Analytics engineer vs data engineer: which role you are hiring, decided by where your data breaks, and the funded standard for each

The vacancy is titled Analytics Engineer. You read down the responsibilities and half of them are ingestion: standing up pipelines, scheduling overnight jobs, keeping a platform alive. The title says one role and the work is the other, and no round of interviews will fix that, because the mistake was made when the job was written, not when the candidates applied.

Analytics engineers and data engineers overlap almost entirely on SQL and diverge on nearly everything else, and the question that separates them is not seniority or salary. It is where your data breaks. When data goes missing, arrives late or lands malformed before it reaches you, the work is data engineering. When data arrives on time but the modelled, trusted layer built on top of it is wrong, the work is analytics engineering. Work out which of those two is your real problem and you have named the role you need.

Where does your data break?

One test predicts the right hire more reliably than any job advert: trace where your data breaks, and hire for that layer.

Upstream breakage is an engineering problem. The dashboard is empty because a pipeline failed overnight. A data scientist spends three days a week wrangling ingestion instead of modelling. A new source needs connecting and nobody owns the plumbing to carry it. The data is late, broken or absent, and everything downstream waits on it. That is data engineering: building and running the systems that collect, move and store data so it lands reliably in the first place.

Downstream breakage is a trust problem. The data lands on time, but three teams define "active customer" three different ways and the board sees three different figures. Analysts hedge every number because nobody has tested the tables they query. The raw data is sound; the modelled layer above it is untested, undocumented and disputed. That is analytics engineering: building the tested, documented models that turn data the business can reach into numbers the business can trust.

What if both ends are broken? Plenty of teams find the pipeline and the modelled layer are each a mess, and the test still gives you an order. Start upstream, because a trusted layer cannot stand on a pipeline that keeps failing underneath it. And where the load at both ends is genuinely large, you are looking at two roles; asking one person to carry both is how you burn out whoever you appoint.

Same SQL, opposite ends of the same pipe. The overlap is real, which is why the titles blur. The divergence is where the hiring decision actually lives.

	Analytics engineering role	Data engineering role
Where the data breaks	Downstream, after data has landed	Upstream, before data lands
What they own	The modelled, trusted layer: the tested, documented tables analysts query	The pipelines, storage and platform underneath that layer
A day's work	Writing, testing and documenting data models; the output is code that analysts' queries depend on	Building and orchestrating pipelines, running the platform that feeds everything downstream
Core toolset	Advanced SQL plus the transformation and warehouse layer (a modelling tool such as dbt on a cloud warehouse)	SQL plus the pipeline, orchestration and platform layer (workflow orchestration and distributed processing)
Funded UK standard	No dedicated Analytics Engineer standard; the work maps to ST0118 (Level 4), or to ST1386 (Level 5) where it leans into pipelines	A standard of its own: the Data Engineer standard, ST1386 (Level 5)
Best-fit signal	Your analysts do not trust the numbers, though the data itself lands reliably	Your pipelines and platform are fragile or missing, and everything downstream waits

Read the two levels as scope, not seniority. ST0118 sits at Level 4 and ST1386 at Level 5 because the standards cover different bodies of work; the number tracks a standard's scope, and each role stands as senior technical work in its own right. A senior analytics engineer and a senior data engineer sit at the same table.

What it costs to hire the wrong one

Hire an analytics engineer for what is really an ingestion problem, and eighteen months later you have someone with a clean, tested modelling layer sitting on data that arrives late, half-formed or not at all. They keep polishing models built on a source that never lands cleanly. The trouble was upstream the whole time, and an analytics engineer carries no remit to reach it.

The mirror mistake costs just as much. Hire a data engineer to settle a trust problem, where the numbers are all there but three teams read them three different ways, and you get robust pipelines feeding a warehouse the business still argues over. A data engineer is not tasked to arbitrate what those numbers mean, or to build the tested models that would settle the argument, so the complaint that opened the search, we do not trust the data, outlives the new hire. Both mistakes share one root: the break got read off the job title while the real fault sat a layer away.

The advert title is the least reliable signal

Job titles for these two roles are not standardised. One firm's "analytics engineer" is another firm's "data engineer", and plenty of adverts mix duties from both. A title tells you what a hiring manager typed on the day; the responsibilities tell you what the work is. Read those, and place the role by the layer it operates on.

Keep the analytics engineer distinct from the data analyst here too, because ST0118 sits behind both. An analytics engineer builds and owns the modelled layer, the tested tables and their documentation; a data analyst queries that layer to answer business questions. One standard funds both, and the job spec the employer writes is what draws the line between them. If that is the pair you are weighing, our comparison of the analytics engineer and data analyst roles draws it in full.

Why salary cannot break the tie

Pay is the worst available tiebreaker here, and the honest reason is a data problem: the numbers you can find do not measure what you need them to. The Office for National Statistics publishes no dedicated occupation code for either title, so the nearest measured figure comes from a broad group, SOC 2133 ("IT business analysts, architects and systems designers"), which pools these two roles in with several others. Its ONS ASHE figure describes that whole pool.

Advertised salaries are a different kind of number: they show what employers post to attract candidates, a step removed from what the workforce actually earns, and for these two titles the advertised ranges sit almost on top of each other. So the two available signals fail in two different ways. The measured figure is too broad, because it lumps many roles into one group; the advertised figure is too wide, because the ranges for the two titles overlap almost completely. Neither one can pick your role for you. The daily work can.

Map the need to a funded standard

Once you know which break you have, the funded route follows, and the two roles lead to two different apprenticeship standards.

If the break is downstream, so your analysts cannot trust the numbers even though the data lands reliably, you are training for analytics-engineering work. There is no apprenticeship standard titled Analytics Engineer, so the role trains through the standard that covers the work: most often the Level 4 Data Analyst standard, ST0118, which iO-Sphere delivers as its Advanced Data & AI apprenticeship.

If the break is upstream, so your pipelines and platform are the bottleneck, you are training for data-engineering work, and that has a standard of its own: the Level 5 Data Engineer standard (ST1386), which describes the occupation as building "systems that collect, manage, and convert data into usable information for data scientists, data analysts and business intelligence analysts to interpret." iO-Sphere delivers it as our Data Engineering apprenticeship.

Both standards carry a typical duration of 24 months; iO-Sphere delivers each in 15 months of training plus a 3-month end-point assessment. The protected off-the-job training differs between them, at 418 hours for ST1386 and 370 for ST0118, a real block of released, paid time the employer plans around. Funding is capped by a band on each standard, up to £19,000 for the Data Engineer standard and up to £15,000 for the Data Analyst standard. If you would rather map roles to standards before committing, our employer team can work through it with you.

One honest limit. iO-Sphere delivers apprenticeship standards up to Level 5. If the role you actually need is a data scientist, a machine-learning engineer or a Level 6 or 7 degree-level route, that is not one we train, and our Level 5 route is not a stand-in for it. The government's Find apprenticeship training service lists providers by standard, so you can find one that delivers the level you need.

Common questions

Analytics engineer or data engineer: which do I need?

The deciding question is where your data breaks: whether your problem is data that fails to arrive, which is data engineering, or data that arrives but cannot be trusted, which is analytics engineering. If pipelines fail, data lands late, or a source has no owner, you need someone to build and run the systems underneath. If the data lands reliably but nobody trusts the numbers because the modelled tables are untested, undocumented or disputed, you need someone to build the trusted layer on top. Name that break and both the role and the standard behind it follow.

Is "analytics engineer" a recognised apprenticeship standard?

No. There is no apprenticeship standard titled Analytics Engineer; it is a market job title rather than a defined occupation. So the funded route is the nearest approved standard that covers the actual work: most often the Level 4 Data Analyst standard (ST0118), and the Level 5 Data Engineer standard (ST1386) where the role leans toward pipelines and platform. Read the role's real responsibilities against a standard's knowledge, skills and behaviours before you place it under one.

Which standard trains a data engineer?

The data-engineering role maps to the Data Engineer standard (ST1386, Level 5), a distinct funded occupation covering the systems that collect, move, store and convert data for everyone downstream. It carries a typical duration of 24 months; iO-Sphere delivers it in 15 months of training plus a 3-month end-point assessment, with training funding capped at a band of up to £19,000.

Should the salary decide which role I hire?

No. The Office for National Statistics publishes no dedicated pay figure for either title, and the two numbers you can find each fail in their own way: the nearest measured proxy, the broad SOC 2133 group, pools these roles with several others, and the advertised ranges for the two titles overlap almost completely. One figure is too broad and the other too wide, so pay cannot separate the roles. The work is what separates them.

What happens if I train an apprentice against the wrong standard?

The knowledge, skills and behaviours they build will not match the ones their end-point assessment tests. Put an apprentice whose daily work is pipeline engineering on the Data Analyst standard, or one whose work is modelling on the Data Engineer standard, and they reach assessment with a portfolio that cannot evidence the standard it is marked against, while the business problem you set out to solve stays open. So map the day-to-day work to a standard's KSBs while you are still choosing the standard, well ahead of enrolment.

If you are still separating the two nearest roles, our comparison of the data engineer and data analyst roles and the wider data engineering topic hub fill in the surrounding map. Once you have named the break you are filling, our Data Engineering apprenticeship covers the Level 5 route and our Advanced Data & AI apprenticeship covers the Level 4 route that trains analytics-engineering work.