The brief usually arrives the same way. The business wants to do more with AI, the board has asked what the plan is, and the first role signed off is a data scientist.
Six months later, that data scientist is spending most of the week chasing system access, reconciling two versions of the same customer record, and writing scripts to clean data that should have arrived clean.
Essentially, they are doing data engineering. They’re being paid to do something else, and they are usually starting to look elsewhere.
Why the sequencing goes wrong
Data science is the visible discipline. It produces models, forecasts and slides that make sense to an executive audience. Data engineering produces pipelines, which produce nothing visible until they break.
The consequence is well evidenced. Gartner expects organisations to abandon 60% of AI projects through 2026 where the underlying data is not AI-ready, and found that 63% of organisations either lack the right data management practices or do not know whether they have them.
Very few of those projects fail because the modelling was poor. They fail because nobody built the foundations first.
What each role actually does
A data engineer builds and maintains the systems that move data from where it is created to where it can be used: ingestion, transformation, storage design, pipeline reliability, and the monitoring that tells you when something has quietly stopped working. Their output is infrastructure other people depend on.
A data scientist uses that infrastructure to answer questions: modelling, experimentation, statistical inference, and increasingly the design and evaluation of machine learning systems.
The dependency runs one way. A data engineer delivers value without a data scientist in place, through cleaner definitions, reliable reporting and fewer arguments about whose number is correct. A data scientist without a data engineer will spend most of their time doing a job they were not hired for.
A simple test for where to start
Four questions worth putting to your team:
- Can you get a trustworthy answer on core business metrics without manual reconciliation? If not, you have an engineering problem.
- Does everyone agree what a customer, an order or an active user is? Inconsistent definitions across systems are the most common blocker we see.
- How long does it take to get a new data source into production? If the answer is measured in months, adding analytical headcount will not change your speed.
- Is there a specific decision you want a model to improve? If the use case is still “we should be using AI”, hire the engineer and build the foundations while the business works out what it needs.
Three out of four pointing at engineering is common, including in businesses that have already hired analytically.
The hire that often makes more sense than either
For many mid-sized organisations the right first hire is an analytics engineer, or a senior data engineer with modelling literacy: someone who can build the pipeline and produce the insight while the team is still small. It is a harder brief to fill, because the skills sit across two traditional job specs, but it removes the dependency problem for the first 12 to 18 months.
Getting the market right
Strong data engineers are among the most contested candidates in UK technology hiring. They are typically employed, well looked after, and unresponsive to generic approaches. They also assess employers carefully. An engineer joining a business with no data platform wants to know they are building one, not firefighting one.
Be specific about the state of your estate and what you want built. Candidates at this level respect an honest brief far more than an optimistic one.
Talk to Intec Select about your data team
We work with CTOs and heads of data to sequence data hires around what the business needs next, rather than around the job title that happened to get signed off first.
