Why this matters
Floodlight decisions are cheap to make and expensive to change. Group types cannot change after creation, tag strings are baked into every deployed tag, and a wrong counting method quietly inflates every cost per booking the DV360 strategies optimise towards. The Floodlight tags guide explains the objects and parameters; the ad serving basics guide explains how the ad server counts the impressions and clicks that conversions are matched against. Here you apply both to one journey before any tag exists.
The company and its booking journey
Roadnest Car Hire is an invented company renting cars from town-centre and airport branches. It is adding display and video through DV360 and a direct sponsorship with a travel website, all served through one CM360 advertiser. The developer has a data layer on every page and wants a specification before building anything.
Roadnest Car Hire: booking journey and data layer
| Step | Page or event | Data layer values available | Monthly volume |
|---|---|---|---|
| A | Search submitted (branch, dates) | Branch code, pick-up date, rental days | 41,000 |
| B | Vehicle selected | Vehicle class, daily rate | 18,500 |
| C | Extras page viewed | Extras chosen | 15,200 |
| D | Payment page reached | Basket value | 6,900 |
| E | Booking confirmed | Booking reference, booking value, rental days, vehicle class, branch code, new or returning customer | 3,100 |
| F | Account created | Customer status | 1,400 |
| G | Booking cancelled | Happens by phone or email, not on the site | 280 |
Notes from the developer:
- Customers often reload the confirmation page to print it, and some return to it from the confirmation email.
- The data layer also holds the customer's email address on the confirmation page.
- The consent platform raises a consent event before any marketing tag may fire.
- The direct sponsorship serves CM360 tags supplied by the trafficker.
Choosing the activities
Track the steps that answer a reporting question or feed an audience, and nothing else. The booking is the outcome; vehicle selection and the payment page are intent signals for remarketing and diagnosis; the search is the broadest signal of interest. Navigation-only steps add tags without adding decisions.
Each chosen step then needs three settings. The group type follows whether revenue matters. The counting method follows the business question: is a person who does it twice one signal or two? And the custom variables carry the dimensions reporting will cut by, typed correctly and free of personal data. Windows follow last, set per activity on purpose, because the view-through window decides how much display gets credit in the AttributionAssigning conversion credit to touchpoints. step.
Task
Answer before reading on.
- Choose the activities and the activity groups they belong to, with a naming convention.
- Set the counting method for each activity, and say what the
ordparameter must hold. - Define the custom variables, with type and source, and name anything that must not be sent.
- Set click-through and view-through windows for each activity, with a reason.
- Say how the tags are deployed and how cancellations are handled.
- Write the test plan to run before any line item optimises towards these activities.
Worked answer
Roadnest Car Hire: Floodlight specification
| Group (type) | Activity | Counting | ord | Windows (click / view) |
|---|---|---|---|---|
| Booking funnel (counter) | Search submitted | Unique (one per browser per day) | Fixed; random value in num | 30 days / 1 day |
| Booking funnel (counter) | Vehicle selected | Unique | Fixed; random value in num | 30 days / 1 day |
| Booking funnel (counter) | Payment page reached | Unique | Fixed; random value in num | 30 days / 1 day |
| Bookings (sales) | Booking confirmed | Transactions | Booking reference | 30 days / 7 days |
| Accounts (counter) | Account created | Standard | Random cache buster | 30 days / 1 day |
Custom variables, defined once at configuration level:
u1branch code, stringu2vehicle class, stringu3rental days, numberu4customer status (new or returning), string
On the sale: cost = booking value with no currency symbol or separators, qty = 1. The email address is never
sent in any variable.
Reconciliation target after a test month: Floodlight booking count within a small, explained gap of the 3,100 bookings in the booking system, less cancellations.
Activities. The extras page is dropped; it informs no decision. Funnel steps count uniquely because a person comparing dates repeatedly is one signal, not several. Account creation is a counter, not a zero-value sale.
The booking. A sales activity with transactions counting and the booking reference in ord is the only safe
design given the reloads: CM360 deduplicates on the reference, so a printed and revisited confirmation remains one
booking. Rental days go in a number-typed custom variable so they can be summed; revenue stays in cost.
Windows. The sale gets a longer view-through window than the funnel steps because a booking can follow an impression by days, but it is a recorded choice, not the default.
Deployment. Floodlight Counter and Floodlight Sales tags in Tag Manager (GTMTag management system.), one per activity on its own trigger, values read from data layer variables, all gated behind the consent event, with a Conversion Linker tag on every page. The trafficker checks that the sponsorship tags were pasted with the cache buster and click macro filled, or its impressions and clicks are undercounted and conversions under-attributed.
Cancellations. They happen off-site, so no tag can see them. Report net bookings by subtracting cancellations from the booking system in reconciliation, and consider the documented offline conversions upload only once the online setup is proven.
Tests. Tag Manager preview and Tag Assistant on every step with consent granted and then refused; a reload and an email return to the confirmation page to confirm one booking; a check that rental days arrive as a number; and a Floodlight report over the test period compared with the booking system, the discipline the media auditing and verification guide applies to delivery and invoices.
Common mistakes
- One activity called "conversion". Searches and bookings become indistinguishable in every report and bid.
- A random
ordon the booking. Every reload becomes another sale. - Account creation as a zero-value sale. It pollutes revenue columns.
- Sending the email address. Prohibited, and a liability the moment it is noticed.
- Accepting default windows. The view-through window is a decision about display's credit.
- Testing only the happy path. Consent refusal and reloads are where tags misbehave.
Next, test yourself with the ad serving and verification checkpoint or return to the intermediate programmatic level.
