How to plan a business intelligence project strategically

Building a business intelligence project plan in 4 steps

In order to appreciate this proposed framework’s nuances, consider a large retailer that’s attempting to leverage retail business intelligence to improve sales while reducing acquisition costs. The chief marketing officer (CMO) has been tasked with this strategic objective and need to develop an approach that earns buy-in not only from senior management, but also from the heads of every department whose cooperation would be imperative for successful delivery. How should the CMO proceed?

Below is a four-step approach to translate this objective into viable and tactical delivery assignments:

Step 1. Translate strategic objectives into specific problem statements

So, what does “improve sales while reducing acquisition costs” mean to someone other than the CEO or shareholder? This is a rather loaded remit, but what could it mean tactically? We’ve broken this down into smaller problem statements:

• How can we utilize our customers’ data to ensure that they stay with us? Clearly, keeping a customer active for longer will reduce acquisition costs while increasing lifetime sales.
• How can we use our marketing budgets more efficiently? Are we spending the optimal amount on each channel? Non-optimized spending can quickly deplete marketing budgets, especially in a highly competitive niche.
• Can we focus on a particular segment of customers in order to maximize overall sales?
• What if we focus more on retention rather than acquisition marketing? Acquisition costs would go down and we may generate more sales over the customer lifecycle.
• What about product pricing? Can we maximize sales by offering more dynamic pricing for specific customer segments?
• What about online merchandising? Could a different merchandising scheme be used to increase sales from an existing inventory and product catalog?

This step involves breaking down one high-level business objective into specific problem statements that make sense from a data perspective. Every possible interpretation above will have different data requirements, modeling approaches, risks, and cost/benefit profiles, but these would be impossible to assess without dividing the larger objective into smaller data problems.

Managers can frame problem statements by analyzing how the larger objective relates to every department and by identifying each department’s specific optimizations. This inclusive approach not only allows the capturing of potential optimization options from all parts of the business, but also helps keep key stakeholders engaged.

Step 2. Conduct a high-level cost/benefit analysis of options

Say, you have over 50 possible analytical options that might deliver on the larger business objective. Marketing, product management, finance, customer service, and supply chain have all contributed, resulting in a bloated agenda. How do you set priorities?

Individual optimization options must be further elaborated and subjected to cost/benefit analysis. Take the example above: suppose that this analysis is needed for implementing dynamic pricing. This would entail the following considerations:

• Assume that catalog pricing was 100% dynamic and based on each individual customer’s data. What is the best-case lift in sales given the total number of customers and average purchase frequency?
• What is the estimated cost of implementing data infrastructure (e.g., a data fabric architecture solution) to create composite customer profiles (combining purchases, refunds, web analytics, customer service, etc.)?
• What would be the margins impact if dynamic pricing were implemented? In almost all cases, the margin-per-product would diminish, which would have to be weighed against the lift in sales.
• Is the cost/benefit projection favorable?
• If it is favorable, can it be implemented within realistic timeframes?
• Would any additional resources or the re-alignment of existing ones be needed?
• Are there any major process dependencies (e.g., departmental politics, other on-going programs, etc.) or legal constraints (such as the GDPR) that might impact our customer data handling capability?

In this step, high-level estimates are developed for the benefits and costs of each option while key project constraints and assumptions are outlined. This information allows managers to make objective assessments of which options to prioritize or reject. For example, implementing personalization and dynamic pricing might benefit business in the long term, but it might be possible to drive higher sales by simply optimizing the campaign spend.

This step’s key deliverables include:

• Ballpark estimate of costs
• Clear definition of benefits
• List of assumptions used

At this point, you should have a clear list of options with high-level cost estimates, but which are still based on many assumptions that need to be further validated. Yet, this step is the first filter in screening out options that don’t make sense to pursue, even on the basis of a cursory analysis.

Step 3. Define SMART business intelligence goals where possible

Once the prioritized list of implementation options is available, the next step is developing SMART goals for each one. For dynamic pricing, these could be:

• Generate total sales in excess of $X from the entire catalog
• Average product margin must be greater than $Y
• Project must be completed within A months and business benefits must start showing within B months of project completion

These goals typically require a combination of acute business judgment, extensive familiarity with data preparation and modeling techniques, project management experience, a highly collaborative working environment, and, above all, senior stakeholders’ acceptance that the final results might be different.

In most cases, the effort is formidable, but having objective benchmarks provides constraints for implementation teams and impacts everything from data integration technology choices and modeling techniques to the quality and quantity of resources deployed.

Step 3 is also the most technical and hands-on step in the proposed planning process. For example, why settle on $X as the sale target? The number would typically come by establishing a relationship between sales and individual personalization parameters (number of browsing sessions in a specific time, pages used in those sessions, time per page, past purchase history, any customer service interactions, etc.). This requires experienced data scientists, ETL specialists, statisticians, and product experts to sketch out analysis details, including a high-level overview of key influencing parameters, data quality requirements, modeling techniques to be deployed, and so on. For now, the objective is to make assumptions and some ballpark calculations to generate a SMART objective.

Managers should also further refine the cost estimates from the previous step. More importantly, this step places numerical values on benefits, thereby making the requirements unambiguous. Additionally, other details including resource requirements, hardware/software purchase required (if any), and a refined list of key assumptions and risks should become available.

Step 4. Define global practices, tools, roles, and responsibilities

Steps 1-3 are repetitive and specific to individual business objectives. However, a successful enterprise-wide BI implementation project plan involves creating a set of global practices, assets, tools, staffing arrangements, and delivery methods that can be applied across projects with minimal customization. Regarding business intelligence, some of these activities might include:

Lining up roles and responsibilities. Business intelligence requires business consultants, data specialists, statisticians, and project managers. Other roles may be necessary depending on specific contexts, but at minimum the key responsibilities for these roles should be clarified as part of global staffing strategy.

Picking up the right data mining tool. For data mining to be successful for business, models must be developed quickly and deployed cost-effectively within current operational systems and business processes. Various tools (SAS, SPSS, R, etc.) exist for automated model development, data integration, and model testing, and it is usually best to decide these at the global level, rather than let it be done differently for individual business intelligence projects.

Standardizing a data mining method. Adopting a consistent data mining methodology lies at the heart of realizing benefits from investments into analytics. This would typically involve:

• Clearly defined implementation steps along with agenda for each
• Inputs and outputs for each step
• Project roles and responsibilities

A structured approach developed along these lines ensures purpose and repeatability in addition to allowing data mining to be implemented as discrete projects with defined budgets and timelines. CRISP-DM is the de-facto method of several organizations worldwide, although many have tailored it to their own specific business context.

Putting it all together

When applied properly, this planning framework results in a concrete, repeatable set of steps for translating high-level business objectives into specific project initiatives and ballpark estimates for time, cost, and resource requirements. Developing this strategic context helps maximize the value of data mining investments and avoid the “ad-hoc trap” that often wastes enterprise resources because of the wrong priorities set or the absence of advance insight into investment returns or business outcomes.