Data Peristence for Submitted-Edited-Resubmitted Responses Despite Question Skip-Logic

[RalphBeishline]

In my survey form, one main select_one question, "Outcome", determines later questions to be answered with the 'relevant' property, and takes on possible values essentially of 'Yes', 'No', and 'Unable to Test'. 'Yes' and 'No' responses receive one line of questioning, while 'Unable to Test' receives its own line of questioning.

However, during data reviews in the current collection stage, two responses were found to have responses to both the 'Yes/No' and 'Unable to Test' lines of questioning, something that the relevant property should not allow. The unusual results were able to be reproduced by submitting the survey as 'Unable to Test', editing responses to access the 'Yes/No' line of questioning, and then resubmitting the survey.

In my opinion, that doesn't seem like intended functionality and could certainly impact frequency results of a variable if the issue isn't noticed. Is there a way to properly NULL out responses when a submitted survey receives an edit that changes its skip-logic?

[JamesTedrick]

Hi Ralph,

Thanks for bringing this up.  Was this survey being filled out via the Survey123 mobile/field app or the Survey123 website?  Was the survey authored via Survey123 Connect or the Survey123 website?  Based on that, we can then examine the appropriate form for it's behavior.

[RalphBeishline]

Hi James,

The survey form was authored via Survey123 Connect and completed via the Survey123 app.

[JamesTedrick]

Hi Ralph,

Thanks for identifying which app.  The app itself should be clearing out values when questions are set to non-relevant (hidden). Can I get information about which version of the app you are using and is it possible for you to share you survey form so we attempt to reproduce?

[RalphBeishline]

We reproduced the issue with Version 2.1.6 & AppFramework version 2.0.15 after first noticing it in data extract reviews.

The survey form is attached. I'd prefer not to disclose the feeding 'SuppList17' .csv file during the survey's collection period unless necessary for your testing.

Thanks!

[JamesTedrick]

Hi Ralph,

No worries on keeping that part of the form private- it isn't relevant to the issue at the hand.  I just re-read the issue you are describing; just to confirm the issue with the additional values is on surveys that have been previously submitted and were then edited?  That's a bug that will be fixed for version 2.4.

[RalphBeishline]

The submit-edit-resubmit method is how we reproduced and seemed like a possible workflow for field staff for revisiting an outlet.

Data extract currently shows 7 responses with the issue... 'Non-completion type' should be relevant only for 'Non-completion' outcome.

What was the outcome?	Non-completion type (select one)
Refusal	OCN
Refusal	ONA
Refusal	BNA
Refusal	DNS
Refusal	BNA
Refusal	LOT
Refusal	UNL

This means some additional follow-up with field staff, but if the fix to prevent this is coming in 2.4, I look forward to it.

[JosephAlberts]

JTedrick-esristaff‌ just wanted to confirm this was fixed in 2.4. We just noticed this issue in a data set collected under 2.3. I don't have any devices still running 2.3, so I can't test under that version to verify this is what caused the issue I'm seeing.

Just to reiterate to make sure this is the same problem: Under a submit, edit, re-submit workflow, switching a value that influences the relevant logic for a later question (hides it under the new answer) wasn't nulling out the previously entered value.

[JamesTedrick]

Hi Joseph,

Yes, as of version 2.4 when a question previously answered is made non-relevant in a subsequent edit, a null value will be submitted for the question.  There is a technical limitation with date questions, as documented in Known issues—Survey123 for ArcGIS | ArcGIS 

[JosephAlberts]

Excellent, thanks!