maple.website — AEO websites & hosting
Telephony

Excel Care Pharmacy and Gateforth Medical Clinic: one address, two businesses, one phone system

A pharmacy and a medical clinic operate out of the same unit in Scarborough, Ontario. They share a building and four copper phone lines, and they must never sound like the same business. This is what we built, and what is still open.

The sixty-second answer

A pharmacy and a medical clinic share one unit in Scarborough, Ontario, and four copper phone lines between them. We built a hosted phone system that keeps the two businesses sounding like two businesses: separate menus, separate caller identities, separate voicemail, ten handsets, and emergency calls tagged by site. Two fax lines were deliberately left out of the system because they carry prescriptions. One defect is open and unfixed: voicemail recordings arrive about 25 dB too quiet, and the fix is blocked on physical access to the on-site gateway.

The situation

Ram owns two businesses at one address: Excel Care Pharmacy and Gateforth Medical Clinic, in the same unit at 15-5800 Sheppard Avenue East in Scarborough. They are not one brand. A patient calling the clinic should hear the clinic, and a customer calling the pharmacy should hear the pharmacy, even though the same people may be a few metres apart.

This was a brand-new build on brand-new lines rather than a rescue. The businesses were not yet open to the public when we started, which mattered: it meant we could take the system down, probe it and rebuild it without anyone missing a call. We signed on 28 August 2026 and this became our first signed hosted-PBX client.

Two constraints shaped everything else. First, the owner is not technical, and does not want to be. A large part of what he was buying was not having to phone a carrier, read a bill back to someone, or interpret a config screen. Second, he explicitly refused an AI receptionist. He raised it himself and ruled it out: he does not want patients reaching an AI voice. We built what he asked for.

Four copper lines is the whole design

The building has four analog lines from the incumbent carrier. Two carry voice, two carry fax. That number is not a budget decision we could design around - it is a physical fact about the copper coming into the unit, and it dictates the system.

The two voice lines terminate on an analog telephone adapter on site, which presents them to our PBX as SIP trunks. The two fax lines run directly into the analog fax machines and never touch the phone system at all. That was deliberate. The faxes carry prescriptions, and putting a prescription fax through a packet network creates a class of failure that did not exist before, in exchange for a convenience nobody requested. The fax numbers are announced by the phone menu; nothing is routed to them. Two of the adapter's four ports sit spare as a result, which is headroom for later rather than waste.

An analog line presents no dialled-number information when it arrives. The port it lands on is the only signal of which business was called. That made the trunk-to-line map load-bearing, and it is exactly where the build went wrong first: the two voice lines were physically crossed at the gateway. We proved the crossing with a port probe rather than assuming it, corrected it on our side by swapping the routing context, caller identity and emergency-call trunk selection for each account, and then went back and realigned every record that still carried the pre-swap labels. The building signage, which shows the clinic's number, gave us a second independent confirmation of which line was which.

One consequence worth stating plainly, because it is the kind of thing a phone vendor usually lets a client discover on their own: each copper line carries exactly one call. A second caller to the same number hears a busy signal generated at the carrier's central office, before the call reaches the building. No amount of configuration on our side changes that. Cross-line contention - clinic busy, pharmacy call arriving - does reach us and is ours to handle.

What we built on top

Ten SIP handsets are provisioned across the two sites, with the extension ranges kept separate so a handset's number tells you which business it belongs to. Each business has its own auto-attendant menu, its own recorded greetings for open and closed hours, and its own caller identity on outbound calls, so the pharmacy never announces itself as the clinic.

Voicemail is arranged around how a front desk actually works. Every outside caller's message lands in a shared reception mailbox for that site, not in an individual's box, because a message from a patient needs to be heard by whoever is at the desk rather than by whoever happened to be dialled. Personal boxes exist for unanswered internal calls. Each handset subscribes to both its own box and its site's shared box, so the message-waiting lamp actually lights when there is something to hear, and a single star code opens the right mailbox from any phone on that site without the user knowing a box number.

Hold and transfer work the way people expect, which took a specific decision rather than a default: a per-handset music-on-hold class, because without one the held caller gets silence. Attended transfer is enabled everywhere and media is kept on the server so transfers do not break.

Emergency calls route with per-site tagging. The address a 911 operator sees for an analog line comes from the carrier's own record, not from anything in our dialplan - so the practical work was confirming that both copper lines carry the correct civic address with the carrier. Because both businesses share one address, that was one confirmation instead of two. Recorded greetings on both lines carry an explicit instruction to hang up and dial 911 in a life-threatening emergency. That line is mandatory and is not a wording preference; it stays in every future revision of those greetings.

We also left the carrier's own voicemail in place rather than switching it off. If the gateway, the server or the site's power fails, the carrier still answers and takes a message. That decision makes the number of rings before our system picks up load-bearing, and it is documented as such so nobody "tidies" it later.

What is still open

We would rather publish this than omit it.

Voicemail is about 25 dB too quiet, and it is not fixed. Recordings arrive at roughly -39 to -42 dBFS RMS where healthy is -20 to -26 dBFS. We confirmed it on 9 September 2026 with two unrelated callers minutes apart, which rules out any single handset and puts the deficit on the shared inbound path. We ruled out network packet loss with a timing test on the silent runs in the recording, and ruled out the storage format by measuring two different encodings of the same message within a decibel of each other. The cause is the analog receive gain on the on-site gateway. The fix requires reaching that gateway's management interface, which sits on the client's own network with no route from our infrastructure - so it is blocked on physical or remote access to the client LAN, not on a technical unknown.

Also still open, and honestly ordinary for a build at this stage: voicemail PINs are still at their default and must be changed before real patient messages accumulate; temporary recorded greetings are in place while the approved professional versions are recorded; fax-to-PDF was requested and deferred in writing to a later date; and the client's own data-handling agreements are drafted but not yet signed, with the gate set at the point his doors open.

What this engagement is not

We are the telephony provider here, not the client's IT provider. The contract is scoped to the phone system, and his own handling of health information sits outside it. Call recording is off by default and would have to be scoped in writing before it was ever switched on. Voicemail-to-email is off on every mailbox for the same reason: a recorded patient message is health information, and it stays on the system rather than landing in an inbox.

No claim is made here about call volumes, revenue, or an outcome measured after the fact. The businesses were not open when the system was built. What we can state is what was delivered, what was verified, and what is not finished.

Why this is on our website

Most agency case studies are a number and a grateful quote. This one has neither, on purpose. If you are considering handing your phones to a small Canadian shop, the useful information is how we behave when something does not work - whether we measure it, say so, and keep it on the list. So the open defect is in this page, not in a footnote, and the client can read it.