Colquhoun.AI
← All guides

Practical website research

Fix an Analytics Import Without Changing What the Numbers Mean

· Colquhoun.shop editorial

An import error is frustrating, but making a file pass validation is not the same as making its data trustworthy. Colquhoun.shop checks the structure of an imported analytics dataset before displaying it. Use those checks to find formatting problems, while keeping the original measurements intact.

This workflow is for a dataset you already have permission to use. It does not collect private competitor analytics or verify a supplier’s traffic estimates.

Keep an untouched original

Save the source export before adapting it to the dashboard’s format. Work on a copy and keep a short note of each transformation. Converting a date format is different from replacing a missing measurement with zero. The first may preserve meaning; the second can change the story.

Use the dashboard’s sample JSON as a structural reference, not as a source of replacement figures. The current importer expects schema version 1. An arbitrary spreadsheet renamed with a JSON extension is not an equivalent file.

Check the month before the metric

Each record needs a unique month in YYYY-MM format, such as 2026-10. The importer accepts between one and 120 monthly records and sorts them by month. Two entries for October do not become a combined October total automatically.

If the source has duplicate months, investigate why. They might describe different properties, channels or reporting definitions. Do not add their totals together unless the source definitions justify doing so. For a partial month, record that limitation outside the imported figures so a later comparison does not treat it as a complete period.

The importer also cannot determine whether two sources used the same time zone or definition of a visit. Resolve those questions before assembling a single series.

Use numbers, not formatted labels

Visits, bounce rate and pages per visit must be finite, nonnegative numeric values. In JSON, a number is not enclosed in quotation marks. A value such as "1,200 visits" is a label, not the numeric value 1200.

Bounce rate uses percentage points on a 0–100 scale. If your source exports a fraction such as 0.42 to mean 42%, convert it only after confirming that definition. The importer will accept 0.42 as a number; it cannot know that you intended 42. This is an example of a file passing validation while still communicating the wrong measurement.

Visits are rounded to whole numbers during import. Do not supply daily averages as monthly visit totals simply because both fields contain numbers.

Reconcile shares without inventing traffic

The source and country lists must each contain percentage shares totaling approximately 100%. Every share must be between zero and 100, with a nonempty name. An optional AI-provider list, when supplied, follows the same total requirement.

If your export includes only the top five countries and they total 82%, find out what the remaining 18% represents. Use an “Other” category only when it accurately describes the source’s omitted remainder. Do not increase the listed countries proportionally just to silence the error; that changes their reported shares.

Small rounding differences are different from a substantial missing category. Reconcile the values against the original report and retain its definitions.

Review the result, not just the success message

After a successful import, compare several displayed monthly totals with the original. Check the domain, month labels and a percentage you know well. Keep your transformation notes with the source file rather than assuming the dashboard has certified its accuracy.

Colquhoun.shop helps organize the supplied data. Validation confirms that the file meets particular rules, not that the measurements are independently verified. For broader context, read Bring Your Own Analytics Without Losing Their Meaning and the guide to chart time scales before interpreting the resulting graphs.

This guide describes the built-in Web Reader and Keyword Density behavior verified for this publication. Available data depends on the page response; these tools do not measure visits or search rankings.