Data stack audit

A data stack audit that tells you what to fix first.

Most teams do not need a rebuild. They need to know which three things are causing eighty percent of the pain, and in what order to fix them. That is what this is.

When this is the right call

These are the sentences that usually precede the enquiry:

  • Two dashboards answer the same question with different numbers.
  • Nobody can say what a pipeline failure actually broke downstream.
  • The warehouse bill grows faster than the company does.
  • Every new metric request takes weeks and touches code nobody wants to open.

What is different afterwards

  • Your top business-critical metrics reconciled source to model to report, with the exact divergence points named.
  • For each divergence, a plain verdict: a real business change, or a bug. Stated explicitly so nobody over-reacts to the wrong one.
  • A ranked list of problems by revenue and risk impact, not by how annoying they feel.
  • A sequenced remediation plan your own team could execute without me.
  • A defensible view of what your current stack costs and where that money goes.

What you receive

  • Reconciliation of your top three to five business-critical metrics, source → model → report, with each divergence point located and classified.
  • A data quality baseline across nine dimensions — accuracy, completeness, consistency, timeliness, validity, uniqueness, reliability, relevance and accessibility — with KPIs you can keep tracking.
  • Observability quick wins: freshness, volume and schema monitoring on your critical models.
  • Current-state architecture diagram of sources, transforms, storage and consumption.
  • A prioritised fix backlog, each item sized by revenue and risk impact, with the guardrail that stops recurrence rather than just the patch.
  • Cost breakdown by warehouse workload, with the specific queries and jobs driving spend.
  • A written readout your non-technical stakeholders can actually read: verdict first, quantified.

How it runs

  1. Access and orientationRead-only access to the warehouse, the transformation repository, the orchestrator and the BI tool. A short session with whoever knows where the bodies are buried.
  2. Reconcile the critical numbersTake the three to five numbers the business actually acts on and follow each one end to end, from source system to dashboard cell, until any divergence is provable rather than suspected.
  3. Measure rather than assumeQuery history, job runtimes, failure rates, model lineage and storage growth. Findings are evidence-backed, and where the data cannot answer a question that is said outright.
  4. Baseline the qualityScore the nine quality dimensions and attach a KPI to each, so improvement afterwards is measurable instead of asserted.
  5. Rank and sequenceEvery finding gets an impact and an effort estimate, then an order that respects dependencies between fixes.
  6. Hand overA walkthrough with the team, and a document written so it still makes sense to them in six months.

Who this suits

  • Teams inheriting a stack somebody else built.
  • Companies about to commission a rebuild who want a second opinion first.
  • Leaders who need an outside, non-political read on where the problem really is.

Questions

Before you get in touch

Do you need production write access?

No. The audit runs entirely on read-only access to the warehouse, the transformation repository, the orchestrator and the BI tool. If a finding needs a change to prove it, that change is proposed rather than made.

What if the conclusion is that nothing is badly wrong?

Then you get that in writing, with the evidence behind it, and a much shorter roadmap. That is a useful answer — it stops a rebuild nobody needed, which is usually worth more than the audit costs.

Can the audit lead into the implementation?

It can, and often does, but the audit is deliberately sold and priced on its own so the findings are not shaped by what would make a bigger follow-on project. You are free to hand the roadmap to your own team or another supplier.

Which stacks do you cover?

Cloud warehouses such as BigQuery, Snowflake, Redshift and Postgres-based platforms, with transformation in dbt or SQL, orchestration in Airflow, Dagster or a managed scheduler, and reporting in the common BI tools. The method matters more than the vendor.

Enquire about data stack audit

Describe the situation in a few lines. You will get a straight answer about fit, timing and rough shape of the work.

The current situation and what "solved" looks like is enough. Detail can wait for the call.

Tell me what is broken.

A short call is usually enough to tell whether this is a two-week fix or a two-month one — and whether I am the right person for it.