Commonwealth AI.
What We Deploy DEP.02

Relational Data Architecture

A single source of truth instead of fragmented spreadsheets.

We migrate the chaos — your scattered Sheets, half-installed CRMs, and disconnected vendor portals — into one relational database the rest of the business can read, write, and trust.

Book a Schema Review 30 min · Bring your messiest spreadsheet.
Stack SupabaseAirtableNocoDBPostgreSQLMake.comn8nQuickBooksStripe
Diagnosis

You need this if nobody on the team agrees what the customer count is.

Relational data architecture means giving your business one database where every entity — customers, jobs, properties, units, invoices, contacts — lives once. One row, one ID, one place. Every other tool reads from it or writes to it. We build it on Supabase by default, on Airtable when the team needs a friendly UI yesterday, on NocoDB when self-hosting matters. We do not sell another spreadsheet template.

The clients who hire us for this are usually one big spreadsheet away from a real outage. Six tabs of “master lists.” A bookkeeper exporting a CSV every Monday. Sales with their own copy in HubSpot. Operations with another in Smartsheet. Nobody has lied — they have all just been forced to maintain their own reality because nothing else worked. We collapse the realities into one. Common signs:

  • Three different people on the team will give you three different customer counts in the same week, and all three will be defensible.
  • Updating an address means changing it in QuickBooks, the CRM, the Google Sheet, and the email signature template — and somebody always misses one.
  • You bought a CRM. Half the team logs in. The other half keeps using the spreadsheet they trust.
  • Reporting takes a person and a Sunday. The numbers come out different every time depending on which sheet they pulled from.
What you get

Six concrete deliverables. No vapor.

01 Deliverable

A relational schema written down before we build it.

A real entity-relationship diagram. Customers, properties, jobs, invoices, contacts — the nouns of your business — modeled with primary keys, foreign keys, and the rules that govern them. You sign it off in writing before any data moves.

02 Deliverable

A clean migration from your current spreadsheet sprawl.

We extract from Google Sheets, Excel, Smartsheet, QuickBooks, your CRM, your vendor portals — wherever the truth currently lives — dedupe, reconcile, and load it into the new schema. Every row gets a provenance log. You can prove where any record came from.

03 Deliverable

A Supabase, Airtable, or NocoDB instance running in production.

Hosted under your accounts, your card, your domain. We pick the platform based on who has to use it daily — engineers get Supabase, operators get Airtable, regulated environments get self-hosted NocoDB. You inherit working credentials.

04 Deliverable

Read-and-write integrations to the tools you already pay for.

QuickBooks, Stripe, HubSpot, Mailchimp, Beehiiv, ActiveCampaign — wired through Make.com or n8n so the database stays in sync with your operational tools. One update reflects everywhere, instantly.

05 Deliverable

A views layer your team can actually use.

Sales sees a sales view. Ops sees an ops view. Finance sees the books. All reading from the same tables, none of them tripping over each other. Built so a non-technical operator can filter, sort, and report without breaking anything.

06 Deliverable

A documented data dictionary and a runbook for the next person.

Every field defined, every rule explained, every workflow documented. Written for the person who'll be running this two years from now when we're long gone.

Architecture

One source. Many surfaces.

The diagram looks like a hub with spokes. The hub is the relational database. The spokes are every tool the business already uses — the CRM, the accounting system, the email platform, the dashboards. Each spoke reads from the hub and writes back. The hub is the truth. The spokes are convenient surfaces.

Schema integrity is the deliverable, not an afterthought. Types, foreign keys, and validation rules enforced at the database level — not in a spreadsheet formula a colleague can overwrite. If a record is broken, you find out at write time, not at month-end close.

DEP.02
Title
Hub-and-Spoke Schema
DWG No.
202.01.00
Scale
1 / 100
Process

Six weeks, written down. No surprises.

  1. Week 0

    Diagnose

    We map every place data currently lives. Spreadsheets, CRMs, vendor portals, that one Notion page. We come back with a single diagram of the current sprawl and the proposed schema.

  2. Week 1

    Model

    We design the schema in writing, walk you through the entity relationships, and lock the data model with your team. Nobody touches data until the model is signed off.

  3. Weeks 2–3

    Migrate

    We extract from the source systems, reconcile duplicates, load into the new database. Every record carries a provenance log. We hold a reconciliation review with finance and ops at week three.

  4. Week 4

    Wire

    We connect the database to your existing tools — QuickBooks, Stripe, the CRM, the email platform — through Make.com or n8n. Two-way sync where it makes sense, one-way where it doesn't.

  5. Week 5

    Cut over

    We move the team off the old spreadsheets onto the new views. The old sheets stay read-only for 30 days as a safety net. Daily standups during cutover week.

  6. Week 6+

    Care

    Optional retainer. We monitor the sync jobs, handle schema changes as the business grows, and onboard new tools as you adopt them. Or we hand over the runbook and disappear.

FAQ

The questions you’d ask on a call.

Q.01 Do we have to throw out QuickBooks, HubSpot, and the CRM we just paid for?
No. The database becomes the system of record. The tools you have already paid for stay as the surfaces your team uses. We build two-way sync through Make.com or n8n so a customer record updated in HubSpot reflects in the database in seconds, and an invoice paid in QuickBooks updates the same record. The tools keep working. They just stop disagreeing.
Q.02 How do we know the migration won't lose data or corrupt records?
Every record migrated carries a provenance log — what source it came from, when, and which dedupe rule applied. We run the migration to a staging environment first, hand finance and ops a reconciliation report, and only cut over once the numbers tie. The old spreadsheets stay read-only for 30 days post-cutover. If something is missing, we can rebuild it from source.
Q.03 Where does our data live?
In your accounts, on your domain. Supabase has US and EU regions and we pick whichever your contracts and customers require. Airtable runs on AWS US-East by default and supports HIPAA on the enterprise plan. NocoDB self-hosts wherever you point it — your AWS, your on-prem server, your Hetzner box. None of it routes through us. We can walk a security reviewer through the architecture in 20 minutes.
Q.04 How long until something works?
Three weeks for the first usable views, six for full cutover. By week three the sales team can log in and see customers correctly. By week six the spreadsheet sprawl is read-only. We don't believe in 18-month 'data transformation' engagements. If the work can't be done in six weeks, we tell you why and either narrow the scope or recommend somebody else.
Q.05 What does this cost?
A typical migration is $25K to $80K fixed-fee depending on how many sources and how dirty the data is. Hosting runs $25 to $300 per month for Supabase, $20 to $54 per user for Airtable team, free to $100 a month for self-hosted NocoDB. The integration scenarios in Make.com or n8n add another $20 to $200 a month. We don't take a cut of any of it.
Q.06 Self-host or cloud?
Default is Supabase or Airtable. Both are managed cloud, both have enterprise security, both are faster for us to ship. We deploy NocoDB self-hosted when regulation requires it, when the legal team has hard data residency rules, or when you have a strong preference for owning the box. Self-hosted is a 1.5x build cost and the ongoing operational burden is yours. We'll tell you which one you actually need.
Q.07 We already started building this in Airtable. Is it salvageable?
Usually yes. About 70% of what we see in client Airtable bases is structurally fine — the schema is right, the relationships are sensible, the team just hit the wall on automation, scale, or reporting. We do a one-week audit, write up what to keep and what to refactor, and quote the rest of the work. If we'd recommend rebuilding from scratch, we'll tell you and explain why. We've talked clients out of rebuilds. It happens.
Engagement

What working with us looks like.

Typical scope
$25K – $80K

Fixed-fee migration after the diagnosis week. Five to twelve source systems merged into one schema.

Typical timeline
6 weeks

Schema in week one, cutover by week six. We don’t do 18-month transformation engagements.

Optional retainer
From $3K / mo

Sync monitoring, schema changes, onboarding new tools as you adopt them.

Not included
Hosting fees · CRM seat costs · ETL of legacy ERPs over 50 GB

Big-ERP migrations are a different animal. We’ll refer you.

Next step DEP.02.99

Show us the spreadsheet you’re embarrassed about.

Bring your messiest source of truth — the one with the merged cells, the color-coded rules nobody documented, the tab named “old don’t use.” We’ll tell you what it would take to give the business one database instead of twelve.