Run an analysis

Use one respondent per row and one product per column. Include column names. Use 0 and 1, yes and no, or true and false for binary reach. Use numeric ratings with a minimum rating threshold for rating data.

The file can have a first column of respondent IDs. The parser detects common ID headers. Use the ID checkbox to change that choice. Empty cells, duplicate product names, and mixed invalid values cause an error. The current valid dataset stays available after an invalid replacement.

Files must be CSV, TSV, or TXT. The limit is 50 MiB and two million cells. Download the CSV template for an example.

Set Portfolio sizes with whole numbers from 1 to the number of products, up to 10. Fixed products must fit in the smallest size. Check a product under Fixed products to put it in every portfolio. The list shows products by name, A to Z. Select Product or Reach to sort by that column, and select it again to reverse the order. Zero reach products cannot be fixed. Stage 1 is always reach. Choose frequency or penetration for stages 2 and 3, or None to use reach alone.

The hint under Portfolio sizes gives the smallest set that reaches every respondent who can be reached. The page finds it when the data, threshold, or fixed products change. Respondents with no product reach cannot be covered.

Select Run to solve the sizes. Several sizes solve at the same time, and each completed size appears in the table immediately. The headline and charts add sizes in size order, so a size shows there when every smaller size is done. Cancel stops the run. Completed sizes remain available. Changing the dataset clears old results. Changing the threshold, stages, or fixed products keeps the current run and its original input snapshot. A notice identifies results from previous settings. Run again to apply changes.

Reach counts respondents with at least one selected product. Frequency counts all selected product responses. Penetration is the harmonic mean of the selected products' individual reach counts. For reach counts 2 and 4, penetration is 2 / (1/2 + 1/4), or about 2.67. It is not a percentage.

The page shows the largest completed size. Select a point on the reach chart or a row in Best portfolio for each size to show another size. Columns shows what each product adds, in greedy order with fixed products first, and keeps product colors across sizes. Download CSV saves names, fixed products, metrics including penetration, and solve time.

Start deep dive compares greedy and most popular portfolios and finds a reach and frequency frontier. The baseline budget is 10 seconds. The frontier budget is 30 seconds and 40 points. Input preparation and solver loading are outside these budgets.

A time or point limit produces an incomplete frontier. The page shows the stop reason. A penetration limit means the best tie may not be proven. A new run discards an unfinished deep dive and replaces the results. Cancel during a run keeps the completed sizes.