analytics)

– CRM data (from Salesforce, Marketo)

– Email marketing data (from Mailchimp, Constant Contact)

– Customer service data (from your chat box, Zendesk, Twitter help, or phone)

– Third party data (from Dun and Bradstreet)

Step 3 –  Define business goals and questions

As you navigate through this transition, don’t forget to keep your end goals in mind – What are the important questions your organization needs to answer, and how can you help answer them more effectively? Determining applicable KPIs for each business unit, conducting a site audit if needed and developing a tagging solution with Google Tag Manager or Adobe DTM for any new metrics that are needed are all important parts of the process.

Step 4 – Ensure consistency in data collection

Without a consistent method for how you collect data, your capabilities for conducting meaningful (and accurate) analyses will be limited, which can create a lack of trust in reports.

For example, if a company changes how it collects website data, it is difficult to produce accurate year-over-year comparisons. The use of virtual pageviews, implementation of photo galleries and changing a website’s structure can all significantly affect traffic and are especially important to consider when thinking about site changes.

Carefully consider any changes that will affect data collection and the potential consequences on reporting. Do not rush into any decisions that may impact your ability to answer important business questions in the future. If a major change is necessary, make sure it is thoroughly and thoughtfully documented.

Another issue with data consistency is having incomplete data, especially in circumstances where data duplication makes it difficult to know where there are gaps. One common example is having multiple names of accounts. ABC corporation may be captured as “ABC corp,” “ABC corporation” and “ABC co” in different platforms. This makes it difficult to accurately combine data.

Make sure that all data is collected and joined appropriately before analysis, taking into consideration potential pain points where duplicate data is involved. A clear and consistent process should be in place to ensure duplicate data is dealt with effectively, whether that means taking care to limit the possibility of duplicating data in the first place, or effectively combining or editing the data where needed.

Step 5 –  Select a data visualization tool

Is your company using the right data visualization tool? Can you create reporting directly from your data warehouse, or easily integrate with a data visualization tool? If not, as I mentioned previously,  it’s time to reconsider. Work towards automating meaningful reporting that will be used. Too often I’ve seen organizations spend time developing dashboards that are not useful in the long term. Find the right questions to answer, and figure out how to best share that information with  the applicable audience.  Consider delivery, timing and takeaways. What actions should this information spur?

The following are some questions you might want to consider before selecting a data visualization tool for your company:

– How can data be directly connected to the data visualization tool?

– How will reports and dashboards be distributed?

– Are there budgetary limits or concerns?

– What types of analysis/visualization capabilities are required?

– Will dashboards need to be interactive? Automatically updated?

– Are there specific visual reporting constraints, i.e. color scheme/branding?

– Will you need to limit access to data by business groups?

Step 6 –  Reporting and analysis

An important distinction to make is the difference between reporting and analysis. Reporting should be largely automated, and include metrics from your data sources, such as site traffic statistics like monthly page views over the past year.

Analysis is adding context to the reported data in order to answer questions or inform business decisions. For example, you might consider unanticipated changes in monthly pageviews — spikes, dips, steady increase/decrease etc. — against your marketing initiatives, seasonality effects, changes in implementation, and other potential causes. If you see a major year-over-year change in site traffic on a particular date, it may be due to a major news event or marketing campaign.

The real fruit of a smooth data architecture implementation is the analyses that come from the reports.  Real analysis of the data with actionable takeaways to improve business decisions should be the end goal. With a comprehensive plan in place, analysis of this kind is not only possible, it’s highly achievable. Lastly, once you’ve got a good structure and flow of reporting, consider how to build in room for iterations – don’t let an evolving business get stuck with an outdated model or analysis capabilities.

Developing a strong and well-integrated data architecture plan is a large undertaking with equally large benefits.  Approach this transformation in manageable pieces, taking care to develop a company-wide, collaborative approach to analytics. The subsequent cohesive data architecture structure will prove beneficial in the end — in time saved and ideas generated and acted upon.

 

This article was originally published on CMSWire.com on September 5, 2017.

Leave a Reply