Keep the original export unchanged
Save the downloaded CSV as a source copy before changing column names or filtering rows. If you need a working version, duplicate it. This makes it possible to trace a surprising result back to the original report.
Check the date window and time zone
Confirm the first and last dates, including whether the latest day is complete. Different account time zones can put late-night activity into different days. Compare equivalent time windows and record the exact filters that were used.
Verify the units and metric definitions
Pageviews, ad impressions, clicks, requests and earnings are not interchangeable. Page RPM divides earnings by pageviews; ad RPM divides earnings by ad impressions. Make sure every numerator and denominator describes the same property, range and currency.
Missing is not the same as zero
A blank value may mean the column was not exported, the metric was not available for that row, or the data is incomplete. Zero is a measured value. Replacing blanks with zero without understanding the export can distort ratios and averages. A responsible analyzer should indicate incomplete data instead of inventing values.
Look for duplicate and invalid rows
Check whether a file contains repeated records from two exports appended together. Do not automatically remove repeated rows based on their text alone; the same page and date can legitimately appear in different country or device rows. Compare the report’s dimensions and row keys first. Also flag impossible combinations such as clicks exceeding impressions under a consistent definition, negative pageview counts, or values stored as text.
Use totals to recalculate ratios
Do not average daily RPM percentages unless you specifically want an unweighted average of daily ratios. To calculate period page RPM, sum earnings for the whole period, divide by summed pageviews for that same period and multiply by 1,000. That method weights the calculation by the traffic denominator and matches the underlying definition.
Illustrative check
Suppose three rows show $10 from 1,000 views, $20 from 10,000 views, and $5 from 500 views. Period RPM from totals is $35 ÷ 11,500 × 1,000, or about $3.04. The unweighted average of the three row RPMs is ($10 + $2 + $10) ÷ 3 = $7.33, which tells a very different story because the rows have very different volumes. Use the statistic that matches the question.
Use Publisher Pulse carefully
The AIVUV CSV analyzer maps common fields and processes the selected file in the browser. It is a convenience layer, not an audit certificate: confirm totals against the original platform, inspect any skipped rows and do not treat the output as proof of account activity.
Open Publisher Pulse report analyzer.
Example: how a duplicate export changes the conclusion
Imagine a clean report shows $80 from 20,000 pageviews, a $4 page RPM. If the same 20,000-view and $80 rows are accidentally appended a second time, the total doubles to $160 from 40,000 views. RPM remains $4, so the ratio hides the duplicate while the total earnings and traffic are both overstated. This is why a plausible ratio does not prove the file is correct. Compare total rows, date coverage and row dimensions with the source export.
Do not deduplicate by URL alone
The same URL may legitimately appear on different dates, in different countries or on different devices. Deduplication requires a key that reflects the dimensions in the report. If you do not know that key, keep the original rows and investigate instead of silently dropping data.
Keep a small reproducible audit note
Record the source report name, export time, date range, currency, filters, mapped columns, row counts and any rows skipped by the tool. When a result changes after remapping or filtering, this note helps you reconstruct the analysis later.