There are three different kinds of entities, to simplify: Conditions, Actions, Targets, which needs to be connected together into triads:

condition-2345 - action-5 - target-4534

And there may be many such triples needed to be created in a "bag", but they can be formed one at a time, if it simplifies matters.

There are many conditions (by many I understand may be too much even for a dropdown), which may be long or complex, say, "quantityA1 > quantityA2 > -1", etc, and more can be added if there is no suitable one. May be labelled by shortened labels though.

There are not so many actions, say, max 15.

And again a lot of targets, known by names and URLs.

Conditions are "consumed" in the triples, that is, can be used only once, if this matters. The number of triples in one "bag" is limited only by number of conditions, can be large too.

I want to get maximum usability from the authoring interface.

So far my best idea is to have typeaheads text inputs and dropdowns, depending on the number of underlying sets ( similar to combination of (One to many) to many form? and Many to many form with hundreds of items ) :

(typeahead condition) - (dropdown action) - (typeahead target)

Another idea is to drag and drop after selecting "a tool", which corresponds to one of the actions. (Simpler, unanswered case: One-to-many list items in trigger-response model )

I think, the UX will be even better if many triples for the same "bag" can be done on one and the same screen. If it were on paper, I would arrange entities in three columns and connect corresponding entities with lines. Not sure what to do with some many of them though.

I do not even hope this will be usable on narrow screen, but nice if it will.

Note: the user may think from condition to target as well as from target to condition. One direction or the other can not be fixed beforehand.


Conditions are almost equal, and user-defined. More than that, some conditions are simply in the form (item=something), when some others my be more elaborate. Most recent / most popular are not applicable either.

However, some conditions may use 2D spatial coordinates, kind of location, and can be grouped by nearness. But I have not mentioned this because it is not 100%. Previous groupings (for previous bags) may induce groupings (and then the grouping is strong - "others" can be used), but many times existing "bags" do have nothing to do with new one.

  • Are all conditions equally likely to be chosen? Can you pick a subset of conditions that are more common and then provide a link to other conditions? – DaveAlger Jun 10 '15 at 16:20
  • Thank you for asking in this way (it did not occurred to me). However, the patterns aren't regular enough to distinguish likelihood. Well, actually there are 2 cases: either we can use previous bags as a hint, or we can not. But is it feasible to build two different interfaces for these two cases? I am not sure. – Roman Susi Jun 10 '15 at 16:33
  • The may be some knob for the user to provide hints like "reuse grouping", "central location", "completely new thing"... But it possibly makes implementation complex and requires feedback (not for the first iteration). With targets it is easier: recently added can somehow be more relevant. – Roman Susi Jun 10 '15 at 16:35
  • 1
    Are users likely to have multiple conditions for the same target or multiple actions for the same condition? -- codepen.io/run-time/full/qdjNzY – DaveAlger Jun 10 '15 at 17:09
  • Don't try and help the user if you are only guessing because it is frustrating when a computer is trying to help but only getting in the way. Sometimes putting favorites in a separate list can be helpful in cases where a user has a few they need quick access to. – DaveAlger Jun 10 '15 at 17:14

While forming a triple, you could have an "Actions" drop-down in the middle of the screen.

One at a time, you could display selectable conditions or targets in a "cloud" around the selected action. If the items in these large sets can be compared to each other in some way, you could create a 2-dimensional mapping of them based on a dissimilarity matrix. Finally, the items in the set could be abbreviated or "tokenized" in some way to fit the screen, but allow further detail to appear with a mouse-over before making a selection. After making a selection from one large set, the other large set could be presented.

  • Thank you for your input! I will come back when I will be able to properly assess it. – Roman Susi Jul 8 '15 at 18:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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