Amplifying Data Value

As we move into the new year, there are growing demands for more integrated skill sets among teams responsible for market intelligence. These integrated skill sets require teams to understand the versatile nature of today’s data, the way data can be combined and synthesized, the many techniques available to manage and render data solution-oriented, and finally, fundamentally understanding the business and organizational problems that can be addressed with data–driven solutions.

Given the technical requirements, building effective teams requires the inclusion of unique cross-functional data strategists. These are individuals that understand the business sufficiently to envision how data assets can be leveraged for business solutions. This means first understanding how to align data, data science, process and organizational structure such that data becomes an active agent in meeting a business objective.

This may seem somewhat abstract, so let me make it very concrete with an example. Starting with common internal assets such as customer satisfaction research and an internal CRM, how does a data strategist convert a simple ongoing measurement program and a common CRM into a powerful customer acquisition solution?

A Data Story

Let us begin with a brief overview of two successful data assets. The first is a CRM that functions as a loyalty program tracking purchase history and basic customer characteristics gleaned from a member sign up survey. The CRM has been used by the marketing team to support customer retention and rebuild loyalty among at-risk customers.  The second key data asset is the CX team’s routine customer satisfaction survey. The survey data provides a regular report on how interactions with company touchpoints contribute to positive customer experiences. Periodic analysis of this data promotes progressive changes to processes and products impacting customer perceptions.

The Problem: Customer Acquisition 

The leadership team is looking for a growth strategy. Customer retention is high, and the ongoing customer satisfaction research shows little change over time, despite ongoing changes to the customer experience program. Sales are flat and not contributing to growth. The focus needs to be on customer acquisition.  The leadership team sees “digital first” as the future and has committed to a growth strategy that prioritizes online campaigns. The promise of a winning digital strategy has not born fruit. The current shotgun approach has not been effective, and the marketing team is considering investing some of their ad spend on new research to better inform more effective targeting. They know demographic targeting and channel tactics (e.g., paid social or publisher focused) have been ineffective. Better targeting, the team feels, will get ads in front of those for whom the company’s products will have the greatest appeal. To do this, the marketing team needs a targeting solution built on an understanding of who their new customers should be and how to find them. The team is indecisive because of concern they will be directing some of their ad budgets to a segmentation study that will be difficult to integrate directly into a digital campaign, and difficult to update as markets change.

Enter the Data Strategist 

The data strategist approaches the problem from the perspective of Occam’s razor: How can I simplify the path from data to solution? This means leveraging existing data more effectively, before creating new data. This also means applying analytical techniques that give us more with less, more solution-oriented insights with fewer data inputs. And, this also involves making the entire process more efficient and dynamic, making it easy to update and apply.

The first task is to breakdown the problem into manageable components: 1) Determine the essential elements of the business objective (i.e., customer acquisition); 2) Inventory existing data assets and determine how each asset can contribute to each element of the business objective, and outline the gaps that may need to be filled by new or external data sources; 3) Explore how integration and analytical processes can transform data into an actual solution that addresses the business objective; 4) Finally, establish the technical and collaborative (i.e., cooperation between business silos and potential partners) roadmap to build the solution.

Following these steps, the data strategist can operationalize the basic steps that make existing data assets an integral part of a business solution that meets key objectives. In this case, the objective is customer acquisition.

OBJ 1: Knowing Your Best Customers [Effective Targeting]

The data strategist wants to understand how existing data can contribute to the business objective. In this case, marketing and sales need to build a more effective digital customer acquisition program. Assuming the messaging and creative decisions are already evaluated, the first step is knowing who it is you want to target with your campaign. The immediate data that comes to mind is twofold: CSat survey and CRM. Both data sources tell a different story about the customer. The first tells us the customer journey and how each touchpoint impacts the overall perception of the brand, the product, and services/support. The second depicts buying patterns (e.g., product quantity, frequency of purchases, when purchases are made, influence of special offers) and basic customer characteristics (e.g., location, point of purchase, demographics, etc.).  Separately, each data source provides an incomplete picture; joined they form a more holistic view. The combined picture helps delineate the high-value customers and the experience they have interacting with the company across an entire lifecycle. This holistic perspective also helps us understand factors such as “share of wallet”, loyalty, price sensitivity, at-risk triggers. While such insights are highly valued and can inform customer retention and grow customer spending strategies, they are only the first step to creating a framework that informs a targeting solution for customer acquisition.

OBJ 2: Moving From Internal to External 

Segmenting and describing your best customers do not address the business objective: customer acquisition. Creating an effective and efficient digital campaign, which addresses the business objective, the data strategist must envision three essential processes:  1) Integrate the CRM and CSat data. This means ID graphing and data imputation scenarios, all directed toward creating a single data environment from which ML clustering algorithms can be applied. 2) Create segments and apply a valuation framework for prioritizing targeting and aligning with different messaging, offer bundles or creatives delivered in digital campaigns. 3) The final process is where the digital strategist makes its unique contribution. This is where the segments that are created using internal data are linked to external data sources. This is done using predictive algorithms to find non-customer equivalencies in large digital platforms that build digital audiences for ad exchanges.

The third process is a critical step and requires some elaboration. Building customer segments in the first two processes is restricted to integrated internal data (i.e., CRM and CSat). While integration and data imputation for these processes are non-trivial, they do not offer an immediate connection to non-customers, which is central to the original business objective.  Traditionally, converting an internal customer segmentation into a customer acquisition segmentation involves two approaches. First, marketers are left with the task of aligning the differentiating behavioral or attitudinal (e.g., satisfaction) characteristics of their internal segments with the pre-defined targeting criteria offered by any given targeting options on a chosen digital advertising platform. This is often an intellectually frustrating and time-consuming exercise, which makes updating and re-evaluation exceedingly frustrating. Second, marketers often default to demographic indexing and simply choose targeting criteria that most closely approximates the over-indexing of their chosen segments. This is not particularly effective in most industries where customer segments are more behaviorally driven. Neither of these traditional approaches are effective. Convenient in the short term, but not effective. A more effective approach, which turns out to be more efficient as well, is building bridges between your segment members (within the internal integrated data) and partners operating large data lakes (1) with direct piping into digital advertising exchanges.  These bridges can be constructed differently depending on how the marketing team wants to partner in the digital advertising ecosystem. Once again, the data strategist earns their pay by navigating through this ecosystem. For example, the bridging mechanism for the CSat data can be obtained using survey pixels to link a respondent to a corresponding ID in a behavioral data lake. Alternatively, CRM data has personal information that can be used through legitimately secure matching services like LiveRamp, which matches to a corresponding behavioral data lake. The result is a triangulated alignment between the CRM and CSat, which are the backbone of the customer segmentation, and an external behavioral data lake tied to ad exchanges which provide a source for targeting non-customers sharing the characteristics of the targeted segments.(2) The next section discusses in greater detail how the internal-external data integration provides the foundation for the actual customer acquisition solution.

OBJ 3: Finding Your New Best Customer

At this point, we have found the customer persona that we want to target in our digital advertising campaign. The next step is building the mechanism whereby devices streaming through digital ad exchanges can be associated with the characteristics of the desired customer persona and efficiently targeted with the appropriate ad. The key to targeting potential customers using internal segmentation is building the predictive algorithm linking segment membership with external and independent behavioral data. Recall the segmentation is built from the CRM and CSat data. However, linking the segment membership to external data requires more than a robust ID graph and piping plan. This next step requires machine learning to build effective prediction algorithms. These algorithms require two essential inputs. The first input is a dependent variable, which in this case is segment membership. This is what we predict. The second input is the independent predictors. These are the behaviors found in the external data lake linked to each segment member. Machine learning (e.g., deep learning) uses behavioral data to predict segment membership. The resulting algorithm can be ported to the data lake and used to build audiences predicted to have the same hallmark characteristics as the original customer segment.  It is important to note that the predictors are external to the original CRM and CSat data used to build the original segments. Since this business outcome is customer acquisition, the algorithm is identifying potential customers that predicted to belong to the original segment but are sourced outside of the CRM/CSat environment. The resulting audiences can now be used to target devices passing through ad exchanges.