Why Do Salesforce Data Cloud Implementations Fail?

Data Cloud has been one of Salesforce’s flagship products ever since its renaming from Salesforce CDP and the short-lived Genie. Since then, Data Cloud has had a stellar track record – at least from a sales perspective – reaching a formidable three-digit year-over-year growth in 2024. Clearly, Salesforce has been successful at bringing Data Cloud to market, but how has it performed against user expectations? 

This is where things get tricky. Salesforce has been open about the failure rate of CRM implementations in the past, but they remain awfully quiet about that of Data Cloud. Whatever the reason may be, the question remains: do the same factors that govern CRM implementation success apply to Data Cloud, or is it a different beast altogether? To find out, I contacted leading Data Cloud experts around the globe and asked their thoughts on reasons why Data Cloud implementations fail.

Reason 1: What Is Data Cloud, Really?

Have you tried to explain Data Cloud to a business stakeholder? Easier said than done. Salesforce themselves shuffle between terminology: some call it a customer data platform (CDP), others use data lake or data lakehouse, and some refer to it as the semantic layer for Agentforce. The real answer to “what is Data Cloud” is probably all of the above and more. This versatility is ultimately a good thing, but also makes Data Cloud difficult to communicate to customers and partners.

“Data Cloud is still something of a grey area. It’s evolving so fast that while other SF products have a quarterly release, Data Cloud’s release is every few weeks! There are constant changes and updates happening that make it difficult to keep up.” Jack Searle, Capgemini

To make things more difficult, Data Cloud doesn’t hold intrinsic value in the same way a CRM or a system of record does. Data Cloud’s value is indirect, depending on the use cases it powers. In essence, Data Cloud is an enabler. 

To build a solid business case for Data Cloud, one must understand in depth the problems to solve and what roles Data Cloud needs to fulfill in the grand scheme. This is something that requires a deep technical understanding of the platform, the customer’s business, and its enterprise architecture as a whole.

“Complexity kills adoption,” Jacob Hayes, Accenture, explained. “Organizations sometimes overcomplicate their Data Cloud setup by building overly sophisticated segmentation logic, unnecessary data pipelines, and redundant processing steps. Build for business usability. If marketing and analytics teams can’t use the data, the implementation has failed.“

Reason 2: Know What You Pay For

One common pitfall of Data Cloud projects is its consumption-based pricing model. Instead of the seat-based pricing more familiar to Salesforce users, Data Cloud charges based on allotted credits, with each data ingestion, activation, profile unification, etc. consuming a predefined amount. 

“Organizations often migrate too much data into Data Cloud without a clear business case,” said Jacob. “[This leads] to increased credit consumption, costs and inefficiencies. Regularly evaluate which use cases require Data Cloud vs. alternative solutions.“

Not knowing how this pricing model works can lead to excessive charges for the customer. This, in turn, may lead to customers and partners alike to be overcautious on building new use cases with Data Cloud. Expertise in Data Cloud’s licensing is needed to strike a balance between cost-effectiveness and agile development.

“Most clients are not clear on how they are charged for Data Cloud usage and development. This often results in an overinflated bill halfway through their implementation. As such, I would encourage implementers to deep dive into the topic of licensing.” Eunice Wong, Valtech

The main downside of Data Cloud’s pricing model is a lack of foresight. Data Cloud’s Digital Wallet is a relatively new feature that helps alleviate this issue. It shows credit consumption versus remaining allowance and when these were last updated. 

The problem with Digital Wallet is that it looks backward on already consumed credits, and it is not updated in real-time. As of the writing of this article, there is no out-of-the-box tool for forecasting or predicting credit consumption in advance. Granular optimization of credit consumption falls on the Data Cloud partner’s expertise, third-party tools, and custom reports.

Reason 3: Plans Are Nothing; Planning Is Everything

A famous quote from Dwight D. Eisenhower states that fixed plans do not work; you need to be able to maneuver and deviate from the path. 

Planning itself is valuable, though, as it serves as the foundation upon which the project is built. In Data Cloud’s context, a certain amount of planning is necessary before getting started (and during the implementation), but fixing yourself on a predefined plan is a surefire way to miss the mark.

“An important thing to know about Data Cloud is this: The vast majority of the work is not spent within the system, but actually planning what you want to do. Scoping out use cases, identifying data sources and connectors to use.” Jack Searle, Capgemini

Planning with Data Cloud is particularly hard because it is more like a system of engagement than a system of record. Data Cloud is not meant to be used as a data store but as an intermediary for ingesting, unifying, and activating data to Salesforce and beyond. 

This means that Data Cloud’s data model shouldn’t be approached as a whole but from the perspective of individual use cases. This is more akin to marketing automation systems, less like CRMs and DWs. Failure to understand this difference leads to a longer time to value and the risk of needing to rework the data model at a later stage.

“Segments and activations come with caveats based on your data architecture,” Eunice added. “Clients often do not understand this in the beginning. They build their data model first and find that they are limited in terms of what they can segment or activate on. To avoid needing to rework the data model, partners should consider segments and activations right from the start.”

Ryan Hernalsteen of Prolicity also stated: “While you can unmap and remap DMOs pretty easily, you really want to think through your mapping to Individual before you start identity resolution. Changing the Individual ID, or unmapping an object from Individual, requires you to delete the identity resolution match rule.

“If you have segments already using match rules, you have to delete and recreate those.“

Reason 4: Use Cases, Use Cases, Use Cases

Use case design for Data Cloud suffers from the chicken-and-egg conundrum. You need use cases with a real business impact for a proper implementation but identifying the use cases is often difficult before business stakeholders have seen Data Cloud in action and understand its capabilities. 

Well-executed demos and customer training beforehand mitigate this issue somewhat, but Data Cloud use cases are still tricky to sketch out. Salesforce has heeded the need for Data Cloud education and responded with an impressive library of content to support customers and partners alike.

“Data Cloud projects are mostly discussion and planning. Much of it is spent on identifying use cases, and at an early stage of the implementation, it can be harder to do without having a full understanding of your Salesforce setup, usage and business processes.” Jack Searle, Capgemini

The thing about use cases is that they are as much about quantity as they are about quality. Identifying a handful of brilliant use cases initially is all well and good, but no single use case will justify the time and effort invested into Data Cloud adoption. Even if you figure out the initial 4-5 use cases, who is to say that you can maintain momentum and keep churning new ones after go-live? 

Keeping the pace of use case development is key in ensuring that Data Cloud continues to serve the overall enterprise architecture.

Final Thoughts

While Salesforce Data Cloud boasts impressive sales growth and market penetration, there is a contrast between its commercial success and a bubbling dissent among customers and partners. 

The complexity of positioning Data Cloud, navigating its pricing model, and the need for meticulous planning and use case design highlight significant challenges. These issues raise questions about Data Cloud’s future role as the engine of Salesforce’s AI agents. 

Can it truly meet user expectations and drive the next wave of innovation, or will it falter under the weight of its own complexity? As Data Cloud continues to evolve, its ability to adapt and deliver tangible value will be the ultimate test of its long-term viability.