- contact@insightaiviz.com
Trusted metrics, dashboards you can review, and what migrating off Tableau really involves.
Most of a BI licence bill is people who never build anything — and rationing their access costs more than the invoice does.
Two teams bring two revenue numbers to the same meeting. Nobody did anything wrong — this is what happens when a definition lives inside a file.
Between half and four fifths of a typical estate was not opened by anyone in ninety days. The cheapest audit available, and it changes the migration scope.
A BI renewal arrives as one number and a request for a signature. The line items underneath answer questions most organisations have never asked.
Every chart can show that a number fell. None says why. What it looks like when the dashboard answers in one click — worked out from the rows, not guessed by a model.
Every BI vendor now has a chat box. The difference is not the model — it is what the model is allowed to touch, and what happens when it does not know.
A dashboard waits to be opened. What it takes for the numbers, the reason they moved and what to do about it to arrive first — and why the email only carries what was checked.
Wrong joins, drifting metrics, invented fields. The mechanism that makes those three failures impossible rather than unlikely.
Why a dashboard should be something you can read, compare, version and let an assistant draft — and what the simpler surface costs.
Someone challenges a number in a board meeting. How long until you can say where it came from, what it included, and who was allowed to see it?
On one test page the query engine accounted for 34 milliseconds and the page took 2.27 seconds. Where the rest of the time goes.
Between a question and a database sits ordinary, deterministic, tested code. It is where the security properties actually live.
Copying has a reputation. Why a governed copy is the right default for reporting, what a staleness budget is, and when to use a live connection instead.
Why a chart is a small object saying what to show rather than code that draws it — and why there is no dual-axis chart, which is a decision.
Eleven days of assembling numbers, and the one question in the meeting that nobody can answer on the spot. What changes when every figure carries its own working.
The largest group of BI users, the least served. What a dashboard has to do for someone who will never open a builder and should not have to.
Four days of re-cutting the same report six ways, and one day of actual analysis. Which half of the queue an assistant takes, and which half it makes worse.
Access set per workbook is a promise. Two reorganisations later nobody can say what the promise was. Setting rules on the data instead, and proving they held.
One container, one bucket, one database. What to back up, what to watch, and what wakes someone at two in the morning — which is almost nothing.
The reference deployment, and the one we run ourselves. What lands where, what it costs, and the two places an ECS deployment goes wrong.
None of the Microsoft warehouses can be queried live, so Azure is an extract-first deployment — which is usually the architecture you wanted anyway.
BigQuery cannot be attached and queried live. Against per-byte pricing that is not a limitation, it is a saving — and the arithmetic is worth seeing.
Most enterprises acquired their way into multi-cloud. What ports, what does not, and the one property that decides whether you are portable or merely hopeful.
Three cases where building your own is genuinely right, and the question that decides it — which is neither cost nor capability.
What you run in the meantime, who has to be in the room, and where it stalls — which is the same place every time, and it is not technical.
Four numbers worth tracking, one almost nobody tracks that is usually the largest finding, and one we refuse to provide.
The data is the easy part. The years of decisions inside the workbook are the migration. What converts, what does not, and how to find out before you commit.
Where a Tableau migration is won or lost. What translates, what gets flagged, what cannot be settled — and the category we refuse to translate on purpose.
A screenshot of the dashboard you have is a better specification than any brief. What gets read, what gets rebuilt over your own data, and what it refuses to pretend.
Every migration runs two estates in parallel. What the overlap is for, how to shorten it honestly, and the four rules that stop it becoming permanent.