Customers tell us what they want with every click, call, trip, interaction, and purchase. But too often we aren’t listening and instead offer the same tired products and solutions our customers have already ignored or told us they don’t want. Paying attention to customer feedback and giving them what they want is critical for business success. And thanks to the lightning pace of analytics technologies today, we can now do so at scale and these five critical factors for successful customer analytics projects will help:
Successful customer analytics projects require clear objectives and well-defined tactics. Unless the entire team is clear on the strategy and tactics to be deployed and how analytics will be used to guide decisions, even the best advanced analytical algorithm won’t improve business performance.
Start by identifying the primary objective. Do you want to build brand capital or generate demand? Is the goal to cultivate and manage sales leads through a pipeline? Is leadership most concerned with personalization and customer insights? Does the executive team want to accomplish all of these or more? If so it will be necessary to rank (and perhaps weight) each of the objectives.
Next, identify the tactics that will be deployed to achieve the objectives and how the insights will drive those tactics. A brand capital campaign might focus on targeted messaging via mass media and measuring the results as part of a marketing mix optimization while demand generation might be measured in terms of incoming inquiries, mobile/web traffic, or additional steps along the customer journey.
Finally, mapping the outcomes based on model performance and the measurement of actuals back to key performance indicators (KPIs) and financial statement metrics are critical to understanding the business performance.
To truly understand your customer it is critical to dig through all of the information you have available and for larger organizations in or near the Petabyte range, requiring significant data infrastructure.
Cloud architecture and infrastructure as code are making it much easier to stand up an analytics lab for model development. Since this data can be highly sensitive containing Confidential Information (CI), Personally Identifiable Information (PII), and (depending upon the industry) Protected Health Information (PHI, appropriate security, and masking are required.
Big data is often required for customer analytics projects which necessitates a significant investment in infrastructure and expenses for data management. The project team will require time to understand the data, place it in the proper business context, and integrate the data into a format for analysis.
Seasoned data scientists are a must for navigating the quantity of data enterprises are tackling and for producing the detailed outcomes and insights expected by leadership. This starts with, but is not limited to:
Finally, cause and effect must be identified for successful customer analytics projects to estimate the model’s performance when deployed to production. Knowing how to code the best big data/data science algorithm is just one small piece of the puzzle.
How do you know if your customer analytics project is successful? Accurately measuring results, of course. Many marketers have learned the importance of avoiding last touch attribution when measuring results since it inflates the impact of the last touch. Multi-touch attribution provides a better understanding, as can the design of experiments.
Other considerations for successful measurement include continuous testing in the business environment and comparison of champion and challenger performance over time. Additionally, tracking the customer journey on both the individual purchase and lifetime relationship can aid the business in understanding how customers transition from one enterprise segment to another. Lastly, it’s important that the outcomes be measured in terms of the core KPIs the business leaders track, and that those KPIs are indicative of the financial benefit the models are driving for the business.
Execution requires discipline and cooperation between the analytics team, IT, and the business. The tools used to wrangle data and create models are often different than the data integration tools and scoring code used in production. Understanding the generated insights and what to do about them will require change management and training efforts. Executives will want to closely monitor performance and determine when to override model recommendations with human judgments.
Additionally, continuous monitoring and testing are necessary to ensure model performance does not slowly or abruptly degrade and to protect against population drift. When results vary significantly from predicted values its time to re-evaluate models.
Becoming a truly customer-centric organization requires significant effort but can be highly valuable to businesses as they work to compete for customer attention and loyalty. Leveraging today’s complex, but advanced customer analytics technologies can be a game changer for enterprises who are ready to provide their business users with the insights they need to seek out and serve their customers in the most desirable way possible.
This blog was originally published on Forbes.