It doesn’t seem the tool works on unconfirmed places or paces set to street address
It might also just be a data issue with nolo data
Ah, more likely it’s because I’ve poorly documented the feature (or rather not documented it at all), so it’s not clear what it’s actually meant to be doing.
I’m going to rush out an update shortly (as soon as I find the cause of that visit edit crash) and hopefully in that update I’ll also get a chance to add an information dialogue that explains what this feature does.
But until then, I’ll explain here: What it does is it walks through all the recorded samples within the visit, skipping over any that already have a confirmed activity type, then for the rest it will convert it to stationary, but only if that sample is within the place radius (ie the orange circle on the map). Samples that fall outside that radius get left untouched.
The reason for that last caveat is to avoid the feature accidentally converting chunks of data to stationary that possibly weren’t meant to be inside the visit in the first place. Once something is converted to stationary, if it wasn’t meant to be, it’s a lot more fiddly to undo the mistake. So I’m quite cautious about anything that automatically does that, potentially leaving you with an even bigger mess to clean up. It’s one of those “oh no, what did I just do!” situations 
Oh, though your screenshot also highlights a bug / shortcoming of the new feature. Those segments/samples in red are “nolos”, meaning “no location data”. The check for “is this sample inside the place radius?” is going to fail on those because they don’t have location data to compare against. I should just make it skip that check and treat nolos as inside the place radius anyway.
K, I’ve added that to today’s todos. It’ll only be a one-liner. Easy to get into the next update.
Any chance you can make it a double confirm for the out of place ones?
I use this most often on site with poor data to clean it up, and in those cases or in first visit cases the circle is quite small and I’d love it if you just marked all things within a time range as stationary
It might not be the safest but I think with a double confirmation or some sort of toggle maybe (or maybe even on a second stationary redefine attempt) it would still work
I’ll have a think about it. I know exactly what you mean - it’s often the first one or two visits to a new place that need the most aggressive cleanup, and in those cases the bulk of the samples will be noisy messy ones that fall outside the place radius. So this new feature isn’t ideal in those cases.
It’ll be a matter of how best to go about it in the UI. Possibly a confirmation menu that offers “Do it”, and “Really do it - all of the things”. Heh. I’ll have a ponder on the wording.
Aside: I’m going to change the feature to be called “Clean up” (or some such), with a sweep/broom icon, and also add it to the Individual Segments view for path items. Basically a single quick button to “clean up the mess” in both visits and paths.
For paths I’ll have it change all samples to the most common activity type. With the above caveat that it doesn’t make changes to samples that’ve already been confirmed with a type, and with the extra caveat that it won’t convert stationary samples that have a high confidence score of definitely being stationary.
So as an example, say you’ve got an item that’s been automatically classified correctly as walking, but which has some short nonsense segments of cycling or car or some such scattered throughout. If you use the confirm/edit view to choose “walking”, that will clobber everything, even samples that you’d already confirmed as a different type, converting it all to walking - quite heavy handed. If for example you’d already confirmed a short segment within the item as running, because you did do a quick dash to the train platform to jump on the train, choosing walking on the confirm/edit view will clobber that bit, undoing your careful editing.
If you instead just want a cleanup of the unconfirmed cruft, only touching samples that haven’t already been confirmed, then this new “Clean up” button will be that light touch.
It could just be on second attempt you have a popup
The natural response to not getting the desired result is to try again, and then you could have a confirmation popup
It would probably prevent needlessly cluttering the UI
Yeah and on that latter issue, I started making stationary segments at the end and start of everything to prevent that. I don’t reccomended doing this as a general practice but it makes my data a lot cleaner looking
1 Like
I was just about to go with your suggestion, because it does seem a decent way to do it, but then realised there’s possibly a less intrusive way. Though it will remain to be seen whether it’s adequate or not.
What I’ve got it doing now is: If it’s the first or second visit to the place, it will use a wider radius - basically radius + 2SD (standard deviation), instead of radius + 1SD. (The orange circles on the map are radius + 1SD).
It might end up being better to use 3SD though (eg including ~99% of samples instead of ~95%). Anyway, no doubt we’ll figure that out after a bit of experience with using the feature.
I’m still waiting for Apple to approve the update I submitted two days ago. Usually they review and approve within 24 hours, but maybe submitting on the weekend slows things down. Anyway, once they’ve approved that one I’ll prep another quick update, so we can feel out how well this change is working.