Skip to main content
NewNMC’s advertising guidelines for doctors are in force from 6 Oct 2026. See what to change, and get a free compliance auditNew NMC advertising rules for doctors: what to change
Analytics

From Lead to Patient: A Closed-Loop Attribution Model Using CRM Stages

A stage-based attribution model for hospitals: define lead, consult booked, consult done and procedure stages in the CRM, store the source on the lead, and credit channels at the stage that matters.

From Lead to Patient: A Closed-Loop Attribution Model Using CRM Stages

This post is for hospital marketing heads and analytics owners who can already see leads by channel in their ad platforms and want to know which channels produce patients. It assumes you have, or are about to set up, a CRM where enquiries are logged. The method is a stage model: a short list of lead states, a small set of fields stored on each lead, and rules for giving a channel credit at each stage.

The reason to do this is that platform reports stop at the form fill or the call. A campaign that produces many cheap enquiries can look excellent in the ad account and still deliver few consultations. Another campaign with fewer, more expensive enquiries can be the one that fills the operating list. Only a record that follows the person from first click to consultation to procedure can tell you which is which.

Nothing here requires patient-level clinical data to reach marketing. The model works on stage names, timestamps, source fields and, later, a revenue band. Clinical content stays in the hospital information system, and any procedure advice stays with the treating team.

Why platform conversions are the wrong finish line

Ad platforms optimise toward the event you give them. If the only event is a form submission, the system learns to find people who submit forms. In healthcare that population includes people comparing prices, people with the wrong condition for the department, students, and duplicate enquiries from relatives. None of them are bad people, but they are not the same as people who attend a consultation.

Google's own guidance on offline conversion import describes the principle: capture the click identifier when the lead is created, then send a later conversion back against that click. The current offline conversion import help page calls the file-based method legacy and points to other routes, but the logic of importing a later stage is unchanged. The enhanced conversions for leads documentation describes the related route that matches hashed first-party data from the lead form to an imported offline conversion.

A closed-loop model is the discipline that makes those imports possible. Before you can send "consult done" back to an ad platform, the CRM must know that a consult was done, for which lead, and where that lead came from.

Define six stages and write them down

Stage definitions fail when they are vague. Agree them with the call centre, the front desk and the department coordinators, then publish them as a one-page definition. A workable set for a hospital:

StageDefinitionWho sets it
New leadEnquiry received by form, call, chat or walk-in logAutomatic or call centre
QualifiedRight department, reachable, genuine intent to bookCall centre
Consult bookedSlot confirmed with a doctor and dateCall centre or front desk
Consult donePatient attended; visit confirmed by front deskFront desk
Treatment advisedDoctor has recommended a procedure or courseDepartment coordinator
Procedure doneAdmission or procedure completed and billedBilling, via a stage update

Two rules keep this honest. A stage can only be set by the person or system named for it. And a stage change always records a timestamp, because the gaps between timestamps are the data you will use later for payback and for channel comparison.

Add terminal states as well: lost with a reason (price, location, went elsewhere, no response, wrong department) and duplicate. A lead without a reason code is a missing data point, not a neutral outcome.

Store these fields on every lead

Attribution quality depends on what you capture at the moment of creation. A minimal record:

  • lead_id: your own identifier, never a phone number or name.
  • created_at and channel_of_entry: form, call, WhatsApp, walk-in.
  • first_source, first_medium, first_campaign: the UTM values at first visit.
  • last_source, last_medium, last_campaign: the values from the visit that produced the enquiry.
  • click_id and click_id_type: the Google or Meta click identifier, where present.
  • department and service_line: for example orthopaedics, joint replacement.
  • stage and stage_updated_at, plus one timestamp column per stage.
  • lost_reason.
  • revenue_band: a coarse band such as A, B or C, filled in at the last stage.

Google's URL builder guidance notes that campaign parameter values are case sensitive and that missing parameters appear as (not set) in reporting. The same is true in a CRM. If one team writes "Google" and another "google", you will count two sources. Normalise to lower case on entry.

Keep personal data out of marketing URLs. Google's guidance on avoiding PII in Analytics says that neither the URL path nor its parameters should contain personally identifiable information, and that this includes campaign parameters. Use the lead_id for joining, not an email address or phone number in a link.

Rules for giving a channel credit

You have two separate jobs: understand which channels start journeys, and decide which channels deserve budget. Different rules suit each.

Origination credit goes to the first source on the lead. It answers: where did this patient first come from? It suits brand and discovery channels such as YouTube, local search and social video, which rarely close the loop on their own.

Conversion credit goes to the last non-direct source before the enquiry. It answers: what prompted the enquiry? It suits search ads, WhatsApp click-to-chat and call extensions.

Report both side by side at every stage rather than choosing one. If a channel looks strong on origination credit and weak on conversion credit, it is feeding other channels. If it is strong on conversion credit and weak on origination, you are harvesting demand someone else created, which is fine until the creating channel is cut.

For phone and walk-in enquiries, a rule is needed for the unknown. Do not leave them as "direct". Tag them with the call-tracking number's source if you use dynamic numbers, or with the visit code process described in the offline attribution method. An enquiry marked "unknown" should be visible on the dashboard as its own line so its share is not hidden.

Want this applied to your practice? Get a free analytics audit.

Book my free audit

A hypothetical worked example

The numbers below are invented to show the arithmetic. They are not benchmarks. Imagine a 200-bed hospital in Pune running two channels for its joint-replacement service over one quarter.

MeasureSearch adsInstagram video
Spend₹6,00,000₹4,00,000
New leads300400
Qualified150120
Consult booked9060
Consult done7040
Procedure done104

Cost per lead is ₹2,000 for search and ₹1,000 for Instagram, which makes Instagram look twice as efficient at the top. Cost per consult done is about ₹8,571 for search and ₹10,000 for Instagram. Cost per procedure is ₹60,000 and ₹100,000. The ranking has reversed by the time you reach the stage that carries revenue.

Two cautions apply. Four procedures is a tiny count, and a single cancellation would move the figure. And elective orthopaedic decisions often take longer than one quarter, so leads from the last weeks are unfinished. This is why the model records stage timestamps: you report cohorts by lead-creation month and read them again after the typical decision window has passed. Do not compare a mature month with an unfinished one.

Decide which stage to optimise on

Different stages suit different purposes:

  • Qualified is the earliest stage you control reliably. It is a good optimisation target for volume campaigns once you have enough events, and it is available within hours.
  • Consult booked is a strong target for most specialties. It is close to intent and arrives within days.
  • Consult done is the most useful stage to send back to ad platforms, because attendance filters out no-shows and casual enquiries. Allow for the lag.
  • Procedure done is the right stage for budget decisions and the wrong stage for bid optimisation, because it is rare and slow.

A practical pattern is to optimise bids on consult booked, review budget at consult done, and judge the channel mix at procedure done. If you want AI help prioritising which leads the call centre should ring first, that is a different problem, covered in patient lead scoring and AI lead scoring. Scoring orders the queue. Attribution assigns credit afterwards.

Failure modes to expect

Stage inflation. Staff mark a lead as consult booked to clear a follow-up reminder. Audit by sampling booked leads against the appointment system each week.

Source overwrite. A returning visitor's second session replaces the first source. Store first and last separately and never overwrite the first.

Duplicates. The same patient enquires by form and by phone. Deduplicate on a hashed phone number inside the CRM, and keep one lead with several touchpoints.

Missing revenue. Billing does not feed the CRM. Start with a revenue band set by the coordinator and improve it later through a minimal, de-identified feed, as discussed in our EHR and CRM integration service.

Consent gaps. Hashed identifiers sent to ad platforms need a lawful basis. Under India's data protection law, notice and consent standards are part of the substantive obligations that, according to the phased commencement timetable summarised by AZB & Partners, take effect from 13 May 2027. Build the consent capture now. The DPDP Act guide for healthcare marketing covers the basics.

Where to start

For a CRM walkthrough of enquiry ownership, follow-ups and pipeline stages in one place, Pioneer CRM is how we show it. The steps below apply to any CRM.

  1. Hold a one-hour meeting to agree the six stage definitions and who sets each.
  2. Add the fields listed above, with first and last source stored separately.
  3. Make source capture automatic on every form, and add a call-source field to the call log.
  4. Add lost-reason codes and make them mandatory when closing a lead.
  5. Report by lead-creation cohort, with every stage on one row per channel.
  6. Review one sample of fifty leads each month against the appointment system.
  7. Only then send consult-done events back to ad platforms.

If the tracking layer on your website is not yet capturing source values reliably, start with analytics and tracking setup, then return to the stage model.

Filed underlead to patient attributionclosed loop attribution hospitalCRM stages healthcare marketinghospital marketing ROI by channelpatient acquisition attribution model

About the author

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.

The Pioneers Brief — one essay, three benchmarks, one teardown, every Thursday. Subscribe free →

Free analytics audit

Your next chapter starts with a conversation.

Three quick questions. We come prepared with a plan for your specialty and city.

  • A 30-minute call with a healthcare specialist
  • Your top three growth opportunities, in writing
  • Honest advice — even if that means not hiring us
  • No obligation and no hard sell
You’ll speak with a senior strategist
Consultations are hosted by Arush Thapar’s team.
Step 1 of 3Takes 30 seconds

What do you want to improve first?

Details stay privateNo obligationReply within 1 working day

Client portfolio · Healthcare websites

Work relevant to from lead to patient a closed loop attribution model using crm.

Explore client work →