Power BI consultancy: what a good engagement looks like

A finance team builds a revenue report in Power BI. Sales builds another. Operations builds a third. Each one is accurate on its own terms, and the three figures do not match. The board asks which number is right, and nobody can answer without a week of reconciliation. That situation is the usual reason organisations start looking for outside help with Power BI.

A Power BI consultancy is a firm that designs, builds and supports reporting on Microsoft Power BI. A good one does more than draw dashboards. It builds the data model, agrees one set of business definitions, sets up governance and access, and helps people use the reports, so the numbers stay trusted after the consultants leave.

This guide is for the person who has been asked to sort out reporting: a head of data, a finance director, or a technology lead in a mid-sized organisation. It explains what a Power BI consultancy should deliver, when you need one, where engagements go wrong, and how to brief one so you get a reporting estate rather than a pile of reports.

Shipshape Data is platform-agnostic. We build on Power BI and Microsoft Fabric, and we also build on Snowflake, Databricks, Google BigQuery and dbt. We choose the tool by fit. That means this guide is honest about where Power BI is the right choice and where it is not.

What a Power BI consultancy does

Power BI is the business intelligence tool from Microsoft. It connects to data sources, shapes the data, holds it in a model and presents it as reports and dashboards. A Power BI consultancy does this work for you, or alongside your own team, and should leave you able to run it.

The work behind the reports

• Data preparation: cleaning and shaping source data with Power Query, or upstream in a warehouse so every tool receives the same clean data.
• Data modelling: building star schemas with fact and dimension tables, and setting the relationships between them.
• Measures: writing Data Analysis Expressions (DAX) that hold business logic such as margin, use or churn.
• Report design: building reports and dashboards that answer named questions for named people.
• Security: setting row-level security so each user sees only the data they are entitled to see.
• Platform setup: workspaces, capacity, deployment pipelines and, increasingly, Microsoft Fabric and Copilot.
• Training and adoption: workshops and support so people use the reports rather than export them to spreadsheets.

Firms in this market vary. Some specialise in Power BI alone. Some work across the wider Microsoft Power Platform, including Power Apps and Power Automate. Others are broader data consultancies that use Power BI as one tool among several. The right choice depends on where your problem sits. If the reports are the problem, a specialist may be enough. If the data underneath the reports is the problem, you need a firm that works on the foundation as well. Our guide to shortlisting a UK data consultancy covers that wider choice.

When you need a Power BI consultancy

Most organisations do not need outside help to build a first report. They need help when reporting has grown faster than anyone planned for. These are the common signs.

• Several reports show different figures for the same measure, and teams argue about which is correct.
• Reports are slow to open, scheduled refreshes fail, or the model has grown too large to manage.
• One person built most of the estate, and that person has left or is about to.
• You are moving from spreadsheets or an older reporting tool and want to set the structure up properly the first time.
• You face licensing and capacity decisions, including whether to move to Microsoft Fabric.
• You plan to put AI on top of your reporting data and need governed definitions first.

When you may not need one

If your data sits in one well-run system and you need a small number of reports, a capable analyst using Microsoft's own training may be enough. Equally, if the real problem is that your source systems disagree, a consultancy that only builds dashboards will not fix it. It will build attractive reports on top of the disagreement. In that case the work starts with the data foundation, and the Power BI work comes second. We say this to prospective clients when it is true, even when it means a smaller engagement or none.

The five deliverables that matter

Dashboards are the visible part of a Power BI engagement. They are also the part that matters least on its own. Trust in the numbers depends on four things underneath them, and on whether people use the result. A good engagement delivers all five.

The data model

The data model decides what questions the reports can answer and how fast. A sound model uses a star schema: fact tables holding events such as orders or jobs, surrounded by dimension tables such as customer, product and date. The consultancy should agree the grain of each fact table, document the relationships, and keep transformations out of individual reports. Our explainer on star and snowflake schemas sets out the design choices.

The semantic layer

In Power BI, the semantic model (previously called a dataset) holds the tables, relationships and measures that reports use. This is where the business definitions live: what counts as revenue, when a customer is active, how margin is calculated. The goal is one governed definition of your business, held once and reused by every report. A common failure is a separate semantic model for each report, each with its own version of the same measure. If several tools, or AI systems, need the same definitions, the semantic layer may belong upstream of Power BI so that every consumer reads the same logic.

A dashboard is only as trustworthy as the definitions underneath it.

Governance and access

Governance decides who can publish, who owns each semantic model, which content is endorsed as the official source, and who can see which rows. It covers workspace structure, row-level security, deployment pipelines between development and production, and lineage from source to report. Without it, a reporting estate grows into hundreds of unowned reports within a year or two. Our guide to data governance explains the wider framework this sits in.

Dashboards built around decisions

Each dashboard should exist to support a named decision for a named group of people. A good consultancy starts by asking what you decide, how often, and what you need to see to decide it. The result is usually fewer pages with clearer purpose. Visual polish matters less than whether the page answers the question in front of the reader.

Adoption

Reports that nobody opens have no value. Adoption work includes training by role, a small group of internal champions, usage monitoring, and retiring the old spreadsheets and reports the new estate replaces. If you want business users to build their own reports, the consultancy should set the guardrails first. Our guide to self-service business intelligence covers how to give people that freedom without losing control of the definitions.

How Power BI engagements go wrong

Failed Power BI engagements tend to fail in the same few ways. Each one is avoidable if the brief and the scope account for it.

• Starting with visuals. The engagement opens with dashboard mock-ups, and the model is built to fit them rather than to fit the business.
• Logic inside each report. Measures are written report by report, so the same calculation exists in several slightly different forms.
• Power Query doing warehouse work. Heavy transformation sits inside Power BI, where it is hard to test, slow to refresh and invisible to other tools.
• No owner after handover. The consultancy leaves, nobody internal owns the semantic models, and the estate decays.
• Self-service without guardrails. Everyone can publish, nothing is endorsed, and the original problem of conflicting numbers returns.
• Licensing surprises. Capacity and licence costs are not modelled at the start, and the rollout stalls when the bill arrives.
• No measure of success. The engagement ends when the reports are delivered, not when a decision is made faster or with more confidence.

Most of these share a root cause. The engagement treated Power BI as the product, when the product is trusted numbers. The tool is the last step in a chain that starts with source systems and a clean, modelled store of data.

How to brief a Power BI consultancy

A clear brief gets you comparable proposals and a sensible scope. It also tells you quickly which firms understand the problem.

What to put in the brief

• The decisions the reporting must support, and who makes them.
• The source systems involved, and an honest view of their condition.
• What already exists: reports, semantic models, a warehouse, spreadsheets that people rely on.
• Constraints: licences already held, security requirements, UK GDPR obligations, timelines.
• What success looks like, stated as an outcome rather than a number of reports.
• Who will own the estate after the engagement ends.

Questions to ask a consultancy

• Where will the business logic live, and how will you stop it being duplicated across reports?
• Will you build a warehouse or transformation layer, or work directly on source systems, and why?
• How will you set up workspaces, endorsement and row-level security?
• What will our team be able to do without you at the end?
• What happens after go-live, and who supports the estate?
• When would you advise us not to use Power BI?

The last question is a useful test. A firm that cannot name a situation where another tool would fit better is selling a tool, not advice.

What drives the cost

The cost of a Power BI engagement depends far more on the state of your data than on the number of dashboards. The main drivers are these.

• The number of source systems and how clean and consistent they are.
• Whether a data warehouse exists, or needs to be built first.
• How many business definitions need to be agreed, and how much they are disputed between teams.
• The complexity of security, especially row-level rules across many user groups.
• The number of reports and the depth of each.
• The amount of training and adoption support needed.
• Whether you need support after go-live, and for how long.

Power BI licences and Fabric capacity are a separate cost from the consultancy and should be modelled at the start.

Power BI in a modern data stack

Power BI is one layer in a larger architecture. Where it sits, and what sits underneath it, matters more than which visuals it offers.

Where it sits in the architecture

In a well-designed stack, data moves from source systems into a warehouse or lakehouse. There it is cleaned in stages, often described as bronze, silver and gold layers in a medallion architecture. A semantic layer then holds the agreed business definitions, and Power BI reads from it. This keeps heavy transformation out of the reporting tool and lets other consumers, including AI systems, use the same governed data. Our guide to business intelligence architecture walks through each layer, and our explainer on data warehousing covers when you need a warehouse.

Where Power BI fits well

• Organisations already using Microsoft 365 and Azure, where licensing and identity are in place.
• Estates moving to Microsoft Fabric, where storage, engineering and reporting share one platform.
• Finance and operations teams that work heavily in Excel and need reports that connect to it.
• Teams that want reports inside Microsoft Teams and SharePoint.

Where another tool may fit better

Some situations are a different shape. If you need analytics embedded in your own product for external users, the licensing and embedding model of other tools may suit better. If your estate is standardised on Google Cloud, a tool built around that platform may be a closer fit. Power BI Desktop runs on Windows only, which matters for analyst teams on other operating systems. None of these is a verdict on the tool. They are questions of fit. Our review of the best business intelligence tools for enterprises compares the main options.

Power BI, Copilot and AI

Microsoft now adds Copilot to Power BI, so people can ask questions of their data in plain language. This makes the semantic layer more important, not less. AI reads the definitions it is given. Over a clean, governed model it gives useful answers. Over duplicated, conflicting measures it gives confident wrong ones, faster. The foundation work is what makes the AI features worth turning on.

AI amplifies whatever it sits on. Clean definitions first, Copilot second.

Where this leaves you

A good Power BI consultancy leaves you with a documented data model, one governed set of business definitions, clear ownership and access rules, dashboards built around decisions, and people who use them. Dashboards alone do not fix conflicting numbers. The work that fixes them sits underneath the reports.

Start with the brief. Name the decisions, the sources and the owner. Ask each firm where the logic will live and when it would advise against Power BI. Choose the firm whose answers show it understands your data, not only the tool. If the foundation is not ready, fix that first, and let the first trusted report pay for the next.

Our data consultancy service covers data strategy, engineering, analytics and governance on a governed foundation, including Power BI work where Power BI is the right fit and other platforms where it is not. We scope honestly, build the model and the semantic layer before the dashboards, and stick around past go-live. If you are not ready for a conversation, the free AI readiness assessment scores your data position on the spot. Talk to us and we will tell you plainly whether you need a consultancy, and if so, what to ask for.

Frequently asked questions