Most patient enquiries in this specialty arrive by phone, so measuring calls is essential. Chapter three explained why naive call tracking damages your entity record. This lesson is the implementation that gets you both.
The conflict, restated
Attribution wants a unique number per channel. Consistency wants one number everywhere. Resolve it with a single rule:
Your canonical number goes in every listing, profile, and structured data. Tracking numbers live only where they cannot pollute your entity record.
Three safe implementations
- Dynamic number insertion on your website only. JavaScript swaps the displayed number by traffic source, while the number in your markup and structured data stays canonical. You learn which channel produced the call and your listings never change. This is the best general answer.
- Primary plus additional numbers. Where a platform supports it, including Google, set your canonical number as primary and a tracking number as additional. Matching uses the primary.
- Dedicated numbers for closed channels. Paid ads, print, direct mail, and sponsorships can carry their own numbers safely, because they are not part of your organic record.
What not to do
- Do not put a tracking number as the primary number on any profile or directory.
- Do not put a tracking number in your structured data.
- Do not let a tracking number reach an aggregator, since it will propagate widely and outlive the campaign.
- Do not use tracking numbers in payer directories or credentialing data. These are slow to correct and the patients are your most valuable.
- Do not switch off an old tracking number without checking where it still appears. Chapter three explains why: the calls simply stop, silently.
Healthcare-specific cautions
Call tracking in a psychiatric practice touches sensitive territory, so decide these deliberately rather than accepting vendor defaults:
- Call recording. Recording conversations with patients raises privacy obligations and consent requirements that vary by state, some of which require all-party consent. Take advice before recording, and if you do not need recordings, do not enable them.
- Where the data goes. A call tracking vendor handling patient calls may be handling protected information, which raises the question of whether a business associate agreement is required. Ask, and get it in writing.
- Transcription and AI analysis. Vendors increasingly offer automatic transcription and sentiment analysis. Applying that to psychiatric enquiry calls is a significant privacy decision, not a feature to switch on casually.
- Do not connect call data to individuals for marketing purposes.
The safe default for most practices: track that a call happened, its source, its duration, and whether it was answered. Do not record, transcribe, or analyse content.
What to measure
- Call volume by source.
- Answered versus missed. Often the most valuable number in the whole system, and the one practices are most surprised by.
- Time to answer, and abandonment.
- Calls outside opening hours, which may justify changing your hours or your voicemail.
- Callback time for missed calls.
- First-time enquiries versus existing patients, which requires staff to note it.
Missed calls deserve emphasis. Practices routinely discover they are missing a substantial share of inbound calls, which means the highest-return action is not more visibility but answering the phone. Chapter eight ghost network lesson makes the same point from the patient side.
Do not over-attribute
Even with clean tracking, be careful about the conclusions:
- A patient may see your listing, ask their therapist about you, search your name, and then call. The tracked source is the last step, not the cause.
- Referred patients often search you before calling, which can make search look responsible for referral volume.
- Payer directories and print pass no source data, so they under-report structurally.
This is why the intake question from chapter eight matters more than the tracking system. Ask every new patient how they found you and record it, then use call tracking to understand call handling rather than to settle attribution arguments.
Attribution for long consideration cycles is next.