Why the single source of truth is not the answer (for real-time use cases)
The dream of a single source of truth might sound compelling, but it’s just that: a dream. And it’s not even the right dream. The fact is, businesses have multiple data sources, and that’s okay. In fact, it can be healthy: it reflects the multi-faceted reality of how businesses and data function.
The single source of truth model: Where did this even come from?
In the mid-1990s, Tom Siebel popularized a utopic vision: replace siloed systems with a single source of truth. Imagine a wealth of customer data, instantly available, driving seamless customer service and marketing.
The single source of truth is sometimes called “the golden record” or “360° view.” It was supposed to cure all ills: data silos, IT problems, and stagnant profits. To be fair, siloed systems really did cause problems, like data redundancy, inaccuracy, broken messages, and missed opportunities when necessary data wasn’t in the right place at the right time. Pretty much everyone was ready to jump on the single source of truth train. The only problem is… the real world doesn’t work like that.
Myth: A single tool can do it all. Reality: Use cases determine the best tools.
In reality, there will never be a single source of truth. Ever.
People want to use best-in-class tools for their purpose. Developers, in turn, naturally strive to distinguish themselves by building tools optimized for a particular market and purpose.
Think about it like this: you get a flat tire and go looking for a lug wrench. But instead, all you find is a hammer-wrench-screwdriver combo. Needless to say, it doesn’t work very well. What you need is a lug wrench: a specialized tool that performs a specific job exactly as it should.
Consolidating all customer data within a single, all-purpose tool will result in a suboptimal business. It’s simply too much data, not just for real-time processing, but also for people to effectively use.
The solution is to choose specialized tools based on use cases that deliver the right data to the right teams, at the right time, in the right tool. For example, marketers need up-to-date insights on customer traits and behaviors. Customer service needs instant access to in-progress orders, previous purchases, and shipping status. Both groups need a system that collects, processes, and outputs data optimized for their particular purpose.
Myth: A single data warehouse can do it all. Reality: Warehouses are useful, but can’t keep up with real-time analysis.
At this point you might be thinking that a data warehouse solves the single source of truth problem. After all, using a data warehouse lets you ingest data from disparate sources and reuse it across a variety of tools, processes, and teams.
But data warehouses aren’t suitable for real-time analysis. And specialized tools mean there will always be new data islands: sets of data locked into whatever tool is best for the purpose. As such, a data warehouse will not solve your silo problems or meet all your data needs. Know its limitations; it can be useful, but isn’t the be-all, end-all of data management.
Myth: A CDP can do it all. Reality: CDPs limit the power of data.
Customer data platforms (CDPs) like Segment, RudderStack, and Hull are touted as a potential solution to the silo “problem,” offering a single source of truth. But most are focused on a specific use case: marketing. They aren’t optimized for the rest of your business: sales, customer service, and more. They’re just another silo.
Further, CDPs force all data models to conform to the lowest common denominator, based on whatever model the CDP determines. That means you can’t model all business problems accurately to deliver the right information to the right places at the right time. For example, companies with a two-sided marketplace of sellers and buyers can’t naturally model the relationship between the two.
Myth: Fragmentation must be eliminated. Reality: Silos can be healthy.
The anti-fragmentation theory goes like this: businesses lose money by duplicating information (or failing to recognize it) across systems. Fix the fragmentation, instant profit boost. But fragmentation isn’t necessarily a bad thing.
As organizations grow, they naturally become siloed. Departments emerge to take advantage of the efficiencies offered by specialized functions, and each department reasonably selects best-of-breed tooling optimized for their particular needs. They’re never going to agree on a single tool. They can’t even agree on a single methodology.
And that’s just fine. No solution can excel at meeting every need. And even if creating one was financially feasible (which it isn’t), forcing a business into a single system would be like a companywide potato sack race where everyone’s hopping in a different direction.
The solution is to choose a tech stack that matches your differing needs. Some systems might be used by multiple teams, but others will be specific to a particular function.
Cutting to the chase at Customer.io
Here’s how we’re evolving the Customer.io platform to support the reality of multiple sources of truth and our customers’ need for real-time use cases.
Data in. It’s time to expand our data model. Customers have multiple data sources, which can create contradiction and ambiguity. We’re working on:
- Easier ways to connect data sources. For example, recurring SQL syncs and webhook ingestion.
- Faster ways to reshape data. For example, merging logic and workflow actions that reshape data in Customer.io, with no custom code or infrastructure required.
- Accessible ways to relate data. For example, marketer-friendly automation triggers that can message people who belong to an account, order, appointment, or other business object.
Data out. Customer.io is the source of truth for some data, like messages and message metrics. We’re working to make that data easy to access from whatever best-of-breed tools our customers choose. We won’t trap data on our platform.
Multiple sources of truth: how businesses and data really function
It can be very tempting to believe there’s a single source of truth that can do it all. But data and businesses just don’t function that way, and don’t need to. Instead, we should know which data is relevant and impactful for a particular purpose at a particular time. The way to stay agile is to leverage multiple tools optimized for specific use cases and recombine data only as necessary. Don’t fall for the single source of truth illusion. Your business is more sophisticated than that, and deserves an approach that makes the most of the best tools out there.