Understanding dataset weighting in a lens

A lens combines several datasets into one view. By default every response carries the same weight, so a dataset of 50,000 responses shapes the result far more than one of 500. Dataset weighting lets you decide how much influence each dataset has, no matter how big or small it is.

Turning it on

Open your lens's configuration (Profile > Manage lenses > Configure Lens) and go to the Datasets section. There you'll find a Dataset Weighting setting with two options:

• Unweighted — no weights are applied, so bigger datasets count for more.
• Weighted — each dataset gets a ratio that you control.

When you switch to Weighted, each dataset gets a Ratio box. Enter a whole number greater than zero, then save.

Ratios do not need to add up to 100. Only their relationship to each other matters — ratios of 45, 10, 10 produce exactly the same result as 9, 2, 2.

If a dataset has several comment columns, the ratio you enter is split equally across them. The interface marks such a ratio with an asterisk (*).

To check the weights while you're looking at your analysis, hover over the weighting icon on the analysis page — it shows each dataset's share as a percentage.

How the numbers are calculated

Theme percentages

Each dataset's percentage is worked out inside that dataset first, then blended according to the ratios.

Example — two datasets, "Delivery delays" theme:

• Unweighted: 700 out of 11,000 responses → 6.4%
• Weighted, ratios 1 : 1: (5% + 20%) ÷ 2 → 12.5%
• Weighted, ratios 3 : 1: (5%×3 + 20%×1) ÷ 4 → 8.75%

A dataset with no mentions of the theme still counts toward the calculation and contributes 0%, which pulls the blended percentage down.

Scores

A theme's score is worked out separately in each dataset, then blended by ratio.

Example — the NPS of everyone who mentioned "Delivery delays":

• Unweighted: (20×500 + −10×200) ÷ 700 → 11.4, pulled toward the dataset with more mentions
• Weighted 1 : 1: (20 + −10) ÷ 2 → 5

One important difference from theme percentages: a dataset only contributes to a theme's score if it actually has responses mentioning that theme. Datasets with no mentions are left out of the score calculation entirely, rather than counting as zero.

Sentiment splits follow the same rule as theme percentages. Impact is derived from the score, so it is recalculated from the weighted scores to stay consistent with them.

What weighting changes

• Theme and sub-theme percentages, scores, impact and sentiment
• The order themes are listed in
• Over-time charts, the cards highlighting unusual movements, the breakdown of what moved your score, and dimension breakdowns — wherever they show percentages rather than counts

What weighting never changes

• Response and comment counts. Every count stays the real number of records. Some screens stop displaying counts while weighting is on (see below), but the underlying numbers are untouched.
• The comments themselves, and the summaries written from them.
• Per-dataset figures. Where a breakdown shows each data source separately, those rows show that dataset's own unweighted numbers. It is the combined total that carries the weighting.
• Individual dataset reports. Weighting is a lens-only setting.
• Revenue at risk. It is a total added up across your datasets, not an average.

Things worth knowing

Filters can change each dataset's actual share. If a filter leaves a dataset with no responses at all — for example, a dataset with no reviews in the selected month — that dataset is excluded from the calculation and the remaining datasets' shares are worked out between themselves. The percentages behind the weighting icon always reflect your saved ratios, not the filtered result.

Counts are not shown next to weighted percentages. A weighted percentage no longer corresponds to a specific number of responses, so the comment count beside it is left out — in chart tooltips, under a theme's volume figure, and in chart CSV exports. The counts themselves are unchanged; they are simply not shown alongside a weighted figure.

Turning weighting off keeps your ratios. Switching back to Unweighted does not delete the ratios you entered. Switching to Weighted again restores them.