Updated: Feb 14
Maybe you're like me - analytics minded, but only to a point. Marketing, Sales, Human Resources, and plenty of other business units love to keep their mind on their analytics (and their analytics on their mind), but it's often far from an area of deep expertise for the bulk of their team members.
I can nerd-out on Google Analytics, CRM metrics, and digital advertising data. Comparing CTR, pageviews, and other performance KPIs and developing high-level readouts in Excel and PowerPoint are core competencies I've refined over a career in marketing.
Maybe I'll use a pivot table. Reluctantly. I know V-Lookups are a thing, but using one is way out of my depth. And that's to say nothing about Tableau data visualizations or the data management skills needed to set up a real-time analytics environment.
I mean seriously - regardless of the CRM or ERP you use, what level of confidence do you truly have that your data is reliable? 80%? Let's not bury our heads in the sand here. Do you have a consistent format for your columns? Is every entry uniform 100% of the time? Do you Quality Control the data being entered externally, for example when a new client fills out a form?
Senior Marketing Manager
Sr. Marketing Manager
Senior Marketing Mgr
Sr. Marketing Mgr.
Senior Manager, Marketing
You know what I mean? And that’s just the beginning.
Or, what about this - (I'm actually asking ping me on LinkedIn if you have the answer) - is there a difference between a Director of Human Resources and a Head of Human Resources? A LinkedIn search for the former yields about a million results. The latter yields 600k. Are they the same? If so, which one fits your usecase better? If they're not the same, how do you differentiate them to ensure coherent reporting on the topic? Does it matter to different end-users of the analytics?
That's the essence of end-to-end analytics: critically examining your data from the point of entry and carrying that same attention to detail through to the point of usefulness. But this isn't always easy when there are so many different tools involved in gathering that data, curating it into something usable, analyzing it from multiple angles, and ultimately presenting the results back to stakeholders within the organization. That's where end-to-end analytics comes in: It's an approach for analyzing business intelligence across an entire organization.
Can your organization be successful by tweaking ad hoc reports on a case-by-case basis? Probably! I've certainly seen it. But I think most would agree it's not the smartest use of either your abilities or your time. So, what next?
End-to-End Analytics: The Basics
End-to-end analytics is a technology-agnostic philosophy. It doesn't matter if your organization is using on-premises servers and simply exporting to Excel - there's plenty of room to implement the core best practices that are also employed by enterprise organizations using cloud storage and robust data transformation and visualization tools. This process is always going to start at the data source and conclude with end-users creating reports, spreadsheets, or data visualizations.
Data Ingestion & Master Data Management: Set Yourself Up for Success
Master data management (MDM) involves the creation of processes and systems to ensure that data is complete, accurate, and consistent across the organization and that it is used effectively to support business processes and decision-making. It helps organizations to ensure that they are working with high-quality data and that they can trust the data they are using to make important decisions. MDM involves defining what data needs to be collected, organizing it in a way that will support future analysis, and making it accessible to the people who will be responsible for analytics.
A crucial part of setting up your data for analytics is ensuring high data quality. Quality data is vital to any organization that uses analytics. Unfortunately, no matter how great your operations and source systems are, data often needs to be cleaned to ensure that it is accurate, consistent, and usable. There are several reasons why data can become dirty or incomplete, such as data entry errors, data format inconsistencies between systems, missing or incomplete data, and redundant data between systems. Any of these can lead to inconsistencies and inaccurate analytics. Extraneous data can also be removed during this process to ensure that all available data is accurate, valuable, and ready to use.
KPI Definition & Gathering Real Requirements: Understand Who Is Using Your Data & Why
Having good requirements is crucial for the success of a data project. It helps ensure that the final product meets the needs of the business. Gathering requirements begins with involving the right stakeholders. This should include the end users as well as subject matter experts. The stakeholders will define the Key Performance Indicators (KPIs) and success criteria for a project, and failure to include the right people can create confusion or missed requirements.
To define KPIs, you should first identify the specific goals and objectives that you want to achieve. Then, consider the metrics most useful for tracking progress toward those goals. The metrics you choose should be measurable, relevant, and actionable. It is important to choose a small number of KPIs that are most indicative of progress toward the stated goals, and ideally to include benchmarks or goals for each KPI.
The requirements should also define who will be using the data solution, how they will use it, and what action will be taken based on the information presented. To have a successful solution, you may need to address the needs of multiple types of users, with different goals, processes, and pain points (personas). Identifying the personas early in a project is important to a timely and successful project. A representative of each persona should be involved in the requirements process, but every individual user does not need to be involved, which can help make the requirements-gathering process more efficient.
Other tools can be used to help ensure the final product meets the goals of the project. User observation and interviews will provide insight into the current state. Here the most valuable information is what is currently working, and what pain points need to be addressed. These will often come up in one-on-one or small group user sessions, while they are missed in typical requirements gathering meetings. Wireframes, Mockups, and Prototypes are particularly useful tools in discussing and refining requirements, setting expectations, and facilitating conversations with non-technical stakeholders.
Transformational Data Layers and Data Modeling: Enrich the Data in an Intentioned & Scalable Way
Data layers separate the raw source data from the presentation layer. Structuring data layers in a modular fashion allows for a more maintainable design and provides higher confidence in the lineage of the data being used for analytics. Data layers handle the storage, access, manipulation, and enrichment of the data.
A foundational (data collection and storage) layer provides the foundation for downstream manipulation. This layer typically has all relevant fields and data. An organization will often have the most granular data available with minimal transformation. In an ELT (Extract, Load, Transform) process, this encompasses the Extract and Load stage.
The data processing layer handles the transformation, cleansing, and combination of data. This can be implemented modularly to set the stage for multiple use cases and improve the maintainability of downstream analytics. This layer can be aggregated to the level necessary for the resulting analysis, as well as incorporate business logic, standard row-level calculations, and any supplementary data to be integrated. Data aggregation will reduce the size of the resulting data set and can improve performance and mask confidential information prior to exposing the data to end-users. However, not all aggregation and calculations can or should be done at this layer, and this will depend on the analysis being performed. If you used wireframes or mockups in your requirements process, these will be invaluable to informing what aggregations and transformations are needed at this layer.
The presentation or analysis layer surfaces the prepared data to the data visualization or analytics tools used by the organization. This layer does not usually perform any further transformation but allows for a modular design.
Testing should be performed at each layer or stage of the processing to ensure there is no record loss, different components of the model are functioning and integrating as expected, the model can handle the volume of data and types of queries being performed, and no bugs have been introduced causing existing functionality to break.
Data Visualization & Presentation Best Practices: Ensure Adoption & Agile Iteration
Data Visualization is a useful way to present data because it can help make large and complex datasets easier to understand, allowing users to see patterns, trends, and relationships in the data that might not be clear if the data was presented in a more raw form. Visualizations can also be more engaging and memorable than raw data, which can help people better understand the information. Typically, data visualization will appear as a dashboard or suite of dashboards. Dashboards allow users to monitor KPIs, track progress, see trends and patterns in the data, and even aid in diagnosing problems. An effective dashboard should focus on the core KPIs identified in the requirements and follow a logical flow from high-level aggregated KPIs with benchmarks or goals to more granular information used to drill into areas of concern.
Are we on track for our KPIs? This is where benchmarks or goals are crucial to knowing what is expected. Visual cues can quickly inform the user where to look further.
If not, what is contributing to the failure to meet the goal? This is where drilling down into additional dimensions for analysis becomes necessary. It’s important to know the audience and provide the information that is relevant to them. Time-series views and other dimensions such as regions, product lines, customers, etc. help refine the search for a cause.
What actions can be taken to address the cause? This will depend on the audience. For an executive dashboard, a link to the process owner or a different dashboard may be enough. For an operational dashboard, this may involve links to source applications, ticketing systems, or more detailed views.
Documentation: Dashboards should typically include helper text, instructions, and/or links to SOP documents. User training and processes established by the business domains are also critical at this stage. The users will need to know what actions are available or expected and often requires training and change management by an organization to get strong adoption and ideal outcomes.
Focusing on the needs of the personas identified in the requirements phase and including the least amount of information needed to meet those needs results in a clean, easy-to-use product that will support business decisions. If you have multiple personas involved, you may need to create dashboards tailored to each, rather than trying to meet the needs of all users with a one-stop-shop. Keep in mind that “one-size-fits-all” often results in “one-size-fits-none”.
As tempting as it may be, dashboards don’t need to be packed full of metrics or extraneous functionality – they just need to make the users’ jobs easier and supply the necessary information to take the next action. To this end, highlighting areas that require attention, along with the ability to drill down into those areas will make a dashboard an invaluable tool.
End to End Analytics & Digital Transformation
Data analytics are transforming businesses across every industry—from retail to healthcare to financial services—because of their ability to drive better decision-making at every level of an organization. This organizational shift has come to be known as digital transformation, a concept with many varieties, but a shared common goal: integrating data driven decision-making at every level the organization.
You can see this in Industry 4.0, where automation, data exchange, cloud computing, IoT (Internet of Things), and artificial intelligence converge and show us what the fourth industrial revolution could look like.
Even the biotech and pharmaceutical industries, typically reluctant to adapt to new technology and embrace digital solutions, are accelerating towards digital transformation in the wake of the COVID-19 pandemic as a way to ensure broad access to safe medicines.
We are certain to witness the expected improvement in productivity with a digital transformation strategy, robust levels of digital diffusion, and personnel up-skilling. From the top of the corporate ladder to the bottom, now is the time to embrace digitization more than ever. After all, it is essential to our economy. During digitization and digital transformation, your company will need to develop cybersecurity, artificial intelligence, and other strategies. To manage these technologies, you'll need a current skill pool that can put them to work as soon as they arrive. Up-skilling and training can make a huge difference in the game. To increase productivity, it is vital that all businesses, not just a few, adopt digitalization and technology.
Source: Glenn Hole, Anastasia S. Hole, Ian McFalone-Shaw,
Digitalization in pharmaceutical industry: What to focus on under the digital implementation process?
International Journal of Pharmaceutics: X, Volume 3, 2021, 100095, ISSN 2590-1567,
Suffice it to say, training and upskilling are critically important to the adoption of new technologies and the overall success of digital transformation efforts. Analytics Centers of Excellence (CoE) have become a vital administrative structures for organizations to manage strategy, innovation, methodology, enablement, administration and operations. Additionally, the development of an "analytics community" where users can share knowledge and ideas, learn best practices, and contact the COE for support and guidance is often a needed step towards true self-service analytics.
I started off by asking you how confident you were in the consistency of your CRM data. If you brought that number to 100%, how far would that move the needle organizationally? Do you struggle with siloed data? It would be more surprising if you didn't. But, gradually, even the slow-to-change behemoths are seeing connected data as a fundamental differentiator amongst competitors.
Looking for job opportunities in the analytics space? We're always looking to add data visualization, data engineering, and application development consultants - learn more on our careers page. There are also several platforms like Jooble where you can find excellent job opportunities to put your knowledge about end-to end analytics into action.