This article is for marketing and operations leads at diagnostics chains, hospitals and multi-branch clinics where most of the revenue arrives outside a web form. A person sees an ad, reads the test page, calls the centre, and books a home collection. Or they save the address, walk in two days later and pay at the counter. Your ad account sees a click. Your billing system sees a patient. Nothing connects them.
It assumes you already have a CRM or a lead log and some form of UTM tagging. The method is a matching hierarchy: a set of identifiers, ranked by reliability, that let you link an offline event to an earlier digital touch. Each step depends on consent, and each step has a failure rate you should measure rather than assume.
We do not cover the basic funnel map from click to appointment; the post on patient conversion funnels does that. Nor do we repeat the introduction to offline conversion tracking for clinics. This piece goes narrower: how to stitch identities, which keys to use, and how to estimate what you cannot match.
What offline attribution can and cannot do
Offline attribution produces a probable source for an offline event, with a confidence level. It will never be complete. Some people will call from a number they never used online. Some will walk in because a friend recommended the centre. A good system matches what can be matched, labels the rest honestly, and keeps a visible "unattributed" line.
Do not treat the matched sample as the whole population. If you can attribute only a minority of walk-ins, the matched group is probably skewed towards people who filled in forms, which may differ from people who simply turn up. Use matched data to compare channels against each other, and use the unattributed share as a warning on how much weight to put on the result.
The matching hierarchy
Rank the keys from strongest to weakest, and always record which key produced the match.
| Rank | Key | Where it comes from | Typical weakness |
|---|
| 1 | Booking or visit code | Shown to the patient after an online booking or enquiry | Patient forgets or does not bring it |
| 2 | Click ID held on the lead | Hidden form field at enquiry | Only exists for web-form leads |
| 3 | Phone number match | Caller number or form number matched to billing registration | Shared family phones, different numbers |
| 4 | Dynamic call-source | Tracking number shown on the page the caller saw | Number pool exhaustion, cached numbers |
| 5 | Front-desk question | "How did you hear about us?" | Recall bias, coded loosely |
| 6 | Modelled | Branch-level uplift during campaign windows | Cannot be attached to a person |
The first four give you person-level or call-level links. The last two are weaker evidence and should be reported separately, not blended into the same column.
Step 1: Give every digital enquiry a visit code
A visit code is a short, human-readable reference issued when someone books or enquires online. For a diagnostics centre it might be printed on the confirmation SMS and quoted at the counter. Design it to survive real conditions: six to eight characters, no ambiguous letters such as O and 0, and no personal data.
The code maps to a lead record that holds the first and last source fields, the click ID where present, the branch and the service interest. At the counter, the front desk asks for the code or searches by the registered phone number. If either matches a lead, a source is copied across to the patient visit record.
Staff behaviour decides this step. Put the code field into the registration screen rather than a separate sheet. Add a "no code" option so that blank entries are visible. Report the share of visits with a code by branch, and treat low-adoption branches as a training problem.
Step 2: Stitch on phone number, carefully
Phone number is the workhorse key in India because it is nearly always captured at registration. It is also the riskiest key because it is personal data and can be shared.
Three rules keep it workable.
Normalise before matching. Store numbers in one format: country code plus digits only, with spaces, dashes and leading zeros removed. A mismatch between "98xxxxxx01" and "+9198xxxxxx01" is the most common cause of a match rate that looks worse than reality.
Match on a hash inside your own systems. Where possible, compare hashed values in the CRM rather than passing raw numbers between teams. Hashing is not anonymisation, so the hashed value remains personal data and must be handled as such. It does limit who can casually read the number.
Treat shared numbers as a risk. In many households, one phone books for several relatives. If one number maps to several patient registrations in a short window, flag the match as low confidence and do not attribute revenue for each registration to the same click.
When sending data to an advertising platform, follow that platform's rules. Google's description of enhanced conversions for leads says that user-provided data is normalised and hashed with SHA256 before sending, and that the hashed information is matched to data collected on your website and to signed-in Google accounts that engaged with your ads. Matching needs the lead data to have been captured at the website in the first place, which is another reason to keep the hidden-field discipline.
Step 3: Import the outcome, not the person
Platforms need only an outcome and a way to match it to a click. They do not need the patient's name, the test ordered or the diagnosis. Send the minimum: the click identifier or hashed lead data, a conversion action name such as "sample collected" or "consult completed", the timestamp, and optionally a value.
Google's offline conversion import help page explains that a click ID column must be present for file-based imports, that GCLIDs and GBRAIDs are both accepted, and that the file method is considered legacy, with Data Manager, the API and enhanced conversions for leads preferred for new setups. Check the current options before building, because import routes have been changing.
Sensitive health detail should stay out. Do not name a test, specialty or result as the conversion action in a way that describes an individual's health condition. Use neutral action names such as "completed visit" and group by service line only at an aggregate level. Whether a particular conversion naming is acceptable under platform policy is for your compliance owner to confirm against the current policy.
Step 4: Consent and notice come first
Offline matching reuses information that a person gave for a different purpose. A phone number given for booking is not automatically available for ad measurement.
- Capture a clear notice at the point of enquiry that says contact details may be used to measure which campaigns led to visits.
- Record consent against the lead, with a timestamp and the wording version shown.
- Do not match or export records without it. If consent is missing, keep the person in your operational system but exclude them from advertising uploads.
- Store the withdrawal path and honour it in the next sync.
India's data protection framework is moving in phases. According to the timetable summarised by AZB & Partners, the Rules were notified on 13 November 2025 and the notice, consent and data fiduciary obligations commence on 13 May 2027. Design for those obligations now, since changing a consent flow later is harder than building it once. The DPDP guide for healthcare marketing explains the concepts, and your legal adviser should confirm the details for your organisation.
Step 5: Estimate what you cannot match
For unmatched walk-ins, use a conservative, labelled estimate. Two methods are common.
Match-rate extrapolation. For each branch and month, calculate the share of visits with a verified digital source among those who were asked. Apply that mix to the unmatched visits only for planning, never for billing or client reporting, and always show the matched and estimated columns separately.
Campaign-window comparison. Compare branch visit counts in the weeks before, during and after a geo-targeted campaign against comparable branches without it. This is a rough check, not proof, and it is confounded by seasonality, health camps and local events. Pre-register the comparison before the campaign starts so that you do not choose the window after seeing the result.
Whichever you use, show the unattributed line on the dashboard. A report with no unattributed line has hidden it somewhere.
Failure modes
- Stale click IDs. Platforms accept uploads only within a stated window, and click IDs for old clicks may no longer match. Upload promptly and check the current limits in the platform's documentation.
- Duplicate conversions. Importing both a form-based and an offline outcome for the same patient double counts. Pick one conversion action per outcome.
- Call centre overwrites. An agent edits the source field to "phone" during a call and erases the original source. Lock source fields as read-only.
- Branch mix-ups. Two branches share a phone line. Include the branch on every record.
- Over-reading small numbers. A branch with a handful of matched visits has no useful channel breakdown. Pool months before drawing conclusions.
A rollout plan
Staff time is the main cost, so pilot at one branch first.
- Add visit code and consent fields to the lead record and the registration screen.
- Print or send the code on every online booking confirmation.
- Train the front desk on a one-line script and a "no code" option.
- Normalise phone numbers in both systems and test matching on a sample.
- Report matched share by branch weekly and fix the lowest performers.
- Add a consented upload of one neutral outcome to a single ad account.
- Review after a full decision cycle, then roll out to other branches.
A CRM that holds the lead, the source fields, the consent status and the follow-up history in one place makes steps 1 to 5 much easier to run; Pioneer CRM is the one we use for that walkthrough. If you need help with the call side of the picture, see our call tracking and attribution service.
Filed underoffline attribution hospitalwalk-in patient attributionphone call attribution diagnosticsidentity stitching healthcare marketingoffline conversion tracking clinic

About the author
Arush Thapar
Co-Founder & CTO · Gurugram, India
Arush builds the systems the rest of the company runs on — the patient-acquisition pipelines, the AI chatbots, the analytics that tie a single click to a booked appointment. If a campaign reports a number, it's because something his team built is tracking it.
Free guide · 15 pages
The Patient Acquisition Blueprint
A 90-day plan from first search to booked appointment.