Memory Care Provider
A national memory care and senior living provider
THREE SITES MIGRATED, SERVER RESPONSE CUT 60 PERCENT ON THE FIRST SITE TUNED
Strategy
- Site Migration
- Performance Engineering
- Conversion Tracking
- Marketing Operations
- Technical SEO
A multi-community provider, locked into a vendor it was leaving.
The client is a national memory care and senior living provider operating more than 25 communities, with an average daily census near 1,715 residents at roughly 82 percent occupancy at audit time. Its digital estate ran across three separate WordPress domains. All three sat on a proprietary parent theme frozen at its final version and never to be updated again, hosted on infrastructure the client did not own, with no upstream path to maintain or secure it after the prior agency walked away.
The tracking went dark the day the prior agency exited.
Discovery surfaced a crisis hiding under the surface. The senior living CRM exposed only a SOAP API with no retry logic, error alerts, or logging, leaving roughly 50 percent of lead attribution unreliable. The call-tracking platform was never connected to the CRM. When the prior agency exited, both conversion streams broke: zero conversions recorded across 19 paid-search campaigns since the prior agency's setup went dark, the last form-fill upload weeks stale, and Smart Bidding starving for signal. The live site carried no GA4 instrumentation at all, and three brand-sharing domains were cannibalizing each other in search.
A staged cutover under a hard external deadline.
The prior host's contract terminated with the old sites kept live for only a few more weeks, forcing a hard cutover window. We secured all three site archives before the vendor pulled the download links, then ran an after-hours, two-track DNS cutover sequenced hospice, supportive, then corporate. Apex records flipped with TTLs lowered to provider minimums, snapshots and DNS-level rollback paths defined, and per-page smoke tests on every site. The migration plan held total downtime to its under-one-hour target, with all lead forms routing into the CRM correctly through the flip.
We replaced the vendor lock-in with code the client owns.
We replaced the prior vendor's proprietary plugin layer with two custom drop-in replacements that keep the old function names as backward-compatible shims, so themes run untouched. A performance pass on the first site we tuned cut server response time roughly 60 percent, from 1.16s to 0.44s, via opcache, caching-header fixes, Brotli, and HTTP/2. That playbook is documented and rolling out to the other two. We right-sized the web servers across all three boxes, dropping hospice memory use from 80 percent to 41 percent. We hardened the brittle CRM integration with field validation, retry logic, logging, and email fallback the prior setup lacked.
The failures we found because we went looking.
Discipline surfaces what dashboards hide. We discovered nightly database backups had silently never run for the first 12 days post-migration, because the cron mechanism the server image does not support failed without warning. We rebuilt them on verified timers. We caught brute-force protection reporting zero attempts while logs showed thousands of real attacks, and a bot flood of 30,424 requests in 24 hours bypassing the CDN, locked out by firewalling the origin to CDN ranges only. We fixed a months-long CRM error filing one community's leads under another's record.
Foundation shipped. The business targets are now trackable.
With the foundation rebuilt, the client's growth mandate is finally trackable, with clean conversion data starting to accumulate after the cutover. The provider's goal is to lift census from roughly 1,798 toward 2,000-plus residents inside its existing bed footprint, holding cost per web-influenced move-in at or below 2,500 dollars, against a 2025 best-year baseline of 1,808 dollars. Our audit identified roughly 61,000 dollars a year of geo-leaked ad spend and 2.5 million monthly impressions the site was not capturing. We rebuilt the conversion tracking, cleaned the call-tracking estate, and codified a 99.5 percent uptime SLA. The outcomes are pending, and we say so.
The data
Server response time (TTFB), first site tuned
| Baseline | 1.2s |
|---|---|
| After tuning | 0.4s |
Web server memory use, before vs after right-sizing (MB)
| Corporate before | 1,228 MB |
|---|---|
| Corporate after | 819 MB |
| Supportive before | 678 MB |
| Supportive after | 391 MB |
Call-tracking number cleanup (81 numbers across the network)
| Total numbers | 81 |
|---|---|
| Stale retired | 13 |
| Survey calls filtered | 52 |
Census target (client mandate, not a delivered result)
| Baseline | 1,798 |
|---|---|
| Mandate | 2,000 |