How to Choose and Structure a Directory Database

Webflow CMS can be a reasonable starting point for a directory. Choose a separate database when your operating needs require one.

In Unicorn Factory’s early days, freelancers emailed me their details. I added each profile to Webflow manually.

A profile took five or ten minutes. As submissions grew, manual entry competed with other work.

That was an operating bottleneck. It does not mean Webflow cannot support automation or backups.

Choose tools around the work

Airtable helped me organize and automate the original directory workflow. That experience does not establish unlimited scale.

Webflow has a CMS API and CSV import/export.

Airtable has API limits. Estimate calls from your reads, writes, retries, and sync frequency.

Check each plan’s current record and storage limits. Choose around expected workloads rather than a universal tool ranking.

Design around the listing page

Open your page layout and identify fields that change between listings. Typical fields include title, description, location, images, and dates.

Start with four related records:

Use a linked-record field in Airtable where a relational database uses a foreign key. Map values deliberately between systems.

Matching names alone do not ensure matching field types. A Webflow reference expects the destination item ID.

Keep Webflow ID beside your source listing. Preserve source IDs so a retry updates the same item.

Decide where edits belong

Choose one authoritative place for each field. For example, operators might approve listing descriptions in Airtable before syncing to Webflow.

Keep Draft, Approved, and Published distinct. Record sync failures without claiming the public page changed.

Two-way syncing needs conflict and deletion rules. Test those rules before allowing edits on both sides.

Backups need a recovery test

A second live copy is not automatically a backup. A bad sync can change or delete both copies.

Keep independent exports and document how relationships and files are restored. Test recovery in an isolated environment.

Airtable snapshots have plan-dependent retention and restore behavior.

Supabase database backups do not include the actual objects stored through its Storage API.

Back up those files separately. Check your chosen plan’s retention and recovery options before depending on them.

Use AI as a schema draft

The original article suggested asking ChatGPT for a coworking directory schema. That can help identify candidate tables.

Use this illustrative prompt:

Draft a schema for Canadian coworking listings.
Include listings, locations, categories, approval status, source URLs, and checked dates.
Explain relationships and uniqueness rules.
Use clearly fictional sample data.
Do not assume CSV import creates relationships or permissions.

Review the result before importing it. Create field types and relationships explicitly, then test five fictional listings.

Check duplicate slugs, missing locations, and unapproved records. Confirm each update reaches only its intended CMS item.

Next, source useful listing data and build the listing page.

Original video

Watch the original demonstration.