Think Build Implement Repeat
London, UK +44 7367 067226
WhatsApp FOLLOW f in X
  1. Home
  2. Blog
  3. Machine Learning Without Building a Data Warehouse First
Data & Scraping

Machine Learning Without Building a Data Warehouse First

A warehouse project can cost more than the model and delay it by a year. When you genuinely need one, and what works in the meantime.

Updated 2 min readBy SpiderHunts Technologies

Free estimateNo obligation

Get a free estimate

Tell us what you need. A senior engineer reads every enquiry.

Takes under a minute. We never share your details.

  • Free consultation
  • No commitment
  • NDA on request

Prefer to talk? Book a free 30-minute call →

Quick answer — TL;DR

A first machine learning project rarely needs a full warehouse. A repeatable extract into a single working dataset is usually enough, and it proves the value before committing to infrastructure. Build the warehouse when several projects justify it.

The prerequisite that is not one

A common sequence: a business wants a prediction, is told it needs a data warehouse first, and starts a large infrastructure project. A year later the warehouse exists and the original question is still unanswered.

Sometimes a warehouse is genuinely needed. Frequently a first project needs a well-defined extract of a few tables, which is weeks of work rather than a year.

What the model actually requires

  • The specific fields it uses, not everything in every system
  • Enough history for the patterns that matter
  • Outcomes recorded reliably so there is something to learn from
  • A repeatable way to produce the same extract for retraining
  • At prediction time, the same fields available for current records

That is a considerably smaller requirement than a warehouse. The fourth point is the one that separates a working system from a one-off analysis - if the extract cannot be reproduced on demand, you cannot retrain.

A proportionate middle ground

ApproachEffortSuits
Manual exportLowestFeasibility only - not sustainable
Scripted extract to filesLowA first project, batch predictions
Small dedicated databaseModerateSeveral related models, scheduled refresh
Full warehouseHighMany consumers beyond machine learning

The second and third rows cover most first and second projects. Moving to the fourth should be driven by having several consumers of the data, not by a single model.

When you do need the warehouse

There are real triggers, and recognising them prevents both premature building and indefinite deferral.

  1. Several models and reports need the same joined data, and each is rebuilding it separately.
  2. History must be preserved as it stood, and source systems overwrite.
  3. Extracts have started affecting the performance of operational systems.
  4. Nobody can say which version of a number is correct because several pipelines produce it differently.
  5. Compliance requires a consistent, auditable record of what the data was.

The fourth is the usual tipping point in practice. Once the business is arguing about which report is right, a single well-governed source pays for itself.

Do not paint yourself into a corner

Starting small should not mean starting badly. A scripted extract should be in version control, documented, scheduled, and structured so it could feed a warehouse later.

The failure to avoid is a one-off extract produced by hand that nobody can reproduce. That gives you a model you cannot retrain, which has a short and disappointing life.

Build the warehouse when several things need it. Not because one model might.

FAQ

Frequently asked questions

The questions readers ask us after this guide.

Still have a question?

Ask us directly — a senior engineer will get back to you.

Ask about your project

Can we use a spreadsheet extract to start?

For feasibility, yes. For anything ongoing you need it scripted and repeatable, or retraining becomes impossible.

Will we have to redo the work later?

Not if the extract is written properly. The transformation logic usually carries over into a warehouse.

What about querying production systems directly?

Acceptable for small extracts at quiet times, risky for anything heavy. Check with whoever owns the system first.

Does a data lake avoid this?

It changes the storage approach rather than the problem. You still need the data joined and cleaned to a usable state.

Keep reading

More on Data & Scraping

Start here

Want machine learning project details from us?

Tell us what you are trying to predict and roughly what data you hold. We will come back with an honest view on whether machine learning is the right tool, what the work would involve and a realistic cost range. If a spreadsheet would do the job, we will say so.

  1. You tell us what you needTwo minutes on the form, or a message on WhatsApp.
  2. A senior engineer reviews itAnd comes back with questions, a realistic range and an honest view on fit.
  3. Free 30-minute scoping callWe talk through scope, options and a realistic estimate — with no obligation.
Free estimateNo obligation

Talk to someone who builds this

Send a short brief and we will come back with an honest view and a realistic range.

Takes under a minute. We never share your details.

  • Free consultation
  • No commitment
  • NDA on request

Prefer to talk? Book a free 30-minute call →