iSixSigma

What factor affects my results? A case for DOE?

Six Sigma – iSixSigma Forums Old Forums General What factor affects my results? A case for DOE?

Viewing 13 posts - 1 through 13 (of 13 total)
  • Author
    Posts
  • #35921

    Tim G
    Member

    Dear Six Sigma Gurus,
    I am working on a Guest Laundry project in a hotel. We have noticed that the number of pieces of Guest laundry processed per day is not related to the number of guests in the house. The suspiscion is that it is related to the type of guests (or a combination them)  in the hotel on a given day e.g. Business Guest, Corporate guest, Government Delegation, Convention Group Guests, or Tourists on a Leisure trip alone or in groups.
    I need to investigate this relationship. I have investigated the DOE in Minitab 13 and must say I am stuck when it comes to determining the levels. I assumed that factors would be the different Guest categories and we identified 8 of them. I thought I would analyse the fugures for the month of May and would consider each day as a run of the experiment. Am I right? How do I set it up in Minitab so that I can get some result to accept or reject the hypothesis?

    0
    #102022

    Shereen A. Mosallam
    Member

    Hello Tim
    You have a very intersting situation.
    What I understand here that your only concern in your study is the type of guest(8 of them)
    The levels refer to each factor:
    *First decide, will you take the 8 types as your factors? this means you will have 8 factors. Screening experiment ill help you to choose from them the key factors (types)
    * Levels are for each factor. This means levels of each guest type.
    You have to go back to your historical data and check there levels (varies from — to –)
    * Then the very important question comes which should be in the beginning, what is the responce? What is the output that you are intersted to study here?

    0
    #102024

    Tim G
    Member

    Hi,
    I would like to know which type of Guest (Factor) affects the number of Laundry pieces on a given day. Like I said earlier there is no relationship between the total number of guests in house and the number of pieces going for laundry. Sometimes we have high occupancy but we do not experience a corresponding increase in Laundry business. Sometimes we are very busy in the Laundry and yet the hotel is not running a high occupancy. So which factor or Combination of factors affects the number of Laundry pieces?

    0
    #102035

    Robert Butler
    Participant

      Before thinking about a DOE you should give more thought to the data you have collected for May. You should also continue collecting data for several more months – perhaps even as long as a year. 
      Rather than trying for a numerical analysis of some type I’d recommend you plot the data in as many ways as you can imagine and then really examine the plots. For example, use your Minitab program to set up side-by-side boxplots using the 8 types of guests as you grouping variable.  Take a look at the summary statistics with respect to each group (mean, median, mode, standard deviation, etc) to see how they are similar and how they differ.
      When you have two months worth of data do the same thing all over again. Check for month-to-month equivalence within a grouping.  With two months worth of data you can also begin thinking about month-to-month-variation.  If you accumulate data for summer and fall you may begin to spot seasonal trends which may actually overwhelm group-to-group variation. 
      Think about other ways to characterize your guests and plot the data against these groupings as well. There is a very good chance that once you have done this, and discussed the trends in variance and center location by group, time, etc., you will find your graphs have pointed you to the answer to your question.

    0
    #102039

    Shereen A. Mosallam
    Member

    Hi
    So now you know that your response is number of laundry pieces.
    All what you have to do is to go back to your historical data to decide on the levels of each factor (guest type).
    Then you go back to Minitab to design an experiment. Minitab will give you a table of combination of the factors at different levels.
    You have to try these levels actually and insert the response (number of laundry pieces to your model) for analysis
    What I see here that you can use multiple regression instead of DOE. As you are not going to design experiments and run it in your case. You will not control number of guests of each type!!!
    So by using multiple regression for your historical data will help you to make your predictions.
     

    0
    #102055

    Manee
    Participant

    I am not sure you can do that in Minitab.  But you can use rdExpert software to DOE with multiple responses  and analyze the data for whole month. 
    Manee

    0
    #102058

    Tim F
    Member

    Tim GThis looks like a case of mulitple regression. You have eight different data points each day (the numbers of guests for each day for each of the eight guest types) whcih we could call x1, x2 … x8. Your response is the amount of laundry, y. You want an equation like
    k0+ k1*x1 + k2*x2 + … + k8*x8 = ySo put the daily figures into 8 columns, the amount of laundry into the next column, and do the regression. You could add more factors, like the # of guests in the restaraunt, attendance at conferences, daily high temperatures, etc. Then, as Robert wisely suggested, make lots of plots. Look at the predictions vs the actual laundry load; look at residuals; look at different days of the week; look at different seasons. Once you can predict the the load well enough for your needs, then you have done enough analysis. If you really like data, you could start a control chart of the difference between theory and reality.
    When you get outliers, look for a special cause. If there is some assignable cause for the variation, then add it to the model so that next time you will be able to predict the variation. Tim F

    0
    #102060

    PB
    Participant

    Tim,
    Until May 2004, I owned a dry-cleaner and I had account for all area hotels. My expeirence points to the following:
    1 – Length of stay of the guest. (Type of guest is negligible as typically the guest laundry is getting done by corporate customers)
    2 – Laundry Price (Hotels are typically overpriced by 20-25% than area dry-cleaners). If the prices are comparable, you may see increased volume.
    3 – Type of hotel. If the hotel is upscale it attracts more corporate clients who carry only enough clothes for a day/two day trip but stay longer as they rotate their clothes through cleaning.
    4 – Frequent guests to area will check out different hotels and amenities each offers. One thing they will do is check out laundry prices. If there is a match in terms of comfort and price (laundry as well), they will switch. (I have seen that as I reconize the name and the clothes.)
    PB
     

    0
    #102950

    R Peffley
    Participant

    You mention that you have data for one month on the number of pieces and the types of guest.  Do you have data on the number of pieces of laundry/day over a longer time period?  If so have you done a control chart?  Are there any time periods that appear to be out of control.  If so focus on the out of control days and see if you can identify differeces. 
     

    0
    #102995

    JC
    Participant

    PB is on the mark regarding guest type.  However, I disagree on his point regarding laundry prices.  Most hotel guests, do not make hotel decisions based on laundry prices.  Having worked in the hotel industry for over 15 years, from resorts to upscale properties, the breakdown by guest type will only show that it is the corporate guess who impact the amount of laundry being processed.  Since, they are the only traveler who travel with as little as possible, and have an allowance for such expenses. 
    I recommend gathering data from the rooms operations on the Length of Stays of corporate guests.  Then I would check to see if there is any relationship between length of stay of corporate guests and laundry receipts overtime.
    I would also be interested in how accurate Reservations forecasts are compared to actual.  Afterwards develop a formula for scheduling laundry personnel based on the factors you discover.
     
    JC

    0
    #103002

    Kim Niles
    Participant

    Dear Tim: You’ve got some great ideas here and I’d follow them first.  However, after you’ve performed multiple regression analysis and other forms of plotting to highlight what controls your situation, you might want to try to generate a mixture DOE from passive data.   It might be tricky to find representative data but if you are able to do that and if there are only three key factors that affect your situation then you’re end result will be a nice plot that will allow you to predict based on those three factors how much work to expect on a daily basis.  I’m going to attempt to attach an example plot in a MS Word document.  See attached if I’m successful. Sincerely,Kim Niles http://www.KimNiles.com https://www.isixsigma.com/library/bio/kniles.asp Attachment(s): MixturePlotExample.doc

    0
    #103016

    VAD
    Member

    Tim,
    Has there been a brainstorming on this issue with people that serve the customer in various areas?
    PB has given some good ideas in the thread of the mail, and i would rather derive more than one factor affecting response (laundry volume).
    say you find 7 such factors, then determine what factorial design with resolution you could go for your design of experiment.
    The problem at hand “laundry volume” appears more complex and would require a lot of data capturing to arrive at a good solution (season may also affect the laundry volume, just a thought).

    0
    #103023

    Jonathon L. Andell
    Participant

    There are some good commercially available software packages out there. One of my worthy competitors has a good one that mounts macros on Excel. The trouble is, it takes some time to come up to speed – it’s powerful but quirky.

    0
Viewing 13 posts - 1 through 13 (of 13 total)

The forum ‘General’ is closed to new topics and replies.