creating pivot-like table with nulls in repeated rows

Hi Anonymous ,
for performance reasons, I would recommend not to replace, but use this method instead:

let
    Source = Table.FromRows(Json.Document(Binary.Decompress(Binary.FromText("pY87CoAwEETvktrCxHzPIhZqFDX+7184q5AuKNgMD14mw5Ylq5uWZYyLAum5QDpNzFmVpW3+2/quB0ulkYdWSKtE3E3ZpztOAWysQ56WrJE8dudlpZf34uDMzfLV5h/stp/gQtJiMHSLUDb+nLLoVhc=", BinaryEncoding.Base64), Compression.Deflate)), let _t = ((type nullable text) meta [Serialized.Text = true]) in type table [Column1 = _t, Column2 = _t, Column3 = _t, Column4 = _t, IsNewData = _t]),
    #"Changed Type" = Table.TransformColumnTypes(Source,{{"Column1", type text}, {"Column2", Int64.Type}, {"Column3", type text}, {"Column4", Currency.Type}, {"IsNewData", Int64.Type}}),
    TransformFieldNames = {"Column2", "Column3", "Column4"},
    #"Added Custom" = Table.AddColumn(#"Changed Type", "Projection", each if [IsNewData] = 1 then Record.SelectFields(_, TransformFieldNames) else null),
    #"Removed Columns" = Table.RemoveColumns(#"Added Custom", TransformFieldNames),
    #"Expanded Projection" = Table.ExpandRecordColumn(#"Removed Columns", "Projection", TransformFieldNames),
    Custom2 = Value.ReplaceType( Table.ReorderColumns(#"Expanded Projection", Table.ColumnNames(#"Changed Type")), Value.Type(#"Changed Type"))
in
    Custom2

It could be that buffering step "Changed Type" and "TransformFields" would further improve performance.
You define the column names to be transformed once in the list "TransformFieldNames" and can reference them by that name throughout the logic.