Dealership offline sales missing from Ads bidding
A multi-store auto dealership group spends $60,000 per month on Search and Performance Max. It tracks test-drive form submissions and phone calls longer than 90 seconds as conversions. The CRM shows 1,200 form leads per month, an 18% test-drive show rate, and 610 vehicle sales attributed to paid media over the quarter. Google Ads reports 380 conversions per month and a 4.1 ROAS using last-click; GA4 reports 290. The group recently implemented Consent Mode v2 but has not enabled enhanced conversions for leads. Sales cycles run 30 to 90 days, and salespeople often record the lead source in a free-text field. The group has no offline conversion import yet. The CFO wants to shift budget toward campaigns that produce 'real sales' and asks why Ads and CRM disagree.
1. Explain the three-way discrepancy and rank the fixes that would make the numbers comparable.
Read the worked answer
The three systems are counting different things. Google Ads reports 380 conversions per month, but those are mostly form fills and qualified calls, not vehicle sales. The CRM's 610 sales over the quarter are a later-stage outcome. GA4's 290 may differ because of consent, cross-device behavior, attribution windows, tag coverage, and how forms and calls are recorded. The free-text lead source also makes CRM attribution inconsistent, and duplicate leads can inflate one system without inflating another.
Rank fixes by impact:
- Define a shared conversion ladder: lead, qualified lead, test-drive shown, sale. Label each system's numbers by stage so nobody compares leads with sales.
- Enable enhanced conversions for leads and capture GCLID plus hashed email at form submission. This improves matching while respecting consent.
- Import offline conversions from CRM for shown test drives and sales, with deduplication and a fixed lag window.
- Clean lead-source capture: replace free text with controlled picklists and store the original click ID.
- Reconcile Consent Mode v2: confirm tags fire correctly and distinguish observed from modeled conversions.
Do not 'fix' the gap by changing the Ads conversion action to match CRM sales before offline import exists. That would leave bidding blind for another quarter.
2. Design an offline conversion import and bidding approach. Which stages and values would you use?
Read the worked answer
Import a staged funnel rather than only final sales. Use these stages: qualified lead, test drive booked, test drive shown, and vehicle sale. The first two can be imported relatively quickly; the later two need a 30–90-day lag and should be backfilled daily or weekly. Match with GCLID where available and enhanced conversions for leads with hashed email. Deduplicate by CRM opportunity ID and exclude canceled deals.
Assign values deliberately. A shown test drive might carry a modest value, while a sale should carry expected gross margin, not total vehicle revenue, because margin varies by model and financing. If margin data is reliable, use Maximize conversion value with tROAS; if not, use tCPA on shown test drives or sales after enough volume. Do not switch to sale-only bidding before the account has sufficient observed conversions; early on, use the staged funnel to keep learning.
Adjust conversion windows to the sales cycle rather than the default. Report cost per shown test drive and cost per sale separately by store and campaign. Keep lead-form conversions as diagnostics, not the primary bid signal. This gives the CFO a defensible path from ad click to gross-margin sale.
3. How should the group respond to the CFO's request to shift budget toward campaigns that produce 'real sales'?
Read the worked answer
Agree with the goal but reject last-click budget shifts until the measurement is fixed. Last-click ROAS of 4.1 is not comparable to CRM sales because it credits early clicks, ignores assisted pathways, and may overcount leads that never show. A campaign can look strong on form fills and weak on sales, or vice versa, especially with a 30–90-day cycle.
Build a store-level or geo holdout: increase budget in matched markets, hold others constant, and compare incremental shown test drives and gross-margin sales. Use the CRM to measure sales, not Ads conversions. Include sales capacity as a constraint; sending more leads to a store that cannot answer calls will not create incremental sales.
Once offline values are flowing, shift budget gradually toward campaigns with better cost per incremental sale, not merely more leads. Keep brand and lower-funnel search protected during the test. Report results to the CFO as incremental gross profit per dollar, with a confidence range, and state the consent-related limitation. This answers the request without optimizing to a misleading proxy.