Big Data and Business Intelligence have been hot topics for many years and there have been significant advances in these areas. But many businesses don’t realize that constant changes in people, processes, systems, and data can introduce significant errors in reporting, including implementations and data sets that were previously accurate.
In my 20 years of implementing and auditing systems, I have seen many issues and developed some practical recommendations for improving data quality and reporting. I had the honor of sharing some of my experiences by speaking at the August Financial Executives International’s (FEI) Summer Breakfast Professional Development Session. The focus was educating financial executives on the high probability of unknown data issues. This blog is a summary of key points.
To Master the Output, First Master the Input
The way to master business intelligence and data visualization is by mastering the inputs to your operations, systems, and data. I've found that some of the biggest issues can be the easiest to control. This includes change in processes, assigning qualified staff for system changes, data load failures, timeliness of data, lack of design documentation, incorrect calculations, incorrect filtering, and lack of data governance.
Change in Processes
Even minor changes in processes can cause data to be dead-ended. Even if changes gain efficiencies, bypassing old processes can require systems and reporting to be modified. Changes in processes must be communicated to a team that can determine if systems and/or reporting needs to be updated.
System Changes, Updates, & Upgrades
Make sure your area has a qualified person representing your team for the upgrade. Be an advocate for your area and dedicate appropriate staff to complete the upgrade/testing. Ensure that all staff members have been trained on any new processes and/or systems.
Data Loads Failing
Network/system outages, delays, and changes in data, up-streams systems, upgrades/updates, processes and other things can cause loads to fail. Design systems and build alerts to notify the (subject area) owner of the data and IT that the data either was not received, failed to load, did not complete and/or trend with historical data.
Scheduling & Timeliness of Data
Your company should have controls/alerts that confirm that all data is received timely and completely before downstream data consolidation/reporting begins.
Lacking Design Documentation
Design documents can include work instructions, system and/or view design documents, data dictionaries, data catalogues and data transformation details. As employees change jobs (more frequently) and as many of our employees age-out, the need for documenting processes and systems increases. You should require documentation of processes and systems. These will need to be cataloged and updated as changes are made. This makes it easier to transition employees and document the design at various points-in-time.
IT people aren’t accounting or finance people. IT’s business language is very different from finance, accounting, marketing and other business subject areas. Always confirm that data is calculated correctly. Design documents can be used to translate code into business rules which ensures that the various departments are speaking the same language.
Be sure to give specific advice for the data you want included or excluded from the report. Include assumptions and request footnotes on the report if the data you’re presenting differs from the official version. This allows users to easily explain why the data is different than the official version. You don’t want to spend meeting time trying to decide whose numbers are right. It’s always a good idea to tie a new report to a trusted data source (like the GL).
Lack of Data Governance
Place key employees on system, operational, and reporting initiatives. Create a formal process to control the flow of work requests and changes to your systems, hierarchies, list of values, reports, and data warehouses. Standardize core metrics across the organization and always push for “one (correct) version of the truth”.
Many companies are developing business intelligence initiatives in order to improve their operations, but when the above issues are not addressed I have seen these mistakes cause between $1M and $2B in reporting errors. Quality systems, processes, and data are the pillars of operational and strategic reporting. Building a great data warehouse is like building a house; make sure to spend 80% of your time planning the build to ensure your data is high-quality and structured in a way that makes data visualization easy. There are a lot of great BI tools to visualize data, but the quality of the data is more important than the appearance of the report.
More strategic and financial executives are requesting that BI data be used for planning and forecasting. The documentation of systems, implementing data governance, and similar initiatives will give you a greater understanding of your data and processes and make your planning initiatives easier.