Foundation • Activation Sprint • Acceleration

Three doors into the same work.

Every engagement starts with one fixed piece of work: a defined scope, a stated duration, and a document at the end you can act on without us. Which door you take depends on where the reporting currently breaks.

The route

Each door ends somewhere you could stop.

  1. Step 1

    Foundation

    Make the numbers agree.

    One definition per metric, one place it is computed, and an owner for each.

  2. Step 2

    Activation Sprint

    Turn them into a decision.

    Reporting aimed at a question someone has to answer, rather than at the data that happened to be available.

  3. Step 3

    Acceleration

    Put AI on top of both.

    An assistant that reads a model it cannot misread, and says the same thing twice.

Most engagements start at Foundation and some stay there, which is the right outcome rather than an unfinished one. What the tier above inherits is not a licence or a platform, it is a set of definitions somebody has agreed to.

01/03

Foundation

A written second opinion on your reporting estate.

It ends in a document rather than a proposal.

Who it is for

Finance teams who inherited a reporting landscape nobody designed. Numbers that disagree between two reports, a month-end that runs on spreadsheets nobody can audit, definitions that change depending on who you ask. You do not need a verdict on the tooling, you need to know which of those is structural and which is habit.

What you receive

Scored findings
Each of the 12 readiness questions answered for your estate, with the evidence behind the answer.
Data-gap list
What the reporting you want cannot be computed from today, named rather than described.
Ownership map
What is owned today and by whom, and what is owned by nobody at all.
Staged plan
The work ahead in stages, each carrying an effort band rather than a guess.

How the 3 days run

  1. Access and conversations4 to 6 people: whoever owns the numbers, whoever pays for the capacity, whoever builds the reports, and one person who actually uses them.
  2. AnalysisAn inventory of the models in use, where the capacity goes, where the same metric is defined twice and differently, how refreshes are designed, and who owns which workspace.
  3. WritingThe document, and a call to walk you through it.

What you bring

  • Read-only access to the tenant on day one, arranged before we start rather than during.
  • The 4 to 6 conversations booked, including whoever owns the numbers.
  • If the estate is large, we agree one steering area in the first call and scope the 3 days to it.

What it unlocks

The gap list is what makes an Activation Sprint worth running, because the report we build in it is then aimed at a decision rather than at whatever data happened to be available.

Proven in practiceA CFO cockpit over a group that closed its books separately

  • Semantic modelling
  • DAX
  • Power BI
  • Microsoft Fabric
  • Excel migration
Talk about a Foundation Review

02/03

Activation Sprint

The report the decision actually needs, before anyone builds a pipeline.

A target report on sample numbers, agreed in 2 weeks. The build that follows is quoted once both of us know what the data holds.

Who it is for

Teams who already know something is missing from their reporting and have watched a build run for months without answering the question that started it. The expensive mistake is not a slow build, it is finishing one nobody opens.

What you receive

A clickable target report
Real structure, sample numbers, and every screen aimed at the decision named in the workshop.
Metric definition sheets
One page per surviving metric: what it means, who owns it, and how it is computed.
Data-gap list
Which of those the data you hold can produce today, and what is missing for the rest.
Staged build plan
The real build in stages, with an effort band on the first one.

How the 2 weeks run

  1. A half-day workshopWhich decision is this for, and what question sits behind it? Then one test on every proposed metric: what would you do differently if this number were red? No answer, no metric. 14 wishes usually become 4.
  2. Week 1: the target reportClickable, with real structure and sample numbers. You see whether it answers the question in days rather than months.
  3. Week 2: definitions and gapsA one-page definition sheet for each surviving metric, and an honest list of which can be computed from the data you have today and which cannot.
  4. HandoverA staged plan for the real build, with an effort band on the first stage.

What you bring

  • The person who owns the decision, in the room for half a day. Without them the sprint produces a good report aimed at nobody.
  • One decision to aim at. Two would be two sprints.

Why the scope holds

Nothing in the 2 weeks depends on your data being ready. The prototype runs on sample numbers, the definitions are decisions rather than extracts, and the gaps are a finding rather than a fix. The build that does depend on your data is quoted afterwards, when both of us know what is in there.

Proven in practiceFinance reporting that keeps up with the day

  • Executive dashboards
  • KPI definitions
  • Self-service BI
  • Governed templates
Talk about an Activation Sprint

03/03

Acceleration

A scored test of whether your model can carry an assistant yet.

“Not ready, and here is what would make it ready” is a result rather than a failure.

Who it is for

Organisations who licensed an AI assistant, pointed it at their reporting and found the answers plausible but unverifiable. The question is not whether the assistant is good. It is whether the model underneath it says the same thing twice.

What you receive

A scored answer sheet
All 20 questions, each marked correct, wrong or refused, with the answer that was given.
A cause per failure
Why each wrong answer was wrong, traced to the model rather than to the assistant.
A written verdict
Ready, ready with named fixes, or not ready, dated and tied to the version of the question set behind it.
An access-rules finding
Whether the assistant respects who is allowed to see what, tested rather than assumed.
The use case itself
Where the model carried it, the working use case stays with you.

How the 2 weeks run

  1. Days 1 and 2: one use case, 20 questionsWe pick the use case with you and write 20 questions a real user would ask. That set is the scope, and it is agreed before anything is built.
  2. Week 1: the grounding testThe assistant is pointed at your existing model and every one of the 20 is scored: correct, wrong or refused. Each wrong answer gets a diagnosed cause.
  3. Week 2: the verdictReady, ready with named fixes, or not ready. Where it scores well, you keep the working use case.

What you bring

  • The assistant licences you already hold, and access to the model it should read.
  • One business area willing to judge the answers, since only they can say whether a number is right.

Why the scope holds

20 questions is 20 questions whatever shape the model is in. A weak model produces a shorter engagement rather than a longer one, because most of the 20 fail early and the diagnosis is the deliverable. Where the verdict is “not ready”, the named fixes are Foundation work, and we quote them as stages rather than as a rescue.

Proven in practiceThe semantic model that became the AI on-ramp

  • Copilot
  • Fabric Data Agents
  • Row-level security
  • Forecasting
Talk about an Acceleration Pilot

Selected engagements

What was broken, and what replaced it.

Written out rather than summarised into a logo wall, and anonymised by sector and size, because the confidentiality that protects these clients would protect you too. Where an outcome is not published below, it is because nothing exists that would let you check it, which is a better answer than a number that sounds good.

  • Multi-brand retail

    Foundation then Acceleration

    The semantic model that became the AI on-ramp

    Situation
    A retail group running several brands, each with its own systems, its own reporting habits and its own answer to what a margin is.
    What was broken
    Every brand could report on itself and the group could not report on any of them together without a week of manual work. That made group-level questions expensive enough that people stopped asking them, which is the quiet cost nobody puts in a business case.
    What we built
    One governed semantic model spanning the brands, with a single definition per metric and the brand differences that genuinely exist modelled rather than argued about. Then, and only then, the AI work on top of it, which continues as a retainer.
    Where it left them
    The textbook path from clean data to defensible AI, and the only one of these engagements that has walked all of it. The model built for consolidated reporting turned out to be the thing that made AI answers trustworthy, which is the argument this whole site makes, running in production at one client.
    • Semantic modelling
    • AI enablement
    • Retainer

  • Hospitality & real estate

    Foundation

    A CFO cockpit over a group that closed its books separately

    Situation
    A holding group across hospitality and real estate, with entities that each closed their own books and a CFO who had to answer for all of them.
    What was broken
    Reporting was fragmented across systems that had grown up separately, so consolidation happened in spreadsheets, by hand, by people who were good at it and could not be replaced.
    What we built
    A ten-month rollout replacing that with a governed platform, and a CFO cockpit on top of it that consolidates every entity in the group on shared definitions.
    Where it left them
    One number per metric, arrived at the same way in every entity, and a consolidation that is a refresh rather than a fortnight. What it moved in hours and headcount is not published here, because nothing we could point to would let you check it.
    • Microsoft Fabric
    • Databricks
    • Semantic model

  • Industrial manufacturing

    Foundation

    Governed reporting across a production network

    Situation
    An industrial manufacturer reporting across a production network of sites, each with its own view of what was running.
    What was broken
    Site-level reporting existed and network-level reporting did not, and the access question sat on top of it: the numbers that make a network comparable are also the numbers a site manager should not see for a neighbouring site.
    What we built
    Unified governed reporting across the network, with enterprise row-level security carrying the access rules, so every stakeholder sees exactly their own numbers and only those.
    Where it left them
    One reporting layer for a network that had been reporting in pieces, with the security model doing the work that used to be done by simply not sharing the file.
    • Row-level security
    • Power BI
    • Governance

  • Omnichannel e-commerce

    Activation Sprint

    Finance reporting that keeps up with the day

    Situation
    An omnichannel retailer whose finance leadership was steering on yesterday’s numbers in a business that changes within the day.
    What was broken
    The reporting was correct and late. By the time the daily figures landed, the decisions they informed had either been made on instinct or postponed, which is the version of this problem that never shows up as a data-quality complaint.
    What we built
    Near-real-time financial reporting on the governed model, so the figures update as the business moves rather than after it has stopped.
    Where it left them
    Finance leadership sees performance as it happens, not a day later.
    • Direct Lake
    • Eventstream
    • Finance BI

Other engagements span SaaS on Snowflake, self-service BI in logistics and long-standing German Mittelstand clients. Some meant rescuing stalled Power BI rollouts; others, rebuilding KPI and budgeting frameworks. Those are not written up here, and the detail is available on request under NDA.

Who does the work

I sat on your side of the table first.

Before the BI years I spent four years inside controlling, closing books and defending the numbers in them, which is where you learn that a reporting problem is almost never a reporting problem. It is a definition nobody owns. Two people build the same figure from the same system, both are right under rules neither of them wrote down, and the meeting that was meant to decide something opens with a reconciliation instead.

Eleven years of BI consulting across Europe came after that, in Power BI, DAX, Microsoft Fabric and the architecture underneath them. Plenty of people can write the DAX. The part that is scarce is knowing what the number is supposed to mean before anyone writes any, and that is the part I was hired for long before I was hired to build models.

So I lead every engagement myself, from the first call through to handover, and you are never explaining your business twice.

Niklas Hoeppener

Niklas Hoeppener

Founder • Ariavor

Eleven years
BI consulting across Europe
Power BI • DAX • Microsoft Fabric
And the data architecture underneath them
Remote
For companies across DACH and the rest of Europe
English and German
Engagements run in either

All three end the same way.

A document, and a staged plan with an effort band on each stage. That is deliberate. You can stop after any one of them and act on what you have, which is what makes a package a purchase rather than the first instalment of something open-ended. It also means anything that follows is quoted with knowledge instead of a guess.

Not sure yet

Not sure which door you need?

12 questions, 10 minutes, no email required. Under each one, what a “no” costs you.

Readiness check

Is your reporting ready for what you want to do next?

Twelve questions about how your data actually works today. No email and no sign-up: the result appears on this page and goes nowhere else.
Questions
12
Minutes
10
Stages
3

Part one of three

Foundation: is the ground solid?

Whether a number means one thing, and whether anyone owns that meaning.

  • Question 1 of 12

    Do revenue, margin and headcount mean the same thing in every report your leadership sees?

  • Question 2 of 12

    Can you name the person, not the department, who is allowed to change the definition of a core KPI?

  • Question 3 of 12

    Can every report showing financial figures be reconciled to the statutory accounts without a manual step?

  • Question 4 of 12

    If a number looks wrong, can someone trace it back to the source system and the load that produced it within an hour?

Part two of three

Activation: is any of it being used?

Whether the reports you built change what anyone does on a Monday.

  • Question 5 of 12

    Do you know how many of the reports in your workspace were opened in the last ninety days?

  • Question 6 of 12

    Can you name a decision that was made differently last quarter because of a report?

  • Question 7 of 12

    When someone in the business needs a new view, can they build it themselves from a governed model, without raising a ticket?

  • Question 8 of 12

    Does a refresh failure reach a named person rather than a shared mailbox?

Part three of three

Acceleration: could AI safely touch this?

Whether an assistant answering from your data would be right, and whether you would know if it were not.

  • Question 9 of 12

    Is there one semantic model per steering area, rather than one per report?

  • Question 10 of 12

    Are your KPI definitions written down somewhere a person, or a machine, can read them, outside the DAX?

  • Question 11 of 12

    Would an AI assistant querying your model be bound by the same access rules as the person using it?

  • Question 12 of 12

    If an AI answered from your data today and was confidently wrong, would anyone notice before the decision was made?

Scope and price

The questions that come up first.

What does it cost?

Price is a conversation rather than a page. Every package has a fixed scope and a fixed duration, so the number is fixed too, and we tell you what it is on the first call once we know which door you are standing at.

What if we want to stop afterwards?

Then you stop. Each package ends in a document and a staged plan you can hand to your own team or to another firm. That is the design rather than a concession.

What if our data turns out to be a mess?

It usually is somewhere, and none of the three packages can be blocked by that. Finding out precisely where is the deliverable, which is why the work that genuinely depends on your data gets quoted after we have looked rather than before.

Free consultation

Know Where Your Data Stands In 30 Minutes.

A free 30-minute consultation. No pitch, no invoice. You’ll leave knowing whether your data is ready for AI, and whether we’re a good fit.

Book Your Free Consultation(opens in a new tab)