Preserve the two numbers and their provenance
Write down the exact headline value and the spreadsheet sum that disagree. Attach the original export, extraction time, property identifier, filters and selected dimension. Keep a separate working copy for calculations. Without the original file, a later reviewer cannot tell whether the discrepancy came from the source or from spreadsheet editing.
State which metric you are reconciling. Summing clicks is a different operation from averaging CTR or position. Start with an additive metric and inspect the transformation steps before investigating calculated rates. Do not replace a source total with a row sum merely because the latter is easier to reproduce.
Establish what the exported rows cover
Google documents several reasons chart and table totals can differ: aggregation, table row limits and omitted anonymized queries. Its current documentation states a 1,000-row limit for the displayed table. Check the mechanism you actually used; do not assume that a UI table, an API response and another export format all have identical coverage.
Record the number of rows obtained, any filtering performed before export and whether the dataset is a selected subset. Label limitations as coverage limitations rather than missing customer activity. A remainder between a headline and visible query rows is not a list of identifiable searches you can reconstruct.
Audit the spreadsheet transformations
Compare the raw row count with the working sheet. Check hidden rows, text-formatted numbers, duplicate imports, excluded blanks and formulas whose ranges stop before the last row. Recalculate from a clean copy when the transformation history is unclear. Keep each correction in a short change log.
Google notes that unavailable values displayed as a dash or tilde may become zeros in downloaded data. Preserve a record of source labels when that distinction matters. A numeric cell is not sufficient evidence that the underlying observation was a measured zero; avoid using such a conversion to claim a loss.
Example: an arithmetic correction does not explain the entire gap
Illustrative scenario: the headline shows 500 clicks and a spreadsheet sum shows 430. A formula omitted the last imported rows; correcting its range changes the sum to 460. The team can explain 30 clicks of the mismatch from its own calculation error, but the remaining 40 still require a coverage review.
The reviewer documents the correction and checks the source's grouping and omissions. The team does not label the remaining 40 as nonbranded clicks, lost leads or a hidden campaign without evidence. These fictional values demonstrate partial reconciliation; they are not a measured account result or a universal ratio between tables and charts.
Publish a reconciliation note, not a forced equality
Use four fields in the final note: source headline, comparable exported sum, verified adjustments and remaining limitation. Explain which number is appropriate for the report's stated question. If the coverage does not support equality, say so and retain both values with their labels rather than inserting a balancing row.
Keep this note with the export procedure so the next reporting cycle uses the same rules. If a previously stable difference changes unexpectedly, investigate that new evidence. A documented limitation can be an acceptable analytical outcome; pretending to have recovered unavailable query details cannot.