What is the optimal number of SKUs to include for a conjoint study in competitive context?
As a rule of thumb, we recommend including SKUs that makes up ~80% of total volume share. The reason we are doing this is that by closely replicating real market, we can estimate preferences much more accurately for each SKU than if we were to only include a small number of brands.
The drawback here is that typically including more SKUs would mean that we are building a much larger conjoint study and therefore would typically require a larger sample size.
Exception: When the market is highly fragmented (when you see that the top 5 brands make up less than 50% of market share and there are a lot of SKUs or brands with ~1% share or less). In this case, we recommend to include only the top brands / SKUs that has more than 2% share even though the total market share of the brands included does not add up to 80%.
The reason for this is that when you include too many small SKUs (with less than 1% share), we will be presenting more options (brands or SKUs) that consumers are not aware of or care about. This will likely result in inattentiveness and low quality survey response behaviour (e.g. random clicking, etc.).