Data and analytics
Offline sales attribution: from ad click to counter
Offline sales attribution links advertising to sales at a dealer's counter. What can be joined exactly, what cannot, and the honest methods in between.
In this post
The short answer
- Only a sale recorded against an earlier enquiry can be joined to an ad exactly.
- That join survives only if the source, the enquiry and the dealer's outcome stay on one record.
- Walk-in buyers who saw an ad cannot be joined. Estimate them with tests, not guesses.
- Asking at the counter and channel codes are rough, cheap and better than nothing.
- Report joined sales and estimated effects separately, and never add them together.
Offline sales attribution is the work of connecting advertising to sales made away from the website: at a dealer's counter, a showroom or a store. Some of it can be joined exactly, when a buyer left a phone number in an enquiry and the dealer recorded the sale against that enquiry. Much of it cannot, because most buyers who saw an advertisement simply walk in. Between the two sit honest methods: asking, codes, regional tests and careful comparison.
What can be joined, and what cannot
Start by being honest about which sales can ever be tied to an advertisement, and which never will.
Can be joined exactly
- A buyer clicks an ad, sends an enquiry, and the dealer records the sale against it
- A buyer calls a number used only in one campaign, and the call becomes a booking
- A buyer brings a code printed only in one channel's offer
- A marketplace order placed after a click on that marketplace's own ad
Cannot be joined
- A buyer sees an ad on a phone and walks into the showroom a week later
- A buyer hears about the offer from a neighbour who saw the ad
- A buyer searches the brand's name after seeing a hoarding
- A buyer researches online, then buys from a dealer the brand never hears from
The left column is smaller than most reports pretend, and that is fine. Count it well, then estimate the right column with methods built for estimating.
The chain from an ad click to the dealer's counter
An exact join is a chain. Every link must hold, and each one breaks in its own ordinary way.
- Ad clickThe link carries tags naming the channel, campaign and region.
- EnquiryThe form keeps those tags and takes a phone number.
- Lead recordOne record, with source, outlet and owner, created once.
- Dealer ownerThe outlet's named owner calls and records what happened.
- Visit or test rideRecorded against the same record, not in a separate register.
- SaleThe booking or invoice is recorded against the same record, with its value.
Where it usually breaks: the form drops the tags, so every lead reads "website". A second record is created when the buyer calls the dealer directly. The dealer records the visit in a paper register. The sale is entered in the dealer's billing system with no reference to the enquiry. Each of these is a process fix, not an analytics one. Our guides to lead dispositions and the test ride booking process cover the recording side.
The fields that keep the join alive
These are the fields that must survive from the first click to the sale. If one is lost, the join breaks at that point.
| Field | Captured where | How it is usually lost |
|---|---|---|
| Channel, campaign and region tags | On the advertisement's link | The form or the page does not pass them on |
| Enquiry reference | When the enquiry is created | The dealer creates a fresh record instead of using it |
| Phone number | On the enquiry form | Typed differently, so matching fails |
| Outlet and owner | When the lead is routed | Reassigned without a record of who held it |
| Outcome and reason | By the dealer, after each contact | Left blank, or chosen from a list too long to use |
| Sale date and value | At booking or invoice | Entered only in the billing system, with no enquiry reference |
When the platform supports it, a business can also send its recorded outcomes back to the advertising platform, so the platform's own reports reflect sales it could not see. That depends on the platform's current rules, and on the join above being right first.
Honest methods for the sales that cannot be joined
For the walk-in buyer who saw an advertisement and never enquired, no record links the two. These methods estimate the effect instead. Each has a use and a limit.
- Ask at the counter"How did you hear of us?" from a short fixed list. Cheap, rough, and biased towards what buyers remember.
- Channel codesA code or offer printed only in one channel. Counts only buyers who bring it.
- Campaign phone numbersA number used in only one campaign. Counts callers, not walk-ins.
- Regional testsAdvertise in some districts and hold back in comparable ones, then compare dealer-recorded sales.
- Before and afterCompare sales before and after a campaign. Weak alone: seasons and festivals move sales too.
- Compare like with likeSet each estimate beside the same period last year and the same districts.
Of these, the regional test is the strongest, because it compares places that differ in one thing: whether they saw the advertising.
Running a regional test without fooling yourself
- Choose comparable districtsSimilar in sales, dealers and season, chosen before anything is run.
- Fix the measureDealer-recorded sales, the same definition in every district.
- Hold back in someAdvertise in the test districts and not, or less, in the others.
- Run a full buying cycleLong enough for buyers who saw the ads to reach a dealer.
- Compare the changeThe difference between test and held-back districts, not the raw totals.
- Repeat before decidingOne test can mislead; a second, in other districts, tells you more.
Report the result separately from the joined sales. A joined sale is a count; a test result is an estimate with doubt around it. Put both on the cross-channel marketing dashboard, labelled for what they are.
How Intellsys OS handles offline sales attribution
Intellsys OS holds the pieces this method needs in one workspace. Store microsites have lead capture. A business defines its own journeys, activities, forms and dispositions, and dealers record outcomes in them. Advertising accounts connect for reporting: Google, Meta, LinkedIn, Amazon and Flipkart, with a report builder and dashboards on top. See the data part of the platform and the customers part of the platform.
Attribution carries the click to the lead and the dealer invoice. In Hybrid, Intellsys AI recommends the media mix and marketing approves; in AI Ops, agents shift budget inside limits the business sets.
Questions people ask
What is offline sales attribution?
It is connecting advertising to sales made away from the website, such as at a dealer's counter or showroom. Some sales can be joined exactly through an earlier enquiry; the rest can only be estimated with methods such as regional tests.
Can you track which ad led to a showroom sale?
Only when the buyer enquired first and the dealer recorded the sale against that same enquiry, with the advertisement's tags kept on the record. A buyer who saw an ad and walked straight in cannot be tied to it exactly.
How do you measure advertising that drives walk-ins?
With methods that estimate rather than count: asking at the counter, codes used in one channel only, campaign phone numbers, and regional tests that compare districts with and without the advertising, measured on dealer-recorded sales.
Why do our leads all show the source as website?
Usually because the enquiry form or page does not keep the tags on the advertisement's link. Pass the channel, campaign and region from the link into the enquiry record, and check a sample of new leads each week.
Should joined sales and estimated effects be added together?
No. A joined sale is a count of real records; a test result is an estimate with doubt around it. Show them side by side, each labelled, so nobody mistakes an estimate for a count.