Broken Conversion Tracking and a Report Due: What to Tell the Client

Broken conversion tracking vs conversions that never reach analytics

Two different problems look identical in a campaign report: a conversion column lower than it should be. They need different explanations.

Broken conversion tracking has edges. A site update removed the tag, a rebuilt form skips the thank-you page the tag fired on, or someone renamed the event. The gap has a start date and an end date, and everything outside it is still usable.

The second gap has no edges: a physio clinic whose patients book by phone, a roofing contractor whose quotes start on WhatsApp. Nothing is broken; those conversions never reach your analytics. The platform’s conversion count is a partial measure every month, and always was.

The fixes differ. The reporting rule doesn’t: a conversion count that doesn’t cover the whole period or every channel isn’t the period’s conversion count, and it shouldn’t sit in the report as if it were.

Tell the client before the report does

The worst place for a client to learn about a tracking gap is page three of the monthly report, under a chart that suddenly drops. Unannounced, the drop reads as a collapse and the explanation as an excuse.

Once you’ve confirmed the gap and its dates, send a short note: what happened, which dates, what still works, what you’re doing about it and how the report will handle it. If the client’s developer shipped the update that broke it, say the tag stopped firing after the update on the 9th. No blame in writing, and no hiding the cause.

If your team took two weeks to notice, own that too. A broken tag is a forgivable problem. Finding out later that the agency knew and said nothing usually isn’t.

Data notes in client reports: sample wording

The data note travels with the numbers, written for the person who skipped your email. For a tag that broke and was restored in May:

For leads that arrive by phone or chat, the note is permanent rather than dated: Calls and WhatsApp messages are not tracked by the ad platform. Lead counts below are website form submissions only. Phone inquiries come from the clinic’s call log and are not attributed to individual campaigns. It goes in every report, not just the first.

Name the affected dates and pick a baseline source

Exact dates, not “mid-month”

Date the gap from when the tag stopped and started firing: the day the site update went live and the day the fix was verified, in the ad account’s reporting timezone.

Don’t date it from the platform’s daily conversion column alone. Google Ads and Meta both file a conversion, by default, under the day of the click or impression rather than the day it happened, so a lead who clicked on the 20th and converted on the 25th can land inside the gap. The edges in that column blur by a few days. The edges in your note shouldn’t.

Partial breaks are harder. If one form out of three stopped firing, the count drops without hitting zero; say which conversion point was lost rather than rounding it into a clean outage.

One baseline source, labeled

Something outside the ad platform usually knew about those inquiries: the CRM, booking system, call log or shared inbox. Pick one as the baseline for this report and label it everywhere it appears. Blend two and you get a number nobody can reproduce next month.

Note the right-hand column: a baseline source shows that demand existed, rarely that the ads caused it. “212 inquiries recorded in the booking system, all sources” is honest. “212 conversions” quietly credits the whole count to the campaign.

What not to calculate from missing conversion data

This is where most reports go wrong. Take the May example: spend for the month was $6,200. The platform recorded 64 leads, and spend on the 16 days the tag was working came to $3,200. Divide the month’s spend by the recorded leads and you get $96.88 per lead. That figure is wrong by construction: the spend covers 31 days and the leads cover 16. For the tracked days alone, $3,200 ÷ 64 = $50.00, and that number is defensible as long as it carries its dates.

The tempting next step is to double the 64 and call it 128 leads. Don’t. The tracked days aren’t a random sample of the month, and an extrapolated count gets quoted back to you in six months as if someone had measured it.

The default for any gap, broken tag or offline leads, is not calculated, with the reason. In practice that means sorting every figure in the report into one of three groups:

1. 1Measured for the full period. Spend, impressions, clicks, CTR and CPM. The broken tag didn’t touch them.
2. 2Measured for part of the period. Platform conversions and CPL for the tracked dates, each shown with the date range it covers.
3. 3Not calculated. Whole-month CPL, conversion rate, cost per acquisition and ROAS wherever they depend on the missing days, each with a one-line reason.

If you want to say what the gap probably held, present it as the agency’s commentary: labeled as an estimate, method stated, kept out of the metrics table.

Prooflin is built around the same rule. It never estimates conversions, revenue or ROAS, and a metric you don’t supply stays marked Not provided. Before anything is generated you review and confirm each campaign metric, and that step is where the May problem gets decided. Derived figures such as CPL are built only from what you confirm, so confirming 64 leads against a full month’s spend would reproduce the $96.88 mistake. A call-log or CRM count can be typed in, and it’s marked Manually entered so nobody mistakes it for a platform figure.

Prooflin connects to no ad or analytics account and works only from the exports you supply for one reporting period, so the gap’s context comes from you: the affected dates go into the report’s editable text before an Owner or Admin finalizes it. Each finding carries a confidence level and the basis for it, and the limitations close the client-safe campaign report instead of living in a separate email.

Which conclusions survive broken conversion tracking

A gap doesn’t void the report. It lowers the confidence of whatever leans on it, so ask of each recommendation whether it depends on knowing how many conversions happened, and when.

• Still stands: an ad that drew clearly more clicks per thousand impressions than the others, or a placement with high CPM and little click response. Neither needs the conversion column.
• Stands with lower confidence: CPL comparisons between campaigns, if one ran mostly inside the broken window and another mostly outside it. Likewise delivery during the gap, if the campaign optimized for the conversion that stopped recording: the platform was steering blind.
• Hold until a clean period exists: shifting budget toward a “winning” campaign, changing a target CPA, or calling the month better or worse than the last one.

Prooflin’s sample Performance Audit for a fictional gym has no broken tag, but it’s a useful model. It treats missing business outcomes as “unavailable, not zero”, and its first finding spells out what to do when a definition changes between periods.

A broken tag is exactly that kind of break. Next month’s report should show May with its gap marked, not a cleaned-up May that makes June look like a recovery.

The offline case works the same way. If most of a clinic’s bookings start with a phone call, the platform CPL describes the minority who used the form, so don’t let it set the budget on its own. Whether those leads are any good is a separate argument, about lead quality versus cost per lead.

Report the fix as a finding, and scope it separately

The broken tag is often the most important finding in the report, so give it the same structure as any other.

The FitForge sample closes with a measurement handoff checklist for a similar reason: conversion location, event definition, attribution window and duplicates policy get confirmed with the campaign operator before the next analysis, not reconstructed after it.

Then say plainly that repairing the tracking is a separate job. Reporting on campaign data is one piece of work. Rebuilding tags, adding call tracking or wiring a chat channel into analytics is another, with its own owner, scope and often its own quote. Blur the two and the report implies the problem is handled when nobody was asked to handle it.

For phone and chat leads there’s usually nothing to repair. The question is whether the business wants them attributed at all, which involves the people answering the phone as much as any code.

Before the report goes out on the 1st

1. 1Confirm the affected dates against what changed on the site.
2. 2Email the client today with the dates, the cause if you know it and how the report will handle the gap.
3. 3Choose one baseline source, pull its counts for the same period and label it.
4. 4Sort every figure into measured, partly measured and not calculated, and delete any whole-period ratio built on a partial count.
5. 5Mark which recommendations depend on conversions. Hold those and keep the rest.
6. 6Write the tracking finding, and send the repair as its own scoped piece of work.

If a report tool drafts the report, prepare one clean period of exports and decide before upload which conversion figures you’ll confirm. If the gap lands in a month of genuinely weak numbers too, read how to explain bad campaign results without spin before you write the summary.