I’m on handheld device at the moment so I cant be of much help but I think you should search for studies on SF-36 Health Survey by John Ware. Some of the studies discussed about scoring Likert scale items and whatnot.

All the best!

The logic you describe parallels that of "Guttman scaling," on which there is an extensive literature. It is usually discussed as if only one item measured each level, but in principle 10 items per level is analogous. A key issue you'll need to deal with is that some firms may answer yes to (for example) 5 of the first 10, 4 of the next 10, and 2 of the next 10. Or maybe none of the first 10 but 7 of the next 10. These problems are related to the idea of "reproducibility" in Guttman scaling. An Internet search for Guttman scale, scales, or scaling should turn up some helpful sources.

Thanks Burke, I read a few publications and definitely seems what I was looking for. Do you have any hint regarding publications that compare this method with other (e.g. Likert Scale)?

Thanks again for your help

Sorry, I don't. I haven't used Guttman scaling myself; I just know it exists, and I sometimes cover it briefly in teaching courses on measurement. I guess you'll just have to search the web for useful sources.

Within each of your levels, do the 10 items themselves form a hierarchy? That is, do you expect a firm that answers yes to exactly 7 of the first 10 ought to say yes to the first 7 and no to #8, 9, and 10, rather than any 7 out of 10? If so, then each subset of 10 items is itself a Guttman scale, and you could test the reproducibility (as described in those web sources) separately for each subset of 10. Alternatively, if the subsets of 10 do not form hierarchies (7 yeses out of 10 could just as well be the first 2 and the last 5, or any other combination of 7 out of 10), then each of your subsets of 10 could be summed to form an additive scale (and perhaps tested for reliability using Cronbach's alpha or the like). Then you could use only those additive scales as the input to a Guttman scaling analysis and test their reproducibility in the conventional away. Either way (several Guttman subscales, or several additive subscales) would be helpful for data reduction, to get you down to only one measure (one subscale) per level, which is the way Guttman scaling is typically described, I think. Good luck with it! It's an interesting measurement problem.

Thanks for the "food for thoughts" Burke"! it's definitively triggering some new questions that I didn't ask myself during the construction of the questionnaire!

Here's one more idea -- you might find it fruitful to re-post your question on ResearchGate, but this time include the keywords "Guttman scaling" in the header. Maybe that will generate replies from people with direct experience in applying the method.