Your analytics can tell you which channel produced the most form submissions. Your sales team can tell you which deals closed. Nobody can tell you which channel produced the deals, because the two systems have never been introduced.
That gap is the reason marketing in most Indian businesses is reported in form fills, and it is far easier to close than the tooling industry suggests. You do not need a customer data platform, a data warehouse or an engineer. You need a hidden form field and some discipline.
Why they cannot see each other
Analytics works in sessions and anonymous identifiers. It knows a browser arrived from a particular source, viewed four pages and submitted a form. It does not know who that was.
Your CRM works in people and companies. It knows who enquired, what they bought, how much they paid. It does not know how they arrived, unless somebody typed it in, which they did not.
The two sides never meet because nothing carries an identifier across the boundary at the moment the form is submitted. That is the whole problem, and it is one field wide.
What to capture, and when
Five values, captured on arrival and carried to the form.
- The analytics client identifier. The anonymous identifier your analytics assigns the browser. This is the join key.
- First-touch source. Where they came from the very first time they visited, stored and never overwritten.
- Last-touch source. Where they came from on the visit where they converted.
- First landing page. Which page brought them in originally, which is usually more informative than the channel name.
- Campaign parameters, where present.
Store the first-touch values in a first-party cookie on the first visit and leave them alone. Update the last-touch values on each new session. This distinction matters more than anything else here, because in a business with a long consideration cycle the first touch and the last touch are usually different channels, and crediting only one of them is how content work gets written off as ineffective.
Getting them into the form
The mechanics are not complicated.
- On page load, read the stored values from the cookie, setting them if this is a first visit.
- Add hidden fields to every form that generates a lead.
- Populate those fields from the stored values before submission.
- Map them to custom fields on the lead record in your CRM.
- Make sure they survive lead conversion, so a lead becoming a contact or a deal does not lose them. This is the step most often missed and it silently destroys the whole exercise.
Almost every form tool and CRM supports hidden fields and custom fields on the free or standard plan. The work is an afternoon, once.
What you can answer afterwards
With those five values attached to every lead, and the deal value already in your CRM, you can answer questions that were previously unanswerable.
Revenue by first-touch source and by last-touch source. Conversion rate from lead to customer by source, which frequently differs enormously and reverses the ranking you had from lead volume alone. Average deal size by source. Which landing page produces customers rather than enquiries. Time from first touch to closed deal, by source.
That last one is often the finding that changes a budget conversation, because the channel producing the fewest leads sometimes produces the fastest and largest deals.
None of this requires a dashboard. For most Indian businesses it is a CRM export and a pivot table, reviewed monthly.
The spreadsheet version, which is enough for most
Export closed deals with their source fields and their value. Pivot by first-touch source, then by last-touch source. Compare.
Do it monthly and keep the exports. Within a quarter you have something no analytics report was going to give you, built from your own records, unaffected by consent rates and tracking gaps because the revenue side comes from your own system.
That last point is worth dwelling on. This method degrades gracefully. If a visitor declined analytics consent, you lose the client identifier and still keep the source values if you captured them with consent, or you lose that lead’s attribution and keep every other lead’s. Compare that to a reporting approach that collapses entirely when measurement changes.
Phone calls and WhatsApp, which is most of India
A large share of Indian enquiries never touch a form. Somebody calls the number on the page, or taps through to WhatsApp.
Three approaches, in increasing order of effort.
Ask. One question on the call, recorded in the CRM with a fixed set of options rather than free text. Crude, cheap, and better than nothing by a wide margin.
Tag your chat links. Different prefilled messages per page, so the first message itself tells you where the person came from. This costs nothing and works, and we covered it in WhatsApp Business discovery.
Dynamic number insertion. Show a different phone number by channel, so the number dialled identifies the source. Effective and it adds a vendor, a cost and a thing that can break, so weigh it against your call volume.
Consent, handled honestly
The client identifier is only available where analytics is permitted to run, so this is tied to your consent position.
Two practical points. Where a visitor declined, you will not have the identifier and should not attempt to reconstruct one. And the source values you capture are still personal data in combination with a lead record, which means they need a stated purpose and a retention rule like everything else.
Plan for partial coverage rather than treating gaps as a failure. The surviving revenue-side metrics remain complete, which is the point we have made separately about living with partial analytics after a consent banner goes live.
Where it goes wrong
Five failures account for most of the broken implementations we see.
First-touch values overwritten. Somebody stores the source on every visit, so by the time a lead converts the first touch has been replaced by a direct visit. Write it once and protect it.
Fields lost on conversion. The CRM copies some fields from lead to opportunity and not your custom ones. Test the whole path with a real submission before trusting anything.
Forms missed. There is always one form somebody built separately that has no hidden fields. Inventory every form.
Inconsistent source values. Free text produces Google, google, Google Organic and organic-google as four separate sources. Use a fixed vocabulary.
Nobody looks at it. The most common failure of all. Build the monthly review into somebody’s calendar before you build the tracking.
The part this cannot see
Being honest about the limit. A growing share of research happens inside assistants that send no referrer and create no session. Somebody who read an answer, remembered your name and typed it into a browser two weeks later arrives as direct traffic with no first touch worth recording.
No amount of joining systems recovers that, because nothing was created to join. The practical response is the free-text question on the form asking how somebody came across you, which costs one field and regularly surfaces channels no system can see.
The wider version of this gap is in full-funnel attribution for B2B, and the measurement of that invisible layer in measuring brand awareness from AI search.
What is a waste of effort
Buying a customer data platform to solve this. It is a hidden field.
Building a warehouse before you have the fields populated. The warehouse will faithfully store nothing.
Multi-touch attribution modelling on low lead volume. With a few dozen deals a quarter, a model produces confident noise. First and last touch, compared, is honest at that scale.
Trying to attribute every single lead. Partial coverage, consistently collected, answers the budget question. Completeness does not exist and chasing it costs a quarter.
Your CRM number and your ad platform number will not match
Worth preparing for, because the first time somebody notices the discrepancy it derails a meeting.
An advertising platform counts conversions it believes it caused, using its own attribution window and its own view of which click or impression mattered. Your CRM counts deals that closed. Those are different questions asked over different periods, and they will never agree.
The platform will usually report more conversions than your CRM reports deals from that source, for reasons that are mostly legitimate: it counts form fills rather than customers, it attributes across devices you cannot see, and it includes view-through credit you have not asked for.
Decide in advance which number is authoritative for which decision. The platform’s number is the right input for optimising inside that platform, because that is the signal its bidding uses. Your CRM number is the right input for deciding how much to spend across channels, because it is the only one connected to revenue.
Write that down somewhere, because the alternative is re-litigating it every quarter with whoever is least comfortable with the smaller figure.
A note on multiple forms and multiple sites
Two practical complications that arrive once this is working.
If you run forms on more than one domain, or a booking system on a subdomain, cookies do not travel between them by default. The first-touch values captured on your marketing site will not be present when somebody completes a booking elsewhere. Handle it by passing the values across in the link, or accept that those leads carry last-touch only, and know which is which.
If you have several forms serving different purposes, keep the field names identical across all of them. Divergent field naming across three forms produces three partial datasets that nobody can combine, and it is tedious to repair afterwards.
A sensible order of work
- Create the five custom fields on your lead record in the CRM.
- Store first-touch values in a first-party cookie on first visit, and protect them.
- Add hidden fields to every lead form and populate them.
- Submit a real test lead and follow it all the way to a closed deal, confirming nothing is lost.
- Fix your source vocabulary so values are consistent.
- Tag your chat links with per-page prefilled messages.
- Add the how did you hear about us question as free text.
- Export and pivot monthly, and put that review in somebody’s calendar.
The analytics configuration this assumes is in GA4 settings that matter, and what to do with the resulting numbers in calculating SEO and GEO ROI.
If you want this set up on your own stack in an afternoon, get in touch.