All postsAlle Beiträge
Integrations   Jun 18, 2026 · 7 min read

Your data warehouse is not an execution system

Dein Data Warehouse ist kein Execution-System

Generated illustration for the post Your data warehouse is not an execution system

It is a natural idea, and a lot of sophisticated companies arrive at it. You already have a data warehouse. You already pipe your CRM, your finance system, your project tools, and your product analytics into it. So why not pull your OKRs and your strategic initiatives in too, build a dashboard on top, and finally have one place that shows whether the strategy is on track? The analytics team is capable, the pipelines exist, and the result is genuinely impressive to look at. Then you try to actually run the company off it, and you discover the warehouse is answering a different question than the one execution asks.

A warehouse is built to report on what happened. It is a read-optimized store of historical fact, exquisitely good at aggregation and trend and slicing the past. Execution is not a question about the past. It is a question about changing the present: who picks up this stalled initiative, where does the over-committed engineer's time go, which bet do we kill to fund the new one. Those are not reporting questions, and a system architected for reporting cannot answer them no matter how good the dashboard on top looks.

Read versus write is the whole distinction

The deepest reason a warehouse cannot be an execution system is the direction of data flow. A warehouse is fundamentally read-only with respect to the operational world. Data flows in from the source systems, gets transformed, and is presented for analysis. You do not act in the warehouse. You cannot reassign an initiative, change an owner, or commit capacity from inside a dashboard, because the warehouse is downstream of the systems where those actions happen, and its copy of the data is a reflection, not the thing itself.

Execution requires writing, not just reading. It requires a system where the objects, initiatives, ownership, capacity, dependencies, are live and editable, where changing them changes reality rather than describing it. The warehouse holds a photograph of those objects taken from the source systems, and you cannot change reality by editing a photograph. This is the same confusion as believing an export is an integration, scaled up to an entire analytics platform: the warehouse is a very sophisticated, continuously refreshed export, and it has the export's fundamental limit, which is that it observes the world without being able to act in it.

The warehouse inherits the joins problem

There is a second, subtler issue. The warehouse does not contain native strategic objects either. It contains data copied from the source tools, each of which describes its own slice in its own vocabulary, and the warehouse has to join those slices to produce a strategy view. But the join is the hard part, and the warehouse does not solve it, it just performs it in SQL instead of in a spreadsheet. The analytics team writes transformation logic that maps the project tool's "project" to the OKR tool's "initiative" to the finance system's "cost centre," and that logic is brittle, lossy, and exactly as fragile as the manual joins it replaced.

This is the every tool lies a little problem with a more expensive veneer. The warehouse dashboard looks like a single source of truth, but underneath it is the same set of approximate joins between concepts that do not cleanly map, now encoded in dbt models that only the analytics team understands. When the question gets hard, the joins are where the answer goes soft, and the people who built the models know exactly where the soft spots are.

What execution actually needs

An execution system needs the strategic objects to be first-class and native, not reconstructed from copies. It needs them to be live and writable, so that the act of managing strategy, reassigning, reprioritizing, recommitting, happens in the system itself and immediately changes the shared picture everyone sees. And it needs the connections between objects, goal to work, work to owner, owner to capacity, to be maintained structurally rather than joined on the fly from imported tables.

This is a fundamentally different architecture from a warehouse. The warehouse pulls everything into a read-optimized store to analyze the past. An execution system holds the strategic objects natively and connects the operational tools into it so that the live state of strategy is something you act in, not something you report on. The warehouse and the execution system can coexist happily, and should, because reporting on the past is genuinely valuable. But they are different organs. Asking the warehouse to run execution is like asking your accounting system to fly the plane: it has all the numbers and none of the controls.

What to do this quarter

If your company has tried to make the warehouse the strategy system, find the analytics person who maintains the strategy dashboard and ask them two questions. First, can anyone change a strategic decision from inside the dashboard, or does every action have to happen in a source system and wait to flow back through the pipeline? Second, how much of the dashboard rests on join logic that maps concepts across tools, and how confident are they in those joins when the stakes are high? The answers will locate the wall you keep hitting.

Then separate the two needs explicitly. Keep the warehouse for what it is superb at, analyzing what happened across the whole business. Stop asking it to be where strategy is executed, because execution is a write problem and the warehouse is a read system. The strategy needs a home where its objects are live and editable and connected, and where managing them is the same act as recording them, not a round-trip through a pipeline that can only ever show you where things stood the last time it ran.

FAQ

Why can't a data warehouse run strategy execution? Because a warehouse is built to report on the past and execution is about changing the present. It is read-optimized and downstream of the operational systems, so you cannot reassign an initiative or commit capacity from inside it. Its data is a reflection of reality, and you cannot change reality by editing a reflection.

Isn't a warehouse dashboard a single source of truth? It looks like one, but underneath it performs the same fragile joins between concepts that do not cleanly map, now written in SQL instead of a spreadsheet. It is the every tool lies a little problem with an expensive veneer: when the question gets hard, the joins go soft, and only the analytics team knows where.

How is this different from the export problem? It is the same problem scaled up. A warehouse is a sophisticated, continuously refreshed export, and it shares the export's fundamental limit: it observes the world without being able to act in it. Believing it can run execution is the platform-sized version of believing an export is an integration.

Should we get rid of the warehouse? No. Keep it for what it is superb at, analyzing what happened across the business. Just stop asking it to be where strategy is executed. Execution needs a system where the strategic objects are live, writable, and structurally connected, with the operational tools connected into it, so managing strategy and recording it are the same act.

Es ist eine naheliegende Idee. Du hast bereits ein Data Warehouse und leitest CRM, Finance und Projekt-Tools hinein. Warum nicht auch die OKRs? Dann versuchst du, die Firma tatsächlich darauf zu betreiben, und entdeckst, dass das Warehouse eine andere Frage beantwortet als die, die Execution stellt.

Lesen versus Schreiben ist die ganze Unterscheidung

Ein Warehouse ist gegenüber der operativen Welt grundlegend read-only. Execution erfordert Schreiben. Das ist dieselbe Verwechslung wie der Glaube, ein Export sei eine Integration, hochskaliert.

Das Warehouse erbt das Join-Problem

Das ist das jedes Tool lügt ein wenig-Problem mit teurerem Furnier. Wenn die Frage schwer wird, werden die Joins weich.

Was Execution tatsächlich braucht

Ein Execution-System braucht die strategischen Objekte nativ, lebendig und schreibbar, mit den operativen Tools hineinverbunden. Das Warehouse und das Execution-System können koexistieren, aber sie sind verschiedene Organe.

Was du dieses Quartal tun kannst

Frag die Analytics-Person: Kann jemand eine strategische Entscheidung im Dashboard ändern? Wie viel ruht auf Join-Logik? Trenne dann die zwei Bedürfnisse.

FAQ

Warum kann ein Data Warehouse keine Strategie-Execution betreiben? Weil ein Warehouse gebaut ist, über die Vergangenheit zu berichten, und Execution die Gegenwart verändert. Du kannst Realität nicht durch Bearbeiten einer Reflexion ändern.

Sollten wir das Warehouse loswerden? Nein. Behalte es für das, worin es hervorragend ist. Hör nur auf zu verlangen, dass es der Ort ist, an dem Strategie umgesetzt wird.