Do you have any experience of quantifying the "none of these" for studies with large numbers of SKUs (e.g. soft drinks) where the SKU sizes/types differ and chip allocation is used within a shelf-display conjoint exercise? Respondents are asked how many of each SKU they'd buy on a typical shopping occasion (with no limits on this).
We've used the method of taking the maximum across all tasks and recoding the "none" data according to this (None= maximum across all tasks - number chosen in given task) before running HB on the data. However this doesn't really work when we might have e.g.
SKU 1 - 1 litre bottle of fizzy drink
SKU 2 - Multi-pack of 4 x 1 litre bottles of fizzy drink
These aren't equivalent in terms of unit numbers because you're more likely to buy 4 of SKU 1 in place of one of SKU 2. So when it comes to interpreting and defining the "none of these" volume this can be a challenge. Do you have any suggestions of how we could deal with this situation and make a more accurate estimate of the % none in volume terms in these situations?
We thought about perhaps converting everything to litres but this isn't been ideal either as we might also have other SKUs included such as:
SKU 3 - 4 x 330ml cans
SKU 4 - 4 x 250ml cans
where these could be considered more equivalent in purchasing terms (but we'd be recoding them to be different in terms of litres). We'd also end up with very high numbers in the data which can lead to long HB estimation times.
It would be great to hear your thoughts on this.
Many thanks for your help.