- Nobody owns the attribution seam, the handoff between your media buyer and your finance lead.
- 31% is the gap when platforms report 140 customers and finance counts 96: 140 minus 96, divided by 140.
- Under 10% reads as acceptable, 10% to 20% as worth investigating and above 20% as serious (Usermaven, 2026).
- $714 to $1,042 is how far a 31% gap moves cost per customer on $100,000 a month of spend.
- Seven steps reconcile platform attribution to paid invoices by hand, ending in one weekly number with a named owner.
- 20 of 57 messages with broken tracking were repaired before a launch that produced 4 closes at $6,800.
What is marketing attribution reconciliation?
Marketing attribution reconciliation is the monthly match between the customers your ad platforms report and the customers finance counts. It uses one definition of a customer, one attribution window and one source of truth for money. Its output is a single number: the gap, in percent, between the two counts.
You have already held both reports side by side, the ad account saying 140 and the finance export saying 96, each one correct on its own terms, haven't you? And because each was built by a capable person answering a different question, the difference between them belongs to neither of them, which is how it survives every review meeting you sit through.
Most pages that rank for marketing attribution teach you a model to pick, first touch, last touch, linear or time decay. Reconciliation comes before any model, because a model spreads credit across customers, and the first question on your desk is which customers exist at all. Haus, the measurement vendor, describes reconciliation as a stack of tests with no single agreed number (undated page, 2026 footer); this page treats it as one customer count with a name written beside it.
Which seam is this, and who owns it?
Attribution reconciliation is seam 01, the handoff from spend to attention. It sits between your media buyer and your finance lead, and its owner today is nobody. It passes all four tests of a revenue seam:
- Two capable people sit on either side of it: the person buying attention and the person counting revenue.
- Nobody owns the handoff itself, only the function on each side of it.
- It can be written as a rate: platform customers minus finance customers, over platform customers.
- That rate sits on nobody's dashboard, so it never reaches a performance conversation.
So the quarter the gap went unreported was never anyone's failure, because the job that would have caught it was never written down in the first place, was it?
How do you calculate the attribution gap?
The attribution gap is platform-reported customers minus finance-counted customers, divided by platform-reported customers. Both counts cover the same calendar month. A customer counts once, on the date of the first paid invoice.
| Input | Example | Where it comes from |
|---|---|---|
| Platform-reported customers | 140 | Conversions claimed across every ad platform for the month |
| Finance-counted customers | 96 | First paid invoices in the same month |
| Difference | 44 | 140 minus 96 |
| Attribution gap | 31% | 44 divided by 140 |
Run it on last month's two exports and the percentage arrives before your coffee cools, and from that moment the gap stops being a feeling in the Monday meeting and becomes a figure with a size, a direction and, soon enough, a name beside it.
When your platforms say 140 and finance says 96, which 44 customers are you paying to acquire twice?
What attribution gap is acceptable?
A gap under 10% is acceptable, 10% to 20% calls for investigation and anything above 20% is serious, according to Usermaven (published 20 May 2026, updated 18 September 2026). Contributors to a Shopify Community thread (3 July 2026) call 5% to 15% normal and 20% to 30% worth investigating.
| Gap | Reading | What to do this week |
|---|---|---|
| Under 10% | Acceptable | Publish the number and keep the weekly review |
| 10% to 20% | Investigate | Check windows and duplicate events first |
| Above 20% | Serious | Match every row before the next budget increase |
| 5% to 15% | Normal, per forum contributors | Review weekly and log each cause |
The line matters less than the habit. A 12% gap opened every Monday by a named person catches a broken tag inside a week, while a 6% gap nobody has looked at since the quarter began hides one for three months, which of those two would you rather defend?
What does a platform versus finance disagreement sound like inside a company?
A platform versus finance disagreement sounds like three sentences, each true from where its speaker sits. These are the seam 01 quotes from The Thirteen Seams:
- "Platform says we got 140. Finance says 96."
- "Depends whose dashboard you're looking at."
- "We're profitable blended, I'm not worried about channel."
You have heard at least one of those in a meeting this quarter, and the person who said it was right, which is exactly what makes this seam so durable: your media buyer is right about the platform, your finance lead is right about the bank, and you sit in the middle signing next month's spend with a caveat attached that nobody writes down.
One question exposes it on your own team: what did a customer cost us last month, and can anyone answer without adding a caveat? The length of the pause that follows is the age of your seam.
What does a 31% attribution gap cost a $5 million company?
A 31% attribution gap costs a $5 million company about $677,040 a year in planned revenue that never arrives. The arithmetic below is illustrative, and every assumption is stated.
| Line | Value | How it is calculated |
|---|---|---|
| Monthly ad spend | $100,000 | Assumption |
| Customers the platforms report | 140 a month | Assumption |
| Cost per customer, platform view | $714 | $100,000 divided by 140 |
| Customers finance counts | 96 a month (1,152 a year) | Assumption |
| Cost per customer, finance view | $1,042 | $100,000 divided by 96, which is 46% higher |
| Budget increase planned | $30,000 a month | 30% more spend |
| Customers expected at $714 | 42 | $30,000 divided by $714 |
| Customers delivered at $1,042 | 29 | $30,000 divided by $1,042 |
| Shortfall | 13 a month, 156 a year | 42 minus 29 |
| First-year revenue per customer | $4,340 | $5,000,000 divided by 1,152 |
| Planned revenue that never arrives | $677,040 a year | 156 times $4,340 |
Put your own spend in the first row and your own two counts in the next, and the bottom line tells you what the gap has charged you every month it stayed unowned, so how many months has yours been running, and how many of those 13 missing customers a month has anyone explained to your board?
If you already bought an attribution tool and the gap outlived it, the tool did the job it was built for, modelling credit across whatever customers it was shown. It had no definition of a customer to enforce and no person to hand a number to on Monday, and on a Seam Map that missing person is the whole seam.
Why do platforms report more customers than finance counts?
Platforms report more customers than finance counts for five mechanical reasons, none of them dishonest. Cometly (13 May 2026) describes the same family of causes, and Usermaven's discrepancy guide (2026) documents the default windows.
| Cause | What happens |
|---|---|
| Double counting across platforms | One buyer clicks a Meta ad and a Google ad, and both platforms claim the sale. |
| Default windows | Meta defaults to a 7 day click and 1 day view window; Google Ads to a 30 day click and 1 day view window. |
| Tracking blind spots | An event fires twice, or never, on a page nobody tested. |
| Platform events versus CRM outcomes | A platform counts a purchase event; finance counts a paid invoice. |
| Divergent models | Each platform credits itself under its own attribution model. |
Which window does your Meta account use, and which one does the person closing your month assume? Once you know both answers, the gap has a cause before it has a size.
How do you reconcile platform attribution to bank deposits by hand?
You reconcile platform attribution to bank deposits in seven steps, with a spreadsheet and one named owner. DarwinApps (20 August 2026) publishes a month-end version of the same match across CRM, ad platforms and finance.
| Step | Input | Output |
|---|---|---|
| 1. Define a customer as the first paid invoice | Both teams' current definitions | One written definition both teams read |
| 2. Fix one reporting window | Each platform's default window | One window, recorded beside the number |
| 3. Export platform conversions | Each ad account for the month | Rows with order ID, email and click ID |
| 4. Export paid customers | Finance or your processor, same month | Rows with invoice, email and amount |
| 5. Match the two exports | Both lists | Matched rows on email, order ID and click ID |
| 6. Classify every unmatched row | The leftovers | Duplicate, outside the window, untracked, refunded or unexplained |
| 7. Publish one number weekly | Matched and unmatched counts | The gap in percent, with the owner's name beside it |
Matching on email and order ID works in any stack that sends a receipt, so the accounts you already pay for are enough to begin, aren't they? The first month takes an afternoon, because the unmatched rows are where every surprise lives, and by the third Monday the match runs inside an hour, which is when the gap stops being a debate and becomes a line one person defends.
Which source of truth wins?
Finance wins for money, and the platforms win for optimisation inside one channel. Finance records what was paid. A platform records what it claims, which is the right tool for comparing one ad against another ad on that same platform. Contributors to the Shopify Community thread reach the same split.
| System | Answers which question | Owner today |
|---|---|---|
| Finance or processor | Who paid, and how much | Finance lead |
| CRM | Who booked, and who was sold | Sales lead |
| Ad platforms | What was spent, and what each platform claims | Media buyer |
| The reconciled count | How many customers your marketing produced | Nobody, until you name one |
What does an AI employee do inside the attribution seam?
An AI employee inside the attribution seam runs the weekly match and hands the result to a person. It reads both exports every Monday, computes the gap, lists each unmatched row with its likely class and escalates any week above the agreed line. A named person approves the number before it is published.
That is the difference between using AI and hiring it, the subject of how to hire an AI employee: the work lands on your desk finished, with its exceptions already sorted, so the hour you once spent arguing over which dashboard to believe goes into deciding where the next $30,000 of spend will find a customer your bank recognises.
Who, by name, opens the finance export on Monday and compares it to the ad account?
Has anyone closed this seam?
Yes. Coach Kyle's live launch of 30 August 2026 ran only after 20 of 57 messages with broken tracking were repaired first.
Live launch, 30 August 2026: 412 registered, 142 live, 37 applications, 30 calls booked, 4 closes at $6,800, after 20 of 57 messages with broken tracking were repaired first.
The repair came first because a launch counted on broken messages produces numbers nobody can defend, and because Jon Weberg learned to read a business by following every dollar from the first click to the last payout at NowLifestyle, where the referral program he ran produced $270,000,000 and $336,538.13 was paid out through the affiliate dashboard he directed, the tracking on your account gets fixed before a single dollar gets counted.
You can read the full launch, stage by stage, in the Coach Kyle receipt among the case studies, and as you do, ask the question it raises on your own account: how many of your last 20 tracking messages fired exactly once, and who checked?
Bring one month of platform exports and one month of deposits, and in 45 minutes Jon Weberg matches them on your own numbers and prices your gap in percent and in dollars.
Click the button below.
Answer a few short questions about your company.
Pick a time on Jon's calendar and bring both exports.
What else do people ask about marketing attribution reconciliation?
Is attribution reconciliation the same as multi-touch attribution?
No. Multi-touch attribution splits credit among touchpoints. Reconciliation matches the customer count your platforms claim to the customer count finance recorded. You run reconciliation first, because a model built on an unmatched count distributes credit for customers finance never recorded.
Why does Meta report more purchases than my store?
Meta counts conversions inside its own window, including view-through, and a store counts orders. Duplicate pixel and server events inflate the gap too. Meta defaults to a 7 day click and 1 day view window (Usermaven, 2026), so one customer can appear in two platforms.
Should I trust Google Analytics or my bank?
Trust the bank for money received and the platform for comparing ads inside one channel. Finance records what was paid. Platforms record what they claim. The reconciled customer count uses finance as the anchor and platform data as the explanation.
How often should marketing and finance reconcile?
Weekly for the headline gap, monthly for the full variance list. A weekly review catches a broken tag within days. A monthly close finds it after a full month of spend decisions made on the wrong number.
Do I need an attribution tool to close this seam?
No. A spreadsheet closes it. The tool is a second step, after one definition, one window and one owner exist. A tool without those three produces a faster disagreement.
What is the difference between blended and channel customer cost?
Blended cost divides total spend by total customers. Channel cost divides one platform's spend by the customers matched to it. Blended hides a failing channel behind a strong one, which is why "profitable blended" is a common seam quote.