Data Warehouse in New Zealand
A data warehouse is worth building the moment your reporting has more sources than a person can hold in their head. Before that, it is expensive furniture.
$60k+
Typical project value
2 min
To post a project
What do you need built?
Start here and add detail as you go.
Free - No obligation - Your details stay private
What people usually mean by data warehouse
A data warehouse is one place where data from all your systems is brought together, cleaned, given agreed meanings and kept as history. It is what makes reporting fast, consistent and possible across systems. It is also easy to over-build, so the first honest question is whether you need one yet.
A first warehouse
Two or three sources, a clean model, scheduled loading and one reporting tool on top. Usually enough for years.
A reporting platform
Most of the business's systems, agreed definitions, history that source systems overwrite, and several teams reading it.
Warehouse with modelling on top
Forecasting, machine learning or customer analytics that need a reliable foundation before any of it is worth doing.
What it costs
Indicative ranges from projects posted here. Use them to sense-check a quote, not to budget precisely.
First warehouse
Two or three sources, dimensional model, scheduled loads, documentation and a reporting connection.
$35k – $80k
Reporting platform
Six or more sources, agreed definitions, history, tests, monitoring and role-based access.
$80k – $180k
Warehouse plus advanced analytics
The above with feature engineering, model pipelines and the tooling to keep them honest.
$180k+
Questions worth asking a provider
·Do we need a warehouse yet, or would our reporting tool connect directly for now?
·What will this cost to run each month, at our data volume, in a year's time?
·Who owns the definitions of our key numbers, and where are they written down?
·How will we know a load has failed before someone reads a wrong number?
·Can our own people add a new source later, or does that always come back to you?
Common questions
How much does a data warehouse cost in New Zealand?
A first warehouse with two or three sources, a clean model and scheduled loading is $35,000 to $80,000. A reporting platform covering most of the business, with agreed definitions, history and monitoring, is $80,000 to $180,000. Adding forecasting or machine learning on top starts around $180,000. Cloud running costs are separate and typically $200 to $2,000 a month for a New Zealand mid-market business.
Do we actually need one?
Not if you have one or two sources and your reporting tool can connect to them directly. A warehouse earns its cost when you are combining four or more systems, when you need history the source systems overwrite, when reporting queries are slowing down the systems people work in, when refreshes are getting slow, or when more than one tool needs the same numbers. A consultant who proposes one for a two-source business is overbuilding; one who never mentions it for an eight-source business is underbuilding.
What will it cost to run?
Ask for a figure at your data volume in a year, not today. Cloud warehouses are priced by storage and compute, and the compute part is driven by how often data is loaded and how heavily people query it. Most New Zealand mid-market warehouses sit between $200 and $2,000 a month. The bills that shock people come from refreshing every fifteen minutes when daily would do, so ask what refresh frequency the price assumes.
Which platform should we use?
For most New Zealand businesses the answer follows what you already run. If you are a Microsoft shop, Azure Synapse or Fabric keeps the licensing and the skills in one place. Snowflake and BigQuery are both excellent and widely supported here. At smaller volumes a well-designed PostgreSQL database does the job for a fraction of the cost, and a consultant willing to say so is worth listening to. The platform matters far less than the model built on it.
What makes these projects fail?
Two things. Nobody agreed what the numbers mean, so finance and sales pull different revenue figures from the same warehouse and both go back to their spreadsheets. Or nobody owns it, so a source system changes a field, a load quietly starts producing wrong values, and the reports become misleading rather than absent — which is worse. Insist on written definitions, load failure alerts that reach a person, and a named owner.
Data Warehouse projects open right now
Titles and categories are public. Full briefs, budgets, regions and buyer detail are visible to approved providers only.
Updated daily
No open projects in this category right now. New briefs are posted every week.
Check whether you need one yet
Plenty of New Zealand businesses are reporting perfectly well without a warehouse, because their reporting tool connects straight to two systems and that is genuinely enough.
A warehouse starts to earn its cost when:
- You are combining four or more systems and the joining logic is being rebuilt in every report.
- You need history the source systems overwrite — what the pipeline looked like last March, not just now.
- Reporting queries are slowing down the systems people are trying to work in.
- Refreshes are getting slow, or hitting API limits on source systems.
- More than one tool needs the same numbers and they disagree.
If none of those is true, buy the reporting and revisit this in a year. A consultant willing to tell you that has just talked themselves out of $60,000, which tells you something worth knowing.
Definitions before schemas
Ask three people in your business what "revenue" means and you will get three answers. Invoiced or paid? Including GST? Including freight? What happens to a credit note raised in a later month?
Same again for "active customer", "margin", "on-time delivery", "open job".
These are decided by your people, not your consultant, and they have to be written down somewhere the next person can read them. A warehouse built without this produces numbers that finance and sales argue about, and the moment that happens both go back to their own spreadsheets and the investment is dead.
Do it before the build. It is the cheapest week of the project.
Ask what it costs to run, in a year
Cloud warehouses are priced on storage and compute. Storage is cheap and predictable. Compute is driven by how often you load data and how hard people query it, and that is where surprising bills come from.
Ask each consultant for a monthly figure at your expected volume in twelve months, and what refresh frequency it assumes. Most New Zealand mid-market warehouses land between $200 and $2,000 a month. Bills that shock people almost always come from refreshing every fifteen minutes when overnight would have been fine.
Decide refresh frequency by what the business does with the number. A daily sales figure read at 8am needs an overnight load, not a live one.
The platform matters less than the model
Snowflake, BigQuery, Azure Synapse and Microsoft Fabric are all good. If you are already a Microsoft business, staying in that stack keeps licensing and skills in one place. If you are not, the others are excellent and well supported here.
At smaller volumes — and most New Zealand businesses are smaller volumes than they think — a well-designed PostgreSQL database does the job for a fraction of the running cost. A consultant who suggests that rather than a platform with a monthly minimum is being straight with you.
What actually decides whether the warehouse is any good is the model on top: whether the grain of each table is clear, whether history is handled properly, and whether someone can add a source next year without unpicking it.
Load failures must reach a person
The dangerous failure is not a broken report. It is a load that fails quietly, or half-succeeds, so the dashboard still renders — with yesterday's numbers, or with a category silently missing.
Nobody notices for a fortnight, and then nobody trusts anything.
Insist on:
- An alert to a person when a load fails, not a line in a log.
- Tests on the data itself — row counts within expected ranges, no unexpected nulls in key fields, totals that reconcile to the source.
- A visible "data as at" timestamp on every report.
- A named owner in your business, not only at the agency.
Tools like dbt make the tests routine rather than a special effort. Ask whether testing is in the quote.
Start with one question worth answering
Do not start with a programme to warehouse everything. Pick a question the business argues about every month — margin by product line, revenue by channel, utilisation by team — and build only what answers it.
You will learn what your data is really like, you will get something people use inside a few weeks, and you will earn the credibility to fund the next piece. A warehouse designed around every future question is designed around guesses.
Ask each consultant what they would build first. The ones who name a single question have watched a big-bang programme fail.
Comparing five quotes
Read for:
- Who asked whether you need a warehouse at all?
- Who gave a running cost for next year, with the refresh frequency it assumes?
- Who raised agreed definitions before you did?
- Whose quote includes tests, alerting and documentation rather than only pipelines?
- Who explained how your own team adds a source later?
Describe the systems your data lives in, how much history matters, how many people need to read the result, and the number your business currently argues about. Consultants who do this work will tell you quickly whether you need a warehouse or just a better report.
Related services
Not quite what you are after? These are the briefs most often confused with data warehouse.
Get up to 5 quotes for your data warehouse project
You choose who may contact you. Free to use, no obligation.