To replace a CRM, I first needed to understand the two systems a small charity I work with uses day to day. Donated equipment is tracked in WordPress, in a custom plugin. People, referrals from partner organisations and the casework around them live in a large open source CRM, installed alongside WordPress years ago.
The CRM was costing them in upgrades, hosting and staff confusion. The question was whether it could be replaced with something native to WordPress, without losing anything they rely on.
Before you replace a CRM, measure what is actually used
I started with an audit rather than a plan. With read-only queries against the live databases, I counted what the CRM was really doing:
| What the CRM offers | What the charity uses |
|---|---|
| 188 database tables | About a dozen |
| Nine components | Two enabled, one used |
| 60 custom fields | Around 20 holding data at any real volume |
The system was running at roughly 4% of its surface area.
The more important finding was about the shape of the data. The usual argument against rebuilding a CRM in WordPress is that CRMs are full of many-to-many relationships and WordPress is not built for them. Here, every relationship that mattered was one-to-one or one-to-many: a referral has one client and one coordinator, an item of equipment has one donor and one recipient. The only true many-to-many link was contacts to groups, which is a taxonomy.
Decide what not to rebuild
A replacement that copies everything is just a second CRM. The plan was as explicit about what to leave out as what to build. No configurable case-type engine, no rebuilt permission matrix, no automated rules engine and no general deduplication. Each of those had a measured reason: the charity either did not use it or used one fixed version of it that could be written directly.
A transplant on a running patient
The audit also showed the system was busy. On the day I looked, records had been created within the last hour. This was not an archive to migrate at leisure, so the plan was built around parallel running:
- Existing WordPress records stay canonical, so nothing that already works changes its identity.
- New private record types for organisations, referrals, donations, notes and activities sit beside them.
- Every intake form submits to native handlers that check which form sent it on the server, and never trust a record ID sent from the browser.
- Where a link between equipment and a referral is ambiguous, staff choose. The code never guesses.
- The old CRM becomes read-only before it is removed.
Test it on a full copy first
The work is being tested for acceptance on a full local copy of the live site, refreshed from production with read-only access. Writes are only switched on locally. Each intake form family has been tested end to end with disposable entries, including retries after a failure, uploads stored with private file permissions, and checks that private files cannot be fetched directly.
It is not live yet, and it should not be until staff have worked through their real tasks on the copy and agree it fits the way they work.
What other organisations can take from it
- Measure use before you plan. Count tables, fields and records, not features in a brochure.
- Look at the shape of the data. It often makes the decision for you.
- Write down what you will not rebuild, with the reason.
- If the system is in daily use, plan for running both side by side, with a way back at each step.
- Let staff accept it, not the developer.
A large system is not a problem in itself. Paying for the 96% you do not use, and training people around it, often is.