Cantan Group
Contact us

Power BI

Power BI consulting in Toronto and across Ontario: one governed semantic model over your warehouse, ERP, CRM and the spreadsheets that remain, measures in DAX, row-level security, dashboards designed for the decision they serve — and the Redshift, AWS Glue and Python pipeline work underneath them. Built accessible, refreshed daily, owned by you.

Pipeline · Pythonnightly · reconciledSourcesERP · CRM · S3 · spreadsheetsGlue jobsvalidate · transform · loadRedshiftthe warehouse · history04:00Semantic model · governedrefreshed daily · owned by youTables · relationshipsone version of the truthMeasures · DAXmargin · backlogRow-level securityeach user, their ownExecutivephone · board deck06:00Operationsdaily · by sitePaginated reportsscheduled · emailedthe load · the refreshgoverneddecision-ready
One model, governed, fed by a pipeline we write. The load lands at four, the refresh at six, and every dashboard is already right — every person seeing only their own.

The model is the product

Most Power BI trouble is not a dashboard problem. It is three dashboards, built by three people, each with its own definition of revenue, arguing in a meeting about which one is right. The fix is not a better visual. It is one governed semantic model — tables and relationships built once, measures written in DAX and documented, row-level security so each person sees their own — that every report reads from. Build that first and the dashboards become the easy part.

We build the model over the systems you already run, Salesforce among them, connect the sources properly, and design the dashboards around the decisions they serve rather than the visuals the tool makes easy: an executive page that is right on a phone every morning, an operations view by site or by week, and the scheduled report your board or your funder expects. We build with Power BI’s native visuals where they serve the decision and write custom HTML visuals where they do not, to the same standard, so a report is never limited to what the gallery offers. Then we set up the governance, monitor the refresh, and train your team to extend what we built.

The hard part is upstream

The dashboard leadership reads at six is the last hundred metres. Behind it, on the engagements where the data is large or regulated, sits a warehouse and a pipeline: Amazon Redshift holding years of history, AWS Glue jobs written in Python landing the nightly loads, a Jupyter notebook that began as an analyst’s question and has to become a scheduled pipeline with tests, logging and someone to call. Most Power BI consultancies start where the data already is and work around what is wrong with it. We write the layer underneath — the Glue code, the Redshift SQL, the validation that refuses a bad file, the reconciliation that ties the warehouse to the systems it came from — because a model is only as governed as the data it is given. It is the same Python practice that builds our pipelines and Odoo modules, so the people who model the data are the people who moved it.

Built accessible, because it is a website

A Power BI report is a web application, and the people who read it include people who use screen readers, keyboards and high contrast. Colour is never the only carrier of meaning in our dashboards, contrast is measured, reading order is set on purpose, and every visual has a data table behind it. It is the same accessibility practice we bring to everything else, applied to the page your organization reads most.

What we deliver

The model is the product; the dashboards are how you read it

One semantic model

Tables, relationships and measures built once, governed, and shared by every report — one version of the truth instead of a dashboard per department, each with its own definition of revenue.

Measures in DAX

Margin, backlog, days sales outstanding, utilization — the numbers a decision turns on, written as measures that are correct at every level of the model and documented so the next analyst can trust them.

Sources connected properly

Redshift, Salesforce, Quickbase, SQL, SharePoint, Dynamics, Excel and partner APIs, with the transformation done where it belongs and refreshed on a schedule you set.

The pipeline under the dashboard

Where the data is large or regulated, the hard part is upstream: Amazon Redshift holding the history, AWS Glue jobs landing the nightly loads, a notebook that has to become a scheduled, tested pipeline. We write that Python and SQL — the Glue code, the validation that refuses a bad file, the reconciliation that ties the warehouse to the source — so Power BI reads a warehouse that is right. Same Python practice, same team.

Dashboards designed for the decision

An executive view that works on a phone, an operations view by site or by week, and nothing that exists because the tool made it easy. Every visual answers a question someone actually asks.

Native visuals, or custom ones we write

Power BI’s native visuals wherever they answer the question, and where they cannot — a process map, a layout your regulator prescribes, a chart the gallery does not have — a custom visual we write with the Power BI visuals SDK: an HTML widget designed to the same standard as the rest of the report, accessible, keyboard-operable and right on a phone. Both read from the same governed model, and you own the code.

Governance and row-level security

Each user sees their own region, program or client and nothing else; workspaces and deployment pipelines separate development from what the board reads; refreshes and failures are monitored.

Paginated reports and distribution

The board pack, the monthly statement, the funder’s report — pixel-precise, scheduled and delivered, updated by the same model that drives the dashboards.

How an engagement runs

  1. Decide what the decisions are

    Not what data you have — what you need to decide, how often, and who. Two weeks; the output is a list of measures with definitions everyone signs.

  2. Model

    Sources connected, the semantic model built and tested against known totals, measures written and documented. The model is reviewed with your finance team before a single visual exists.

  3. Design the reading

    Dashboards built in the browser with the people who will use them, iterated weekly, checked for accessibility — colour, contrast, keyboard, screen-reader order — as we go.

  4. Govern and hand over

    Security, refresh, monitoring and deployment set up; your team trained to extend the model and build their own reports on it. We transfer knowledge, not dependency.

Who this is for

Operators with the data and no picture

Organizations running Salesforce, Dynamics or another system that holds the data and cannot show it: the numbers exist, the decisions still happen in a spreadsheet.

Public-sector and funded programs

Agencies, institutions and non-profits that report to a board, a ministry or a funder on a schedule, and need the report to be right and repeatable. We are an Ontario Vendor of Record.

Leadership teams

Executives who want one page on a phone that is correct every morning, and finance teams who want to stop being asked why two dashboards disagree.

Teams with a warehouse and no engineer

Organizations with data in Redshift, S3 or a lake, an analyst with a notebook that works on their laptop, and nobody to turn it into a pipeline that runs every night, reconciles, and tells someone when it fails.

Questions we're asked

Power BI or a spreadsheet?

A spreadsheet is a fine analysis tool and a poor reporting system: it has no refresh, no security, no shared definitions and no audit trail. When the same numbers are produced every week for people who act on them, they belong in a governed model. We will also tell you when a spreadsheet is still the right answer.

Do we need Power BI Premium or Fabric capacity?

Usually not to start. Per-user licensing covers most organizations; capacity becomes worthwhile at scale, for large models, or for embedding reports in your own applications. We size the licensing with you and prefer the cheaper answer that works.

Our data is in Redshift and a data lake, not a business system. Can you work with that?

Yes, and it is often the harder and more valuable half of the engagement. We write the AWS Glue jobs in Python that land and transform the data, the Redshift SQL that shapes it for reporting, and we turn the notebook an analyst wrote into a scheduled pipeline with tests, logging and reconciliation. Power BI then reads a warehouse that is right rather than compensating for one that is not — and the refresh at six lands because the load at four did.

Can our own team build reports on the model?

That is the point of a semantic model. Once it exists, analysts build reports on shared, governed measures instead of rebuilding the logic each time. Training is part of every engagement.

Do you build custom visuals, or only use Power BI's own?

Both. Power BI’s native visuals cover most decisions and we use them first, because they are the cheapest to build and the easiest for your team to maintain. When a decision needs a visual the gallery does not have, we write one with the Power BI visuals SDK: an HTML widget with the same accessibility, contrast and keyboard behaviour as everything else we ship, reading from the same governed model. It is designed, not assembled — the same discipline we bring to a website — and the code is yours.

Are the dashboards accessible?

They can be, and ours are designed to be: colour that never carries meaning alone, contrast that passes, keyboard and screen-reader order set deliberately, and data tables behind the visuals — the custom visuals we write included, tested the same way. Power BI’s own accessibility features are only as good as the author who uses them, which is why we include it in the design step.

How is this priced?

A fixed fee for the model and the first set of dashboards, after the two-week definition step, and a retainer if you want us to keep extending it. The variables are the number of sources and how clean they are — which is why we look at the data before we quote.
Contact us

Have a hard problem and a budget?

Tell us what's driving it. You'll talk to the people who do the work.

A sentence or two is plenty: the problem, the system, the deadline.