1. The document describes using MongoDB's mapReduce to perform aggregation on a denormalized data set stored in a DenormAggCollection.
2. Key steps include emitting dimension fields as keys to reduce working set size, manually aggregating measures in the reduce function, and applying filters and sorting to the final result set rather than in map/reduce.
3. The mapReduce operation calculates sum, count, average, min and max aggregates on measures, with max capping values below 100, and finalizes by dividing sum by count to calculate average.