How I Think About Analytics
Analytics is defined many ways in different organizations. The defining factor in how I see analytics functions designed and utilized has to do with where an organization is in terms of maturity of process and data strategy. For an organization just starting out, extracting simple reports is analytics to them. For a digital-first, AI-driven organization, analytics means something very different and much more sophisticated.
Analytics is about the application of data. It’s a different animal from data engineering, and it’s a different animal from the traditional BI reporting functions. Analytics is where value is created from the data assets available. How insights and decisions are made is affected by how well analytics integrates into existing process. Analytics is best positioned to act as a sort of glue between a lot of different groups. Data and analytics can balance the viewpoints of disparate stakeholders, be those the cost components of the infrastructure versus the objectives of the business leader, or some sort of trade-off that needs to happen between a sales department and a delivery department.
For instance, resources, infrastructure, and platforms used by an analytics team are often owned and administered by an IT department or a technology group. Business requirements of the infrastructure come with costs and may be beyond existing capabilities. Analytics sits between the technology and the business needs to provide balance in what is pragmatic and possible.
Beyond basic statistics and trend reporting, analytics models business processes and relationships in the data — not just outcomes. It defines which levers can be pulled and estimates the effect of pulling each lever. Analytics seeks to answer the why, the dynamics not just the current state, to extend the data story one step further. You may see a metric being up over time — is that a result of a change, would results have been markedly different if a different action had been taken? Then most importantly these models can put tangible values behind this.
Where Analytics Functions Fail
Most often when I see an analytics function underdeliver, it’s because the business stakeholders have lost confidence. And the most common reason I’ve seen confidence be lost is speed of delivery. When you have an analytics function supporting various groups throughout an organization, things will get prioritized and deprioritized based on business needs and direction. The feeling of the constituent base is that nothing’s being done. This creates pressure toward decentralized analytics resources, where the constituent team performs analytics tasks in house — use their own analysts, get direct access to reporting and data, and perform value-add tasks manually.
The drawback of that approach is that a centralized analytics function likely has more expertise in solution methods and more options are possible than an individual effort. So, while an answer is produced by federated analytics resources, it may not be as quality a deliverable as it could have been with the collaborative strength of an entire function.
What has been very clear to me seeing different leaders and organizations is that when organizational leaders truly champion analytics and how it can increase revenues and reduce costs, that’s where analytics teams are most successful and more net value is created. Insightful, quality products are produced with tangible monetary improvements. This positions analytics to solve ‘big’ problems, to fix broken process, to streamline decisions and hit above its weight class. Yet it comes with tradeoffs — meaningful projects take time and individual stakeholders may not get their exact ask timely and on demand. The question is whether you want analytics focused on solving the whole problem for everyone or focused on the immediate needs of single stakeholders, knowing a forthcoming, robust solution may well illuminate their need.
This is not an easy ask. It takes organizational adoption to drive meaningful organizational change. Without high-level buy-in, isolated analytics solutions rely on grassroots advocacy with only pockets of usage and limited benefit from a given analytics product.
Most often, it isn’t the analysis that fails but the presentation, delivery, and adoption. If nobody uses the insight, it creates no value. Analytics works best when it’s embedded, when it is systemic, where natural usage is aligned with incentives to encourage behavior change and organizational improvement. Peripheral analytics lives outside workflows and requires extra effort to use. Trusted analytics is embedded directly into processes at the moment decisions are made.
Building Trust with Business Leaders
Building trust starts with understanding what the pains, pressures, and objectives of business leaders are. Any analytics project, any product that is considered and created, must align to those business objectives. If a business leader has a mandate to increase sales numbers, the models and scenarios produced by the analytics team need to be trained on those same sales numbers. The mathematical objective optimized within the analytics model has to match the business objectives.
Beyond that, everyone should be skeptical of models — be they LLM and AI models or a data science model that forecasts or predicts business outcomes. Anytime a business leader is given a black-box-produced result (whether it was built internally, purchased from a vendor, or coming from an LLM model) they should be skeptical. They should require proof that this is a good model. They should want to know what implicit and explicit assumptions were made and if they are appropriate. They should want to know that it will perform as promised and as designed. I would expect leaders to start with a pilot or a proof of concept. The analytics team, if they’re worth their salt, should be measuring and reporting on the outcomes as they are available, providing the reporting and data necessary to show requirements are met and desired outcomes achieved. Only then would I expect or want a business leader to advocate for adoption or expansion to a larger part of the organization.
This is also why it is paramount to establish success criteria before a project even starts. I tend to see adoption metrics — usage, clicks of some sort — being used as a proxy for success, and I don’t think that’s the right thing to measure. If someone clicks for the sake of clicking, is anything of value achieved? Any analytics product, whether produced internally or purchased externally, needs to have very specific metrics to be affected, defined upfront, and those metrics need to tie to business outcomes and success.
Understanding the business and partnering with business leaders and stakeholders to understand their pain points and problems, then tailoring solutions to robustly address them, is essential to driving the actual and perceived value of the analytics function within an organization.
Driving Meaningful Adoption
There is a two-pronged approach to meaningful adoption. Analytics products and their resulting insight has to meet users where they are, and proven results must be fostered organizationally.
Dashboards, generally speaking, should be used exclusively as reporting tools. Once analytics moves beyond information to a realm of prescriptive and predictive models, results must be integrated with work process and meet workers in their day-to-day work. If we build a recommendation engine, those recommendations should be pre-populated into fields inside of the systems where the work happens, or displayed in system when that information is useful. If people have to seek out information and insight from a separate dashboard, if they must interrupt their regular work process, the model will remain unused and underutilized. Deploying analytics into reporting rather than relevant systems is how analytics stays peripheral, and peripheral analytics does not change outcomes.
Putting organizational support behind proven solutions provides an enormous boost to analytics adoption. Having senior leaders champion analytics products and explain how and why adoption is moving the organization toward larger objectives provides a user base with the context they might be missing when they are diligently working through the tasks and processes. A view from the top — hearing a high-level perspective — means a lot and drives a culture of innovation. Having operational support of local leaders and producing quick guides or otherwise fostering usage helps to drive the how. Communications to answer questions and showcase initial results creates the awareness. Functional townhalls can provide context and detail to help a user base firm up their understanding and ask the right questions about how to affect enhancements and improvements to a fledgling analytics product. Showcasing analytics is not about giving kudos to the build and implementation teams — it is about showing how the new capabilities genuinely create value in all areas.
What the Analytics Leader Actually Does
This brings me back to where I began. Analytics as a function is a strange animal. It isn’t quite IT and Technology; it doesn’t perform the business tasks. Historically analytics resources are deployed throughout an organization to support specific and often niche needs. Huge strides can be made and high-impact projects achieved as a cross-functional organized analytics function. Analytics sits at a crossroads between varied, competing, and disparate needs of business stakeholders and technology. Analytics holds the group together supported by hard data and insights with a singular focus of business optimization — designing and deploying models that meet leadership objectives.
Different teams offer competing priorities. Good analytics acts as the glue that brings those perspectives together. When models are aligned to overall business outcomes, the data becomes a neutral mediator. The role of the analytics leader is often to mediate these conversations and let the data tell the story.
An analytics leader needs to be a stalwart advocate for the resources and design of infrastructure that support the type of analysis most conducive to producing productive and profitable results.
Ultimately, success comes down to culture and allowing decisions to be driven by data with models that provide levers to affect business change and improve outcomes. For products to be used and adopted they need support from the top down. But this is of course conditional on products consistently demonstrating their value. I have found the most important aspect of getting that top-down support is being able to demonstrate meaningful outcomes aligned to the objectives of the business leader. If we can go back and say we heard what you were looking for, and if we implement this process change, introduce this model, provide this product to your user base, we’re looking at this amount of return in the measures you are looking to increase — that’s what it takes to get true organizational buy-in. That’s the job.