SCORM 1.2 Suspend Data--Seems Wonky to My Pea Brain

Thanks, Steve!

One thing I forgot to mention in my ponderous tomb of a first post.

I met yesterday with a new client for a new project.  Like the other client above, they're a major company here in Twin Cities and both are in the Forbes Top 150.  Both companies use Articulate as their e-learning products of choice.  Between the two companies, quite a few copies of Articulate software have been purchased.

For this new project, I've been engaged to help this new client with issues they are having with their LMS (Totara aka Moodle with window dressing :) ) for external learners and document best practices and learners.  And Totara, I believe (still need to get up to speed with it) only supports AICC and SCORM 1.2.

One of the issues they are experiencing may just be the same one I described above.  And there's a possibility if we can't get all the issues cleared they may might give up on Storyline and go back to Captivate.

Not making this up . . .

Hi Gerry,

I wanted to let you (and others who come across this) know that I've also included the thread here as a part of the report for QA so that they have all information accessible and so that when there is additional information I'm able to share it here as well. 

Cool, Ashley.  Thanks so  much. :)

Holy Moly Waz!

Gerry Waz-toy?

"SCORM and Peace"?

LOL, Mike. Whatever it takes, eh? :)

ROTFLMAO, Mike! ;)

Good Afternoon, Gerry.

Engineering has weighed in on your question, and I'd like to summarize what I know for you.

• There is a lot more to our suspend_data than just whether or not the learner got the question right and what they answered.  We store the location on the timeline, the navigation history (to support the Previous button), the current state of objects, variable values, interaction results, and much more.
• There are a lot of factors that go into what we store in our suspend_data, based on the features a particular object uses.  Our algorithm is fairly long and complex, so it would be difficult for us to document the specifics.  Also, it's proprietary.  :)
• We don't see anything to indicate that we have a defect here.  Everything we see tells us that we are being as efficient as we can be with the storage of suspend_data.
• It is expected that the suspend_data for Storyline 2 is larger than the suspend_data for Storyline 1.  Due to added features, Storyline 2 requires additional suspend_data.

Please let us know if you need anything else, and have a great day!

Justin - 

We've reported this before. Part of the issue is the order of storage. With the SCORM 1.2 suspend data floor, we've seen longer modules lose progress on restore. This might not seem like a big deal but it's fairly serious when thousands of support requests come in for the same issue or if you're the person that experiences the issue in a long slog. Ton of wasted time that, in my opinion, is completely unnecessary.

Seems like a relatively simple fix to me. Prioritize certain elements to the front of the string in the "proprietary" algorithm. 

Disappointed that this issue doesn't seem to matter...

Hi, Justin!  :)  Thanks for the detailed reply.  Much appreciated.  Also, please pass on my "thanks" to Engineering for considering my issue and taking the time doing so.

You've always been a great and knowledgeable professional to work with.  And so is Steve.  I really respect his expertise and knowledge, much as I do yours.  So when Steve makes a suggestion like he does above, I give it a lot of weight. 

I can't speak to his suggestion as he's far more knowledgeable on these technical things to me but it seems to have a lot of merit to me. So does Steve's suggestion have weight and possibility?

Also, Steve doesn't overstate the severity of this issue or the sheer hell it can put us folks in.  Support requests for things like this can drain an organization and infuriate learners, learners' management, and course sponsors. 

Or when you discover an issue like this at the 11th hour before releasing a course, it's a terrible and very stressful place to be as a developer as you have to scurry to find a solution and you need more detailed help.

I know that publishing to SCORM 2004 and Tin Can are probably the best solutions.  No argument there.  However, some folks can't use them yet--either their LMS does not support Tin Can or SCORM 2004 natively (like Moodle), or their LMS may not support SCORM 2004 "right" (like Saba) or Tin Can is not available yet for their LMS.

That said, I hope you folks, at the very least, can help us out more here.  The KB article that you folks have here is a great start:  http://www.articulate.com/support/storyline-2/exceeding-scorm-suspend-data-limits-sl2

But, IMVHO, we need more.

I'm not interested in how the proprietary algorithm works and even what all is stored.  But, I'd like to know, as a developer, ALL the things that I could do in Storyline to try and reduce resume data.

Do too many layers on slides impact things?  Layers on masters?  Button or objects with many states?  Does something like taking the answers in multiple choice or multiple select questions or interactions and making them just A, B, C, D, etc., and then add separate text boxes for the rest of the answers and then not randomizing answers, reduce resume data? Etc., etc.

You folks obviously know how resume and the algorithm works better than we do, so I'd like to strongly urge that you folks consider augmenting that KB article and provide detailed and comprehensive suggestions on how to possibly reduce resume data. 

You could probably do that relatively quickly and it wouldn't involve possible software changes.  How to design courses efficiently for AICC or SCORM 1.2 would a fantastic resource for developers to have in hand before starting a new course or in troubleshooting an existing one.

My initial post above lists some things to try.  But I don't know if they are all valid or not.  Or maybe if there are more things we can try.

We need you folks to do that.  Would that a possibility?  :)

One could make a cogent argument that, as you folks have increased the amount of data being sent with Storyline 2, you may have a responsibility to help us out more here.  At the very least, again, document in detail all the things that we as developers can do to potentially reduce resume data when SCORM 1.2 (or AICC) must be used for a course and the LMS strictly enforces data limits per the respective specs.

Also, having things in more detail from you folks would help us when discussing things with clients.  Telling a client that I have to change or remove or do something in a certain way because "I think" it may increase resume data and cause bookmarking/completion issues just does not have as much weight than you folks having something in black and white that I can refer to and show the client.

For the course I was working on, I saw the additional resume data for answering a quiz question or interaction vary wildly--from well over 400 characters per slide initially to over 200 characters per slide as I changed some things to over 100 characters when I created a new, vanilla version of the quiz.

That says to me that's it's clear we as developers can do things that influences the amount of resume data.    We need you folks to help us out here.  (And maybe strongly consider Steve's suggestion.)

Sorry for the long post.  But after living with this issue for a long time, I'm a little bit passionate on this . . .

Hello again, Gerry and Steve.

Thanks for being so persistent, guys.  We appreciate you keeping us in check, and I apologize if I missed the crux of this thread.  I may have had tunnel vision on the question, "Does Storyline's suspend_data really need to be so large, or do we have a potential bug here?"  It seems that there is more to to your concerns than that, and here is what I am hearing:

1. You'd like to have better documentation on precisely which slide objects and properties impact the size of our suspend_data.  For example, is it worthwhile to reduce the number of slide layers, or is it not?
2. You'd like to see us restructure our suspend_data to prioritize the most important data elements at the front of the string.  For example, we might decide that scores are more important than interaction results, so scores should therefore go first.

I'm happy to try and advocate for these requests, but I think I need Steve to clarify one thing first...  Given that we compress our suspend_data, is it really helpful to prioritize the data within?  Here's my train of thought, which may or may not be accurate:

• If I want to email a large text file to a friend, I might decide to zip it first.
• Let's say that the zip file is truncated in transit for some reason.
• Upon receipt, I wouldn't expect my friend to be able to read the portion of the file that didn't get truncated.  I'd expect my friend to not be able to unzip the text file at all.

Is this an accurate analogy, or is there something unique about the way suspend_data is compressed that would allow the intact portion of a truncated, compressed string to be readable?

Thanks again, guys!

Hey Justin -

When the data is truncated, it will consistently drop back to about half of the slides. So I suspect the compression is somewhat serial in nature as *some* of the slides are being tracked as visited when the data flows over the limit. Since variable strings are stored uncompressed, I suspect most  elements are compressed in chunks / blocks.

I haven't yet started poking around the methods used to compress and decompress the string. Knowing what order things are compressed, as a model, could help with construction decision-making. For example, what restore states / layer properties / variables, etc.. could be set to optimize the string.

The storage of question data is also a bit of a conundrum. It seems like it's storing the full string of the response rather than the state of the selected choice. Seems like there's some optimization to be had there, as the player shouldn't care about anything but the state of the choice for review.

S

Thanks, again, Justin, for taking this so seriously and your thoughtful response.  :)  Yes, for the near term, I'm more concerned with #1 as that, if done, could be implemented quickly without software changes, which, if they were to happen, would probably take a while,

For the long term, #2 sounds appealing as a avenue of research, along with Steve's conundrum of how question data is stored.

Again, appreciate your response and Articulate considering this.  THANK YOU!  I know supporting the "past" (AICC or SCORM 1.2) is often not as exciting as the "now and future technologies" (like Tin Can) but, unfortunately, some of us are forced to live in the past due to client requirements.

Thanks, guys.

As always, no guarantees, but I'll certainly share the information internally.  Thanks again!

Hi Justin and all,

I find myself having gone through the above process of discovery, and ending up at exactly the same question: What Storyline elements are the heaviest for adding to the resume data, and which are lightest? SCORM 1.2 is the most stable and most used SCORM version in the industry, so having some informed strategy for managing its resume data would certainly benefit a significant portion of Storyline users. I see that this thread has been quiet for 3 months. Any suggestions?

Thanks

Also-in Gerry's excellent post that started this thread, he listed several ways he'd come up with to reduce the accumulating length of the resume data string, including: 

- Reduce the number of questions in a quiz or cut any freeform interactions that you can.
- Use freeform questions instead of text based questions as less data may be sent.

Are freeform interactions and questions something we should try to avoid, or to use? I'd think the former, but I may be missing a distinction here. 

Thanks again.

An update . . .

First, the list of "several ways" to "reduce the accumulating length of the resume data string" in my post was really the product of some of us Heroes, particularly Phil Mayor, Steve Flowers, and myself.  Wasn't just me.  Have to give credit where credit due. :)

Steve and I got a nice note from Articulate concerning the issue.  We gave them some ideas and thoughts.  Have not heard much since.  

In my case, the issue was affecting two projects for two different clients.  One client was using Saba and the other Totara (aka Moodle 2.6).  The solution for both was to publish to AICC.  AICC is supposed to have the same character limit as SCORM 1.2.   However, for both LMS's, this limit was not enforced.

New version(s?) of Moodle (2.7, if I remember right) have a config switch to ignore the character limit for SCORM 1,2, so for one of my clients that might be their workaround for now.  Other LMS's may possibly have this option also.

I still believe for those LMS's that have the strict character limit for SCORM 1.2, some more further help or documentation from Articulate would be nice.

RE: "Are freeform interactions and questions something we should try to avoid, or to use? I'd think the former, but I may be missing a distinction here. "

I'd say the answer is "it depends. "  So much, at least to my pea brain, is dependent on what else you are doing in the course (e.g, how many other interactions, what kind, how many quiz questions, layers on master slides, etc.). 

Thanks for the update, Gerry.

We're using PeopleSoft's ELM 9.1, and it does enforce the 4096 character limit. So that option isn't open. It does support SCORM 2004 v3, but I haven't tested how robust it is yet, and so for short term am looking to manage the resume data string. 

It seems like a listing of stock Storyline interactions ranked by size of their additions to the resume data string would be most helpful. If this varies with some aspect of its use (size or number of movable graphics etc.) then an added note that calls an author's attention to that fact would be useful. It's like a calorie count on the menu of a fast-food restaurant. Let us choose our own poison based on some known data.

Well, hopefully, you can test SCORM 2004 v3 and it can work well, especially as its character limit is 64,000.

For one of my clients, the one using Saba, even though that LMS was supposed to support SCORM 2004 v3, the way Saba implemented SCORM 2004 was so #@#@!#@!#@!#@!#  that it was not usable.  Got some great help from Articulate, SCORM.COM, and Saba itself but working together we couldn't work around the issue.

Also, back in the days when Articulate 09 first came out we ran into issues when the Saba LMS divided the resume data into multiple database records because the field for the resume data could only hold so many characters. 

Every time the suspend data required more than one data record, a Presenter 09 course would act wonky on re-launch for a learner.  Among other things, bookmarking never worked. 

We never found out what was the problem but we strongly suspected the LMS was doing something with the resume data during initial LMS save or returning it wrong to the Articulate content when it reassembled it into one string on content re-launch.  Most often, the quick fix there was to turn off resume for any embedded Engage interactions or Quizmaker quizzes/questions so the resume data could get down to only taking one LMS data record.

Fun problems to debug! :)

The worst problems are always between products... 

When you say "turn off resume", in Storyline that would be setting the individual screen to "reset to initial state", right? Which I notice doesn't cut all addition to the suspend data string from that screen, but should hold it to a minimum (perhaps just the bookmark to the screen), am I correct? BTW, thanks for the help on a Sunday!

RE: 'When you say "turn off resume", in Storyline that would be setting the individual screen to "reset to initial state", right?'

Yes, that's one way and that's our assumption.  That produces slightly less resume info for the slides with that setting.

And there's also turning off resume at the course level in the player, which means no bookmarking resume for the learner.

Thanks, Gerry. No bookmarking for the learner isn't an option for us. The turn off resume in the individual screen sounds good, I'll need to see if there are side effects to that though. Thanks again for your help!

Well, it's empirical testing is always necessary. I ran a course to and through the 4096 char limit today, only to find that ELM 9.1 had zero problem with it. Bookmarking worked after the limit was passed. I changed computers and went back to take the post test, passed it, and ELM 9.1 recorded the completion as successful. So our hypothesis about a completion problem being due to the 4096 limit was incorrect. Which makes me, personally, happy.

The average char count for each screen was around 20. This went up to about 90 for test items in the post test. Screen-captured animations were still around 20 each, although often several of the animations led one to another before a "next" button came up--each one was itself about 20-30 chars. 

GREAT news, Robert--at least in terms of the suspend data issue not being the cause.

With your completion problem, do you have another thread here about that (if you haven't solved that yet)?  BTW, does the course have a final quiz determining completion?

As for the SCORM 1.2 resume data issue , I reached out to the Articulate person who contacted Steve and I for an update but so far no response.

Thanks Gerry.
At least I know what isn't the problem.
Since I am not sure what the cause is, I haven't started another thread. But I very well may.

The course does end with a 10-question post test. The test is what determines successful completion (at 80%).

I think what would be helpful here (and I anticipate to others) is a thread for specific debugging techniques for courseware LMS completion problems. The measurement of the suspend data string is one of them. I'll go ahead and start that thread, maybe see you there!

Thanks again to you and Articulate for your focus on the Suspend Data string.

Best
Robert