When do you need a data warehouse? (Hint: often not)

Data8 min read

A data warehouse (BigQuery, Snowflake, Redshift) is a central place where data from all your tools comes together for analysis. SMEs often think they need one, but in 90% of cases, it's overkill. You only truly need one if (1) you need to combine data from 5+ sources, (2) you need to store historical data >2 years, (3) dashboards are no longer real-time enough, or (4) you want to train AI/ML models. For everything else, Google Sheets, Airtable, or a directly connected BI tool works fine.

What a data warehouse is — and what it isn't

A data warehouse is a central database optimised for analytics (not for running an application). Typical tools: Google BigQuery, Snowflake, Amazon Redshift, Microsoft Azure Synapse, or self-hosted PostgreSQL with a dwh-schema.

Data from various sources (HubSpot, Shopify, Stripe, Salesforce, custom app) is regularly synchronised to the warehouse, often via ETL/ELT tools like Fivetran, Stitch, or Airbyte. Afterwards, you analyse everything with SQL or via a BI tool (Looker, Metabase, Tableau).

It is not the same as a reporting tool. A warehouse is the storage; a dashboard tool displays the insights.

When do you truly need one?

The four signals that, in practice, justify a data warehouse:

  • 1. Combining data from 5+ sources. If you need to build dashboards that combine HubSpot data + Shopify data + Stripe data + inventory, it becomes unsustainable without a warehouse.
  • 2. Storing historical data >2 years. Many SaaS tools only store 13-25 months of detailed data. For year-on-year analyses, you need your own storage.
  • 3. Custom metrics across multiple systems. "Customer Lifetime Value per acquisition channel" requires data from marketing (channel), sales (first deal), and accounting (total revenue over years).
  • 4. AI/ML models or forecasting. For predictive modeling (churn prediction, demand forecasting), you need structurally accessible datasets.

When absolutely not

SME organisations that do NOT need to consider a warehouse:

  • One or two sources, <1M records. HubSpot Reports + Looker Studio for Google Analytics — done.
  • <20 employees, no data analyst. Who will manage that warehouse? Without an owner, it becomes a dead cost.
  • No clear analytical question. Building a warehouse "just in case" leads to an empty data bucket without insight.
  • Budget <€2,500/month for data stack. A serious warehouse + ETL + BI tool costs at least that in licenses + time.

Alternatives for SMEs

  • Google Sheets as a hub. With Supermetrics ($69+/month) or Funnel.io, you automatically import data from HubSpot, GA4, Ads, Facebook into Sheets. This is sufficient for 80% of SME reports.
  • Airtable. Combine with integrations (Make, Zapier) for a lightweight DWH-like setup. Works up to ~100k records.
  • Native HubSpot Reports + Looker Studio. Virtually free, covers much of what SMEs need.
  • BigQuery only. For the intermediate category: only a data warehouse, with GA4 BigQuery export and HubSpot API via Fivetran. ~€100-€300/month starting costs, scalable without DWH fanfare.

Cost reality (2026)

  • BigQuery: storage $0.02/GB/month, queries $5/TB (Pay-as-you-go) or $2,000+/month flat-rate. For SMEs: €50-€500/month.
  • Snowflake: $2-$3/credit, typical SME usage €300-€2,000/month.
  • Fivetran (ETL): $120-$1,500/month for SME volumes (first free for <500k MAR).
  • BI tool (Looker/Tableau/Metabase): €0-€2,000/month depending on choice.
  • Data engineer (internal or external): €600-€1,500/day or fixed contract.

Total SME data stack: realistically €500-€3,000/month operational, plus €15,000-€60,000 setup.

Common mistakes

  • Building a warehouse without an analytical question. "We have data" is not a reason. "We want to be able to analyse X" is.
  • Starting too early. With <20 employees, a warehouse is usually a waste. It only becomes a real accelerator at 50+ employees.
  • Not assigning an owner. Without a data analyst or RevOps role, the warehouse will be abandoned after 6 months.
  • Wanting real-time where daily suffices. Real-time data is 5-10× more expensive than daily batch. Almost never needed for SMEs.
  • Syncing all data. A warehouse that pulls ALL data from all tools becomes unmanageable. Start with the 3-5 most important sources.

Frequently asked questions

How much does a data warehouse really cost for an SME?

Realistic lower bound for a functional system: €500/month operational, plus €15,000-€30,000 one-off for setup. Below that, there are many hidden costs such as data engineering time.

How long does it take to set up an SME data warehouse?

With a specialist partner: 8-16 weeks from kick-off to the first usable dashboard. Doing it yourself without experience: 6-12 months, with a high risk of a partial solution.

What is the difference between ETL and ELT?

ETL = Extract, Transform, Load (data is processed before storage). ELT = Extract, Load, Transform (data is first stored, then processed). Modern warehouses work ELT-style with dbt for transformations.

Which warehouse is best for SMEs?

BigQuery is usually the pragmatic choice: free tier for experimenting, Pay-as-you-go scales nicely with usage, integrates well with the Google stack that many SMEs already have.

Do I need dbt?

For SMEs: only if you combine 3+ sources with complex transformations. dbt (data build tool) makes SQL transformations manageable as version-controlled code. Requires a SQL-skilled data engineer.

What is a data lake and how does it differ from a warehouse?

A data lake stores raw unstructured data (logs, JSON, files). A warehouse stores structured data for analysis. SMEs rarely need a data lake; for that, you truly need big data volumes (TBs per day).

Ready to get started?

Read our approach to getting data in order, or schedule a brief introduction — we'll take a look without obligation and provide an honest estimate of scope, costs, and lead time.

Back to Journal →
Chat