19.11.26

How to Vet an ETL or Data Migration Specialist for a Legacy-to-Cloud Project

Why ETL and Migration Screening Keeps Failing

A CV that lists Informatica, Fivetran, DBT, and Databricks in the same bullet point tells a hiring manager almost nothing. It tells you a candidate has touched these tools — not that they can own a legacy-to-cloud migration end to end, debug a failed pipeline at 2am, or explain why a transformation model broke after a schema change. We run these screening calls every week for consultancies and investment management clients moving off legacy platforms, and the pattern is consistent: the gap between "used Fivetran" and "built and owned the ingestion layer in Fivetran" only shows up when you ask the right follow-up question.

Why ETL and Migration Screening Keeps Failing — dcoded. Manchester tech recruitment

This guide walks through how we actually vet ETL and migration specialists on live screening calls — the questions that separate genuine end-to-end ownership from surface familiarity, and how to apply that same rigour whether you're assessing Informatica Cloud, Fivetran, DBT, or Databricks-based migration work.

Before you start, have ready:

  • The specific legacy system and target platform for your migration (even partial detail changes what you should probe for)

  • A clear view of whether the role needs hands-on pipeline building, architecture and oversight, or both

  • Knowledge of whether the engagement is inside or outside IR35, since this affects the calibre of contractor you'll attract

Step 1: Separate Tool Exposure From Tool Ownership

Start by asking not "have you used X" but "what was your job within that tool." When we asked one candidate about Fivetran specifically, the honest answer was that it had only been used recently, in the candidate's most recent project — not across their "whole experience" as the CV framing implied. That distinction matters enormously for a migration project, where you need someone who has lived through a tool's failure modes, not someone who configured a connector once.

Step 1: Separate Tool Exposure From Tool Ownership — dcoded. Manchester tech recruitment

The same candidate, when pushed on DBT, gave a much stronger answer: specific companies, a specific use case (transforming data inside the warehouse using SQL), and continuity across roles. That's the kind of answer that signals real depth. A useful rule for the call:

  • Shallow answer: "Yes, I've used DBT" with no elaboration

  • Deeper answer: names the companies, the warehouse it fed into, and what the transformation accomplished

Ask every tool on the CV to be answered this way. If a candidate can give you the deeper version for Databricks but only the shallow version for Fivetran, you now know exactly where their real strength sits — and where the surface familiarity is.

Step 2: Ask Who Owned the Pipeline, Not Just Who Touched It

"Led the end-to-end pipeline creation" is a phrase candidates will claim readily. Push on it directly. In one screening call we asked plainly: "Have you led the end-to-end pipeline creation before?" and got a confident yes — but the value of that question is in how it's asked in combination with everything that came before it, including confirmed hands-on DevOps and CI/CD experience and a verified history of PySpark-based notebook transformation work.

Step 2: Ask Who Owned the Pipeline, Not Just Who Touched It — dcoded. Manchester tech recruitment

End-to-end ownership claims should be checked against three things the candidate should be able to speak to without hesitation:

  • The orchestration or scheduling layer they used (Control-M, AutoSys, CA7 Scheduler, or similar) and whether they configured it or simply monitored it

  • Whether they built the ingestion, the transformation, and the consumption layer, or only one segment of that chain

  • Who they sat with to gather requirements — stakeholders, data analysts, data scientists — before building the pipeline

A candidate with genuine end-to-end experience answers these in sequence without prompting. One contractor we screened described exactly this flow: sitting with stakeholders and data scientists to gather requirements, then constructing the pipelines in Azure Data Factory and Databricks using PySpark, with 75% of the work on Azure and 25% on the legacy SQL Server side of a live migration. That level of specificity — percentages, named tools, named phases — is what separates a true owner from someone who was present in the room.

Step 3: Test Informatica Experience Against Its Real History, Not Its Current Buzz

Informatica Cloud sits in an unusual position: it's a mature product with roots going back decades, which means genuine specialists can speak to its evolution in a way CV keyword-matchers can't. One candidate we screened had been with Informatica "since the day one," had worked through its entire journey including Informatica Cloud since 2018, and could speak fluently about both legacy file-transfer methods (SFTP, FTP) and the platform's newer mass ingestion capabilities.

Step 3: Test Informatica Experience Against Its Real History, Not Its Current Buzz — dcoded. Manchester tech recruitment

That longitudinal fluency is the tell. Ask Informatica candidates:

  • When they started working with the platform, and whether their experience spans the Cloud product specifically or only on-premise Informatica

  • What orchestration tools they've paired it with — Control-M, AutoSys, CA7 — and whether they've worked across more than one

  • Whether they've managed an offshore team as part of delivery, and at what scale

On that last point, don't just ask "have you managed a team" — ask for numbers. One candidate was able to specify team sizes precisely: as small as a one-member team up to a 25-member team at UBS, with 10 to 15 at NatWest. Vague team-leadership claims should prompt a direct follow-up for headcount and client name; specific answers like this are what genuine delivery leadership sounds like.

Step 4: Probe Databricks and Snowflake Claims for Migration-Specific Depth

"I've used Databricks" is the most common line on a migration-focused CV today, and the least informative on its own. The useful follow-up is always about migration mechanics specifically: what was the legacy system, what was the target architecture, and what did the candidate personally build.

Strong answers reference Medallion architecture by name and can speak to how long they've personally used it — one candidate confirmed four years of hands-on Medallion architecture experience and was able to explain it as a longstanding good practice that only recently became a popularized industry term, rather than a new concept. Another candidate described building a Databricks pipeline using Medallion architecture for both batch and streaming workloads via Kafka, with a concrete example: extracting e-commerce clickstream data in JSON format and loading it through Medallion layers into final reporting tables.

Also test whether the candidate can speak to the business side of a migration, not just the technical build. One senior data architect we interviewed was explicit that he was comfortable "in the weeds in delivery" but equally comfortable standing in front of the C-suite discussing technical footprint, cost, and the point at which a platform starts turning a profit — a combination that's rare and worth testing directly if your migration needs someone who can also manage stakeholder buy-in.

For Unity Catalog specifically, don't accept "I've used Databricks" as a proxy. Ask directly whether they've used Unity Catalog, since migrating to Unity Catalog is a distinct and increasingly common brief in its own right.

Step 5: Verify Cross-Platform Claims With Specific Project Detail

Many candidates now claim equal fluency across Databricks, Fabric, and Snowflake. Treat this claim as a prompt for more detail, not a disqualifier — some candidates genuinely have it. One candidate described being "equally experienced" across Fabric, Databricks, and Snowflake, and backed it up by naming a specific employer (BD Group) and project type (a machine learning project where the candidate's job was setting up the data feeding the ML models, built entirely on Databricks).

Ask for this kind of specificity every time:

  • Which employer, which project, which of the three platforms was actually used for which part of the work

  • Whether their experience with each platform is current or from a previous, now-closed project

  • Whether their migration experience includes moving data from a named legacy system into cloud storage — for example, one candidate described migrating from SQL Server to Azure Data Lake Storage Gen2 as part of a from-scratch pipeline build

If a candidate cannot anchor a platform claim to a named employer, a named project type, and a rough percentage of their time spent on it, treat the claim as unverified rather than disqualifying — then dig further on the call.

Tips & Best Practices

A few habits make these calls consistently sharper:

  • Always ask "what exactly was your job within that tool" rather than "have you used it." The Fivetran example shows how quickly this surfaces the gap between recent, shallow exposure and sustained ownership.

  • Ask for percentages. A candidate who can tell you their work was "75% Azure, 25% SQL Server" on a specific migration is giving you something concrete you can verify against their CV and references.

  • Don't assume the client's own requirements are fully formed before the call. In several of our engagements, we've been asked to compile CVs before a full job description or even the legacy system detail was confirmed by the client — which makes the specialist's ability to speak to multiple scenarios (Unity Catalog migration, Medallion architecture, cross-platform fluency) even more valuable.

  • Treat IR35 status and day rate as part of the technical conversation, not a separate admin step. Contractors moving between outside-IR35 engagements often have the sharpest, most recent hands-on migration experience, since contract work concentrates exposure to live, high-stakes migration projects.

Troubleshooting Common Screening Pitfalls

The candidate gives confident one-word answers to every tool question. This is the clearest warning sign on a screening call. Genuine specialists volunteer company names, project types, and timeframes unprompted — when a candidate needs every detail dragged out of them, assume the underlying experience is thinner than the CV suggests.

The candidate's cross-platform claims don't match the role's actual need. If your migration is specifically Fabric and Databricks-weighted, don't let broad "I've used everything" answers substitute for depth in the two platforms that matter. Ask which platform the client's project will actually require and weight your questions accordingly.

You don't yet know the legacy system you're migrating from. This happens more often than hiring teams expect, particularly when a consultancy's end client hasn't finalised scope. In that scenario, screen for breadth and architectural reasoning rather than narrow legacy-system knowledge, since the candidate may need to adapt once full scope is confirmed.

Key Takeaways

Vetting an ETL or migration specialist properly means refusing to accept tool names as proof of capability. The strongest signal on any screening call is specificity: named employers, named projects, rough percentages of time spent on each platform, and a clear account of who owned which part of the pipeline. Whether you're assessing Informatica Cloud depth, Fivetran and DBT exposure, or Databricks and Unity Catalog migration experience, the same discipline applies — ask what they actually built, not just what they've been near. Build this into your own screening process, or bring in a team that already runs this level of technical interrogation on every call.

How to Vet an ETL or Data Migration Specialist for a Legacy-to-Cloud Project

Latest News