Skip to content

Strategy & Systems Design

Data architecture: every question answered from the right place

“What did this customer pay?” and “what did we make by area last month?” look like the same kind of question. They aren't. One needs a single live record; the other adds up thousands, and run against the live system it slows the office down. Data architecture decides where each kind of question is answered, which copies exist, what stays behind and who may see what.
Talk through your dataSee where each kind of data lives ↓
where it's writtenwhere it's shaped for questionswho asksphone and email: never copied acrossoffice app databasecustomers, jobs, invoicesfile storagephotos, documentsaccountingthe books, paymentsreporting copyrefreshed hourlyjob history searchnotes and photos, by who asksnot the live database: it would slow the officeownerjobs and revenue by area,last weekoffice leadwhat did J.H. paylast year?only jobs T3 may seeT3has this furnace hadthis fault before?

Where each kind of data lives

One home for each, the copies listed, and who may read it. This part of the design is the one people check most.
Customers and contact detailsoffice app; copies in accounting and texts; never in the reporting copythe office; technicians for their own jobs
Jobs, notes and photosoffice app and file storage; searchable in job historythe office; technicians for jobs they may see; customers for their own
Invoices and paymentsoffice app; copies in accountingthe office and finance; customers for their own
The booksaccounting onlyfinance
Weekly figuresthe reporting copy, refreshed hourlythe owner and the operations manager

Six decisions, made once

Made once and written down, so every report, connection and AI answer follows them without anyone deciding again.

  1. 01

    Each record's home

    One system writes it

    The rest read copies, and the copies say where they came from and how fresh they are.

  2. 02

    What each word means

    “A job”, “revenue”, “an active customer”

    Defined once, so every report counts the same thing.

  3. 03

    Which copies exist

    Each with a reason

    A reporting copy for totals, a search for history. Nothing copied just in case.

  4. 04

    What stays behind

    Only what each copy needs

    Nobody's weekly totals need phone numbers, so they don't cross over.

  5. 05

    Who may read what

    Checked before anything is shown

    Including in search results and AI answers, not just on screens.

  6. 06

    How long it's kept

    The business's rule, applied

    Each kind of data has a retention rule you chose, and something that actually applies it.

Reports from exports, or from a model

Reports from exports

  1. 01Someone exports to a spreadsheet every Monday
  2. 02Each report counts “a job” its own way
  3. 03Big questions run against the live system
  4. 04Personal details copied into every spreadsheet

The numbers disagree, and nobody knows which spreadsheet is right.

Reports from a model

  1. 01A reporting copy that refreshes by itself
  2. 02One definition for each word
  3. 03The live system left to the office
  4. 04Only the fields reports need

Everyone reads the same figures, and the office app stays fast.

Where data starts, and where it's read

Written in

Office app database
customers, jobs, invoices
File storage
photos, documents
Accounting
the books
Data architecture
homes, definitions, copies, access, retention

Read by

Reports
from the reporting copy
Search and AI answers
permitted history only
Integrations
copies to other systems

Worth doing when data lives in several places

Good fit

  • Reports slow the office down, or disagree with it.
  • You want search or AI over your own records and need to control who sees what.
  • Data sits in several systems and nobody's sure which to trust.

Not the right fit

  • One system whose built-in reports already answer your questions.
  • A one-off analysis. An export is fine.

Where this capability is used

Next to data architecture

Questions about data architecture

Do we need a data warehouse?

Not necessarily. For a small business, a reporting copy beside the live database is often enough. A separate warehouse earns its place when there are many sources or heavy analysis.

Why not report straight from the live system?

Looking up one customer is fine. Adding up thousands of jobs competes with the office for the same database, and slows it down when it's busiest.

How do you handle personal data?

Each field gets a home and a list of who may read it, and copies carry only what they need. We build to the rules you tell us apply to your business.

Who decides how long data is kept?

You do. We make sure each rule is written down and that something actually applies it.