Scaling the Google “improvement”
Reported cost per conversion fell by about a sixth. Raising the budget or tightening the target CPA on that basis buys nothing extra: demos and SQLs didn’t change.
Insight · Measurement
Recovered reporting is not new business
Fix your conversion tracking and reported conversions go up. Nothing about your market changed. Treating that jump as growth is one of the most expensive mistakes a B2B paid media team can make, because budgets and bid targets move on numbers that only look better.
Short answer
When enhanced conversions, the Conversions API or a deduplication fix goes live, platforms start counting conversions they previously missed or double-counted. Reported conversions and cost per lead change, but the number of real enquiries and qualified leads doesn’t. Re-baseline your targets, wait one conversion cycle, and judge campaigns on qualified outcomes from your own records.
Why reported numbers move when nothing else does
An ad platform can only report what reaches it. Browser tags miss conversions when consent is declined, scripts are blocked or a form changes. Server-side events and hashed contact details close part of that gap, so the platform sees more of the conversions that were already happening. Duplicate events do the opposite: the same lead counted by the browser and the server inflates reported volume until an event ID ties them together.
Both corrections are good. They give bidding a more accurate picture. But they change the denominator of every efficiency metric you track. Cost per lead, cost per acquisition and conversion rate all shift the day the change goes live, while your CRM, your sales team and your pipeline see exactly the same leads as the day before.
The risk isn’t the fix. It is the decision that follows it: raising budgets because reported cost per lead fell, cutting a campaign because deduplication made it look worse, or tightening a target CPA to a level bidding can’t actually reach.
Reference
Measurement changes that move reported numbers
Each of these is worth doing. None of them creates demand. Keep a dated log of every one.
| Change | Reported conversions | Real leads | What to do |
|---|---|---|---|
| Enhanced conversions for leads (Google) | Usually rise: more outcomes match back to clicks | Unchanged | Re-baseline Google targets; compare with CRM leads. Google Ads Help |
| Upload route change to the Data Manager API (Google, from 15 June 2026) | Can drop if old uploads fail, then recover | Unchanged | Check upload diagnostics before judging any campaign |
| Conversions API alongside the Pixel (Meta) | Usually rise: fewer missed events | Unchanged | Confirm deduplication, then re-baseline. Meta for Developers |
| Deduplication fix (shared event ID) | Drop: double counts removed | Unchanged | Expect higher reported cost per lead; don’t cut budget on it. Meta Business Help |
| Primary conversion action changed | Rise or drop depending on the new action | Unchanged | Bid strategies re-learn; hold other edits |
| Attribution window or model changed | Rise or drop, often shifting between campaigns | Unchanged | Compare periods only under the same setting |
| Form or thank-you page redesigned | Can drop to zero or double | Unchanged | Test tags and events on the day of release |
| LinkedIn Conversions API added | Usually rise | Unchanged | LinkedIn deduplicates with the Insight Tag; allow for reporting delay. LinkedIn Help |
Deterministic, matched and modelled: know which number you are reading
Not every reported conversion is the same kind of evidence. It helps to rank them before you act on a change:
- Recorded in your own system. A lead in your CRM or sheet, with a stage sales has confirmed. This is the number that pays salaries.
- Matched by the platform. An outcome tied back to a click through a click ID or hashed contact details. Accurate for attribution, but coverage varies.
- Modelled by the platform. Conversions estimated where direct observation isn’t possible. Useful for bidding, weakest for judging whether performance changed.
A measurement change usually shifts the balance between the second and third tiers. The first tier, what actually happened, is the one to check before any budget moves.
Worked numbers
Two scenarios: one number rises, one falls, pipeline stays put
| Month | Google reports | Reported cost per conversion | Demo requests in CRM | SQLs | Cost per SQL |
|---|---|---|---|---|---|
| Before | 150 | US$200 | 160 | 40 | US$750 |
| After | 180 | about US$167 | 160 | 40 | US$750 |
| Month | Meta reports | Reported cost per lead | Leads in CRM | Consultations held | Cost per consultation |
|---|---|---|---|---|---|
| Before | 132 | AED 50 | 120 | 24 | AED 275 |
| After | 120 | AED 55 | 120 | 24 | AED 275 |
Cutting the Meta “decline”
Reported cost per lead rose 10% because twelve leads were being counted twice. Consultations held are identical, so the campaign is exactly as good as it was.
Original framework
The measurement change protocol
Six steps for every tracking, conversion or attribution change, however small.
Log it
Record the date, platform, what changed and who changed it, before it goes live.
Freeze targets
Hold bid targets and budgets steady for at least one conversion cycle after the change.
Check delivery
Confirm events are accepted, matched and not duplicated in each platform’s diagnostics.
Compare with your records
Put platform-reported conversions next to CRM or sheet leads for the same dates.
Re-baseline
Reset cost-per-conversion targets to the new reporting level, not the old one.
Judge on outcomes
Decide on qualified leads and opportunities, which the change could not have moved.
Zephra records every change it makes in an audit log and checks that events sent to Google, Meta and LinkedIn are accepted and not duplicated, where configured. See conversion tracking.
India, UAE and US notes
- India. Many enquiries arrive by phone or WhatsApp and never reach a tag at all. A tracking fix changes form-based numbers only, so compare against a lead log that includes calls.
- UAE. Agencies often manage tags and accounts. Ask for a dated change log whenever reported numbers jump, before approving budget changes.
- US. Consent choices and state privacy laws affect how much server-side and hashed data can be matched, so recovery varies by audience and state. Treat any single recovery percentage with scepticism.
What to measure next
- Platform-reported conversions versus CRM leads, by platform and week.
- Cost per qualified lead across the change date.
- Share of leads carrying a click ID or hashed email.
- Budget or target changes made within a conversion cycle of a measurement change.
Related: iOS 14 attribution for lead generation and the signal gap.
Questions, answered
Recovered reporting FAQ
01Why did my conversions increase after turning on enhanced conversions?
Enhanced conversions match more outcomes back to ad clicks using hashed contact details, so Google reports conversions it previously missed. Those leads were already arriving. Check that your CRM lead count hasn’t changed, then re-baseline your cost-per-conversion targets instead of treating the rise as growth.
02Why did Meta leads drop after fixing deduplication?
Before the fix, the same lead was counted by both the Pixel and the Conversions API. A shared event ID now counts it once. Reported leads fall and cost per lead rises, but the real number of leads is unchanged.
03Should I change budgets after a tracking change?
Not straight away. Hold budgets and bid targets for at least one conversion cycle, confirm events are delivered correctly, and compare platform numbers with your own lead records. Move budget only on qualified outcomes that the change could not have affected.
04Does better tracking improve bidding even if it doesn’t create leads?
It can. More complete and accurate signals help bidding systems learn, which may improve results over later cycles. That improvement shows up as more qualified leads for the same spend, not as the immediate jump in reported conversions.
05What should a measurement change log include?
The date, platform, what changed (tag, event, conversion action, attribution setting, form), who made the change, and a note of the reported and CRM numbers for the week before and after.
Your account
Find the measurement gaps before you trust the numbers.
The free assessment checks what your conversions count across Google, Meta and LinkedIn, with no CRM required and no campaign changes.