Skip to content

No account required

Start a project

Find out where the ad budget went.

What each channel actually cost, once the conversions every platform claims are reconciled against the orders you actually took.

No account needed to start. You only pay when you like what you see.

  • .xlsx. Excel workbook. Every sheet read, including the ones that are notes rather than data.
  • .xls. Excel 97–2003. Read the same as a modern workbook.
  • .xlsm. Excel, macro-enabled. Values and formulas are read. Macros are never executed.
  • .ods. OpenDocument spreadsheet. LibreOffice and OpenOffice workbooks, read like Excel.
  • .numbers. Apple Numbers. Tables are pulled out of the Numbers package sheet by sheet.
  • .csv. CSV. Delimiter, quoting and encoding detected rather than assumed.
  • .tsv. Tab-separated. Read as delimited text.
  • .json. JSON. Nested objects are flattened into tables, keeping the path as the column name.
  • .ndjson. Newline-delimited JSON. Streamed a record at a time, so an export of any size is fine.
  • .xml. XML. Repeated elements become rows; attributes become columns.
  • .parquet. Parquet. Read columnwise, with the file's own types kept.
  • .sqlite. SQLite database. Every table read, with the foreign keys used to work out how they relate.
  • .db. SQLite database. Read as SQLite when the file is one, and reported plainly when it is not.
  • .pbix. Power BI report. The tables inside the report are read out of its model, along with the measures that were defined on top of them, so your own definitions are used rather than re-invented.
  • .pbit. Power BI template. A template carries the model and the measures without the data, and is read for both.
  • .twbx. Tableau packaged workbook. The packaged extract is read as data, and the workbook beside it supplies the calculated fields.
  • .twb. Tableau workbook. Calculated fields and the shape of each sheet are read.
  • .hyper. Tableau extract. Read straight through, table by table.
  • .txt. Plain text. Read as notes, or as data when the contents turn out to be a table.
  • .md. Markdown. Markdown tables become tables; the rest becomes context for the analysis.
  • .zip. Zip archive. Unpacked, and each file inside is treated as though you had uploaded it yourself.

What comes back

google_ads_jan_jun.csv
4,281 rowsChecked

Three platforms claim 8,412 conversions. They describe 4,318 orders.

Nobody is lying and the numbers still do not add up: each platform counts a conversion inside its own attribution window, so a customer who saw two ads is claimed by both. 2,580 of the claims are the same order counted again, and a further 1,514 match no order at all. Reconciled, the cheapest channel becomes the most expensive: TikTok reports $39.34 a conversion and costs $128.40 an order. Set against contribution rather than revenue it loses $35,714 over the six months, because a quarter of what it sells comes back.

1.9×Claims per traced order
$92.26Blended cost per order+84.2%
1,514Claims matching no order
Open the full sample project

What each platform claimed, and what it can be traced to

8,412 claimed against 4,318 distinct orders

  1. 012,580 of the 8,412 claims are the same order counted again.
  2. 02TikTok is the cheapest channel on its own numbers and the most expensive on yours.
  3. 03TikTok loses $35,714 once returns and product cost are taken off.

01

It adds up what the platforms claim

Each platform counts a conversion by its own attribution window, so the same customer is claimed more than once and the totals exceed the orders that exist. The first number this produces is the size of that gap.

02

It reconciles claims to orders

Platform conversions are matched against your own order records, so a channel is credited with what it can be traced to rather than with what it reports. Every unmatched claim is counted and shown rather than dropped.

03

It reports cost against contribution, not revenue

Revenue per channel flatters whichever one sells the discounted, high-return products. Where product costs are present, spend is set against contribution instead, which is the number that decides whether a campaign is worth running.

It expects a mess.

Most files that arrive here have several tables on one sheet, a title block above the data, totals stranded in the middle of it, and dates in more than one format. That is the normal case, not the awkward one.

  • Claimed conversions against matched orders, by channel
  • Real cost per acquisition, with the double counting removed
  • Spend against contribution rather than revenue
  • Every unmatched claim, counted
sales_export_FINAL_v3(2).xlsx
A messy sales export, and the repairs Data Analysis App applies to it.
OrderDateCustomerAmt
1024107/02/26John Smith$84.00
102422 July 2026SARAH LOPEZ$41.00
102422 July 2026SARAH LOPEZ$41.00
1024345840s. lopez41
Not suppliedNot suppliedSUBTOTAL$166.00
102442026-07-03Amara Osei$127.50
  • Date format
  • Duplicate row
  • Serial date · text number
  • Total inside the data
Advertising exports reconciled to one order ledger, with duplicate conversion claims linked together.

Questions it works through

  • How many of these conversions are the same customer counted twice?
  • What is my real cost per order on each channel?
  • Which campaigns look profitable only because of the attribution window?
  • Is the channel that reports the best return actually selling the worst products?

Questions people ask

Do I need a tracking tool for this?

No. This runs on exports you already have: the platform's own report and your order data. It cannot tell you what a customer saw, and it does not pretend to. It tells you how far the claims are from the orders, which is the question the dashboards are quietly avoiding.

Which platforms does it take?

Any that export a spreadsheet, which is all of them. Column names differ by platform and by year, so they are matched by content rather than by an expected header.

What if my orders have no channel field?

Then attribution is matched on time and value, which is weaker, and the analysis says so rather than presenting the same confidence on worse evidence.

Upload ad and order exports

No account needed to start. You only pay when you like what you see.