Creative Delivery Concentration Calculator

Prepare an additive impression export

Use one settled window and one reporting grain. Each row is a unique anonymous creative ID and whole impression count, separated by a comma. Do not paste a header, thousand separators, confidential customer details or overlapping asset-level counts. Include eligible zero-delivery rows. IDs are case-sensitive; merge aliases intentionally before pasting.

The browser tool does not access your ad account or run AI. It accepts at most 100 rows, 80 characters per ID and one billion impressions per row. These are input limits, not recommended campaign sizes. Editing rows or scope clears an old result; Reset restores the synthetic example.

Definitions and formulas

For count n_i, total N=sum(n_i). When N is positive, impression share p_i=n_i/N. Squared-share concentration C=sum(p_i²); effective creative count E=1/C. The top share is max(p_i). With k positive-count creatives, 1≤E≤k. Equal delivery across k creatives gives E=k; one delivered creative gives E=1. Zero-count rows do not alter E, but are retained in roster and zero-delivery counts.

This is the plug-in inverse Simpson calculation, without finite-sample correction or inference. We apply its standard arithmetic to creative categories as an original descriptive use. Values are calculated at full floating-point precision and rounded only for display. With N=0, shares, C and E are undefined; the JSON export uses null, not a misleading zero.

Check the synthetic answers

Counts 7,000, 1,000, 1,000, 1,000 and 0 yield N=10,000, five listed, four delivered, one zero-delivery, largest share 70%, C=0.52 and E=1.923076923. Four equal rows of 2,500 yield C=0.25 and E=4. Multiplying every count by ten leaves the shares and E unchanged; it increases volume, not balance.

One row with 10,000 and four zeros gives E=1. Five zeros give no defined exposure distribution. Adding more zero-delivery ads cannot improve the effective count. Splitting one ad into artificial IDs can improve it numerically, which is why the ID/version contract matters.

Use the result responsibly

Export the JSON with its anonymous scope and attach the source roster separately. A low effective count prompts an eligibility and allocation review, not automatic pausing. A high count is not evidence of creative diversity, unique reach, sufficient samples, conversion lift or independent random assignment. The tool intentionally does not calculate CTR, ROAS or significance from impression shares.

For an experiment question, specify the assignment unit and outcome, estimate sample needs and use the platform's appropriate controlled-testing workflow. Equalizing delivery inside an optimized campaign can change its objective and cost; this tool does not authorize that action.

Sources checked October 8, 2026: Google creative-testing guidance, October 1; Google ad rotation; scikit-bio inverse Simpson definition. The review workflow and marketing interpretation are original synthesis, not vendor endorsement or a statistical test.