I have a set of questions for my research on Importance Menu Design towards Customer Purchasing Decision. After running Cronbach's Alpha for reliability test, it turned out less than 0.5 These are the set of questions in the survey.

Psychology Effects

  1. I read the menu thoroughly before I make an order
  2. I always order the same / familiar item without looking at the menu
  3. I always end up order more items than my initial plan after I look at the menu
  4. The menu is attractive and persuade me to order
  5. The menu is user friendly and interested


  1. The menu has clear dish classification
  2. I am attracted to big pictures on the top or left upper corner
  3. The pictures are appealing and encouraging me to order
  4. I always look at pictures before I read the menu descriptions
  5. The grid format used in the menu help me to read better

Dynamic of Menu Design

  1. The menu book is sturdy and has clean appearance
  2. The menu is too heavy to hold and make me uncomfortable to make decision
  3. The menu size is too big which make it difficult to flip over and take too much space on the table
  4. The menu color is encourage me to order
  5. The book format is suitable to cover menu range offer by restaurant

Menu Information

  1. The dish description is too complicated
  2. I prefer rounded nominal of price without currency symbol (eg. 10, 20) rather than decimal nominal of price with currency symbol (eg. RM 9.99)
  3. I am aware with the price given
  4. I ask waiter for more details before I order
  5. The font type and size can be read easily

All the question is answered on scale 1=totally disagree, 5= totally agree

Can someone help me to pin point my mistake? Did I use the wrong reliability test or something wrong with the question I constructed? Thank you. PS: I am a newbie in research

  • $\begingroup$ Some of your questions are positive and some negative. Did you reverse score some of the questions (e.g., "The menu is too heavy")? As an alternative to relying on one scale score, you might look at the correlation of each of the items with customer purchasing decision. That would let you sidestep the issue of whether menu attractiveness is unitary. $\endgroup$
    – Joel W.
    Sep 2, 2014 at 11:57

2 Answers 2


A few things come to mind as possibilities...

First, did you enter all 20 items in to one reliability analysis? If so, that is likely a big issue here, as alpha will be low if your scale is multidimensional. Try running the reliability analyses for the separate subscales (preferably after doing a factor analysis to make sure that they are actually forming 4 subscales- you may have a different number of subscales, and each would need to be examined separately).

Second, make sure that any items that are phrased negatively are reverse coded. Even a single item with the wrong coding can throw off your analyses.

Finally, you have some items that are double barreled. For instance, "The menu is too heavy to hold and make me uncomfortable to make decision" is actually 2 items combined. You are asking them to think about 1) whether the menu is too heavy to hold and 2) whether the size of the menu gets in the way of their decision-making. A person could reasonably agree with one part but not the other, which would make it difficult for them to respond. I may think the menu is too heavy, but it may not get in the way of my decision-making. In that case, I wouldn't really know how to respond. If this happens, you may have people responding with the midpoint of the scale just because they are confused, and this can make the reliability go down.

If I were you, I would conduct a factor analysis to determine how many factors are present, and then run a separate reliability analysis for each. I would also consider changing the wording of any double barreled items (items 2 & 3 of "Dynamic Menu Design" stand out right away), assuming that you can change items at this point.

Good luck!


Beside the important factors that have been pointed by Katherine S, I would like to add some more.

First, you did not conduct the wrong reliability test. Cronbach Alpha is often used to test the reliability in questionnaires like this in disciplines such as marketing, organizational behavior, etc.

I want to remind you that Cronbach Alpha is based on the interrelation of items. So the items have to be closely related to one another for a good Cronbach Alpha. For example, what are the chances that people who answered totally agree ("5") in Layout item 1 (The menu has clear dish classification) would also give the same answer in Layout item 2 (I am attracted to big pictures on the top or left upper corner)? Another example, what are the chances that people who answered totally agree ("5") in Menu Information item 1 (The dish description is too complicated) would also give the same answer in Menu Information item 3 (I am aware with the price given)? If you think, hmm.. not so much, then you have a problem. Because that means that the items are not "measuring the same construct" and Cronbach Alpha would automatically become low.

Cronbach Alpha can also be low due to the low number of items. You can simply boost your Cronbach Alpha with more items.

Lots of luck!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.