Like a house built to withstand the seasons, a successful and sustainable analytics project must start with a firm foundation, a purposeful plan, a seasoned team, and the right tools and materials. Analytics project owners and budget-holders, much like a homeowner, want to see their large investment result in something that delivers value and supports evolving needs.
Elder Research helps companies accomplish this at a strategic level through our Analytics Assessments, and a similar approach can be used for individual projects. Using the five assessment areas of Culture, Process, People, Infrastructure, and Analytics, we help our clients lay the groundwork for success in machine learning, AI, and other analytics initiatives.
Whether you are building a house or constructing a solid analytics project, you can anticipate and mitigate the most common challenges by asking the right questions.
Culture: Understand your Neighborhood
We’ve all heard the home value adage “Location, location, location.” When building a house, it primarily signifies a property’s proximity to desirable locations, but also speaks to the quality of the substrate and the culture of the neighborhood. A data-driven culture in any organization provides solid ground upon which to build a successful analytics project that will remain relevant as needs change. The right culture affords a strong base on which to build when inevitable change happens, and a shared understanding to sustain the work beyond deployment.
An analytics culture is critical to reaping the full gains of any project, and you’ll benefit from taking an honest look at the Role of Culture in your organization and how that will impact your plan. Is your working environment dedicated to results? To truth? To making a difference? To taking action and making a change if evidence points to that being what will best serve the mission? To using data to make decisions, even if the data contradicts popular belief? Regardless of who gets credit?
Once you understand your cultural core, you can consider the following:
Do you have key stakeholder support?
Think broadly about who needs to support your project. We would all love for our projects to have ubiquitous support, but some stakeholders are more crucial than others when working on analytics solutions. To determine key stakeholders answer the following questions:
- Do you need access to software controlled by another department?
- Do you need subject matter expertise from someone that interacts with the process you’re working on?
- Do you need to have buy-in from the end-users who will ultimately choose to use or ignore project results?
- Do you need support from high-level leaders across the organization to help normalize any related process changes?
It is important to engage stakeholders in a way that respects their other priorities and helps them understand the impact this project will have on their work and the work of others. Keep in mind that support for the most successful projects comes frequently and from many directions. Securing a diverse and influential set of stakeholders for your project greatly increases the likelihood of success.
How will the organization need to change?
Most people are hesitant or averse to change, especially if they don’t know the full impact of the change. In order to navigate this resistance, it’s important to think through who will be impacted and how they will be affected.
- What processes will be different and how?
- What information will be available to whom?
- Who will need to take action on the results?
- Which departments will be impacted by the final solution?
- Who may be concerned about displacement by the solution that warrants career planning or upskilling?
Anticipating the impact of your project should start early so there is plenty of time to communicate the change and prepare those affected. Your project’s firm foundation can only be viable long-term by building bridges to the rest of your organization. Some organizations have entire teams dedicated to communications plans and change management that can be enlisted in this work. If so, get them involved. If you don’t have dedicated personnel, build extra time into your project for this important task.
Culture Summary: Understand your Neighborhood
Process: Planning for Success
Most home-dwellers learn what works best for their needs—how to make the best use of the layout, closet and storage space, walkways, number of bedrooms, and built-ins can make or break a home. An architect can help design a home that will meet the needs. From these blueprints, the build team can plan the necessary steps to reach that shared vision. The design of a successful analytics project isn’t much different. You want to make sure that your project accomplishes the desired goal by considering the following:
Are our objectives for analytics clearly defined?
Well-defined project objectives will give your team a shared vision and will help focus individual efforts. These objectives should be broadly communicated and revisited throughout the project to ensure that your team stays on the right track. Be sure to think about how you will know your objectives have been met, since they should be a measure by which you evaluate success. Setting clear objectives up front can often help secure the resources (people, time, budget, etc.) needed to complete the project by showing decision-makers the benefits of the project.
Are we tackling too much too fast?
Richard Loewy, designer of well-known icons such as the Coca-Cola bottle and the Greyhound logo, created products with just the right balance between the well-known present and an innovative future, coining the MAYA Principle, giving clients the “Most Advanced. Yet Acceptable.” Sometimes a giant leap is exactly what is needed to propel a business forward. However, in many cases a measured approach, with small and frequent early wins, builds momentum across the organization towards long-term success. You may find greater success defining more palatable projects that take incremental steps towards your bold vision.
What level of analytics is right for the project?
Elder Research has identified Ten Levels of Analytics with increasing complexity. From descriptive business intelligence to advanced causal modeling, there are multiple analytical approaches to data-based problems. Finding the right intersection of needs based on the problem, solution, methodological complexity, and available expertise to implement that solution are the keys to designing a successful project. Make sure you have people involved who can speak to the benefits and limitations of each level in the context of available data and your project objectives.
What is the plan for deployment?
Deployment is the valuable end goal of any analytics project—getting the solution into an environment where it is ready to be used. While it may be tempting to think of this as a switch that is turned on when the work is done and it goes live, it’s best to plan early for what must happen before, during and after deployment.
- How will you test and validate your solution?
- What metrics will determine you have met your project objectives?
- What is the training plan to teach end-users how to use the output?
- Who should be involved in integrating the project into the existing environment?
- Who will monitor the model or process in its live state to identify when problems occur?
Deployment will go more smoothly if you include the insights and wishes of your key stakeholders and subject matter experts throughout the project.
Process Summary: Planning for Success
People: Assemble the Team
Once you’ve found your location, settled on a design, and prepped your foundation it’s time to assemble the team that will build your dream house. You’ll want to engage people that have experience building the type of structure you’ve designed. Most likely you’ll enlist an electrician, plumber, carpenter, and a general contractor to keep the project moving. Some things you may want to tackle on your own—maybe painting the guest room your favorite colors. Other tasks, like installing natural gas lines might be better left to the experts. Staffing a successful analytics team is no different. You need to convene a group with the skills required to execute the project plan. The team may consist entirely of people from within your department or organization or may incorporate outside perspectives and expertise.
How should we include domain subject matter experts?
Subject matter experts (SMEs) provide critical context for the project, helping technical experts understand the data, technology, and interface with existing processes. SMEs can also help ensure the team uses a practical approach and understand if the model output has value for end-users. Without the engagement of someone who intimately knows the business, an analytics project may not deliver actionable results that provide value to the business. Our CEO, Gerhard Pilcher, puts it simply in his blog Top 3 Objectives Before starting an Analytics Project, “A data scientist who tries to build an analytic model without consulting SMEs is like a lawyer who tries to handle a case without consulting the client.”
Do we have the right team?
Elder Research has helped many clients with Building an Effective Data Science Team, and it’s important to note that each project will require a unique set of skills and expertise. In general, we find that medium to large projects are best tackled by these types of team members:
- Data Scientists to build models using their unique expertise
- Data Engineers to create a base table of transformed data that is more suitable for modeling
- Software Engineers to create software around the model
- Information Technology Specialists to oversee implementation of the model in a production environment
- Champions to provide leadership for the cultural transformation and training aspects of the project
- Project Managers to keep things moving forward on larger projects and serve as a nexus of communication
One person may have multiple roles, but don’t seat too many roles with one individual. The benefits brought by teamwork and collaboration are nearly as important as the varied skills and knowledge domains. It’s also important to understand the capacity of your existing staff and determine when it will help to bring in additional resources such as external consultants. On one hand, if your team members are unprepared for a critical analytics endeavor, you are setting them up for failure. On the other hand, if you don’t leverage them to their highest level of capacity, you will be leaving valuable brainpower on the table. We find that some of our most successful projects partner internal expertise with our consultants—this allows us to help build internal capacity for our clients.
People Summary: Convening the Team
Infrastructure: Laying out the Tools
Once you have selected your team it’s time to gather the right tools. In a construction project, the tools often come with the build team. In an analytics project, you should evaluate existing tools relative to project requirements and then make an informed decision about the need to invest in new tools.
Do we have the right tools?
Using the wrong software, technology, and tools can increase your project’s cost, timeline, or chances of failure. Rudimentary tools may not have the features you need to accomplish your objectives. Complicated tools may add unnecessary project cost or extensive learning curves that can threaten completion or adoption of your solution. Many tools used for data storage, machine learning, visualization, and DevOps can be overwhelming to learn, in addition to the many options for infrastructure and programming languages. We recommend consulting with someone who has broad experience and can give an informed and unbiased (or at least transparent) opinion on the best tools to accomplish the project’s goals.
Beware of tools that claim to do all things analytics. Don’t let fancy bells and whistles mislead you into acquiring a tool that is more than you need or software that doesn’t match your use case. We often recommend using whatever tools are already integrated into our client’s environment, but we do have some favorites that may reduce development time or improve solution results.
Infrastructure Summary: Laying out the Tools
Analytics: Framing with the Right Materials
You’ll have lots of options when it comes to building the framework for your house—poured concrete or cinderblocks, pine or oak, asphalt or slate. The choices you make will impact how your house is constructed, the skills required, what it will look like when completed, and how long it will last before you need to do maintenance. In an analytics project, that infrastructure includes having the appropriate IT and technical infrastructure and making sure you’re working with the right data.
Do we have the right data?
Once you have determined your project’s objectives, it’s important your analytics team can access the right data. This can relate not only to quantity (making sure you have enough relevant historical data for an accurate forecast, for instance) and cleanliness (how much the data can be trusted or needs refining), but also to the type of data itself. This may mean looking beyond the data you’ve historically held, to seek data adjacent to your project.
As an example let’s say you own a large chain of sports-themed coffee shops and want to use an algorithm to explore possible locations for your next grand opening. You have demographics from your customer loyalty program that suggest college-aged folks love your particular flavor of java. You might even have zip codes or city names that suggest where there are aggregations of those customers. While a data analyst might connect the fact that Daytona Beach, FL has a lot of college students, an algorithm wouldn’t know that strictly based on the city or zip code. If you pulled in external data, your algorithm and decisions could be informed by the fact that there are about 69,000 residents in Daytona Beach, and 38,000 students enrolled in higher education. Connect loyalty demographics with public data on enrolled students by geography and store-level sales, and you might determine that the corner of 92 and S Clyde Morris Blvd is a desirable location! In this example, traffic patterns, median income, media consumption, and even weather patterns might help to focus your pricing strategy, product mix, and promotional approaches as you explore opening a new storefront.
A deeper consideration of data also applies in other settings:
- If you want to forecast summer sales in regional clothing stores, do you know how much was shipped to each store and how much was sold on clearance at the end of the season?
- If your service is geared towards winter sports, do you have data from enough years to give forecasting models a good sense of how your product sells across all four seasons?
- Do you have customer comments from your website that might explain unexpected variances in past sales?
Most projects aren’t blessed with perfect data, but understanding what data you have, its limitations, and what can be added can help you achieve your project goals.
Analytics Summary: Framing with the Right Materials
In our metaphorical house build, you’ve picked a location, made a plan based on a well-thought-out design, and assembled the team, tools, and materials to get started. Congratulations! It’s taken time to get there, but the payoff will be significant. This type of planning will get you the house you want now and will help with future additions, enhance your equity investment, and reduce the need for maintenance.
The same is true for analytics projects. Asking the right questions in each of the core analytics assessment areas will help you build strong analytics projects from the ground up.
Questions Leading to Analytic Project Success