← All services
Power BISemantic ModelsPerformance

Semantic Model Health Check

Is the model underneath your reports helping you — or quietly creating technical debt?

A focused review of the semantic layer behind your Power BI solution, with attention to both correctness and maintainability.

Start with what is actually happening.

  • The model has more tables than anyone can comfortably explain.
  • Relationships work until one more requirement appears.
  • Calculated columns and helper tables keep accumulating.
  • Business logic is scattered between source systems, Power Query and DAX.
  • Nobody is sure whether the current design will scale.

Good work starts with better questions.

01

Does the model represent the business process correctly?

02

Are relationships and filter directions doing what you think they are doing?

03

Where is unnecessary complexity?

04

What belongs in the semantic model — and what should move elsewhere?

The scope follows the problem.

I do not force every engagement through the same checklist. These are the areas that may matter; the agreed scope determines how deep we go.

Fact and dimension design

Grain and star-schema structure

Relationships, cardinality and filter direction

Many-to-many and bridge-table patterns

Calculated tables and calculated columns

Storage modes and model size

High-cardinality columns and compression opportunities

Consistency of business logic across the model

Something you can use after the conversation ends.

The output should help you make a decision, change a practice or move the work forward. It should not exist simply to prove that work happened.

01

Model Findings

The most important design, performance and maintainability observations.

02

Relationship Review

A clear view of filtering behaviour, ambiguity risks and modelling trade-offs.

03

Simplification Opportunities

Tables, columns or patterns that can be consolidated, removed or moved upstream.

04

Recommended Direction

What I would change in the model and why.

05

Walkthrough Session

A technical discussion of the findings with the people who own the model.

Enough structure to make the work clear. Not so much process that the process becomes the work.

01

Context

We start with a short conversation about what is happening, what matters and what has already been tried.

02

Review

I analyse the relevant parts of the solution or operating model at the depth agreed for the engagement.

03

Prioritise

I organise observations around impact, risk and usefulness — not around how many issues I can find.

04

Walkthrough

We go through the findings together, challenge assumptions and discuss trade-offs.

05

Next step

You decide what happens next: implement internally, continue together or stop because the focused engagement was enough.

I keep the client-side time demand deliberately light. A focused assessment usually needs a kickoff conversation, access or materials, and a final review session. We agree the exact involvement before we start.

This is likely to help if...

  • The model has evolved organically over time.
  • You inherited a model and need to understand its risks.
  • Relationships or calculations are becoming difficult to reason about.
  • You want to simplify before adding more requirements.
  • The problem sits mainly in report UX rather than the semantic model.
  • You need implementation capacity rather than an independent review.

If the problem is real but the format is wrong, that is useful to know early. I would rather redirect the conversation than force the challenge into the wrong service.

This does not automatically lead to another engagement.

Sometimes the focused piece of work is enough. If it reveals a useful next step, these are some of the directions that may make sense.

Is the model underneath your reports helping you — or quietly creating technical debt?

Bring me the problem, not a polished brief. A few sentences about what is happening and what you would like to change are enough to start.

Tell me what's happening

Send me a message.

You do not need a polished brief. Whether the topic is Power BI, Fabric, AI, team training, mentoring, coaching or simply a challenge that needs untangling, leave a few details and we can start from there.

SIGNALMESSAGECONVERSATION

AI & TRANSPARENCY

AI supports selected parts of my work — it does not replace my judgement, responsibility or relationship with the client.

I use AI tools for tasks such as research, organising information, structuring content and preparing materials. I treat AI as support for my work — not as an autonomous author or decision-maker. Anything I publish or use professionally remains my responsibility.

In coaching, I do not use AI to automatically assess, diagnose or make decisions about a client. AI does not replace attention, confidentiality or human responsibility for the coaching relationship.

AI may support the process. Human responsibility remains.