Hart, after Mbeki
A conference prints its check-in packs overnight, and desk 3 covers surnames H to M. The flow on the Grid reads Sign-ups closed → Prepare packs → Packs ready. In the morning a volunteer at desk 3 cannot find Hart's badge. Here is how it got lost:
- The print job sorts the 1,800 registrations by surname, inside "Prepare packs".
- A colleague adds the late sign-ups to the same job, after the sort, because nothing on the drawing says there is one.
- The job gives each desk the names in its range, in list order, so Hart lands at the end of desk 3's pack, after Mbeki.
- The volunteer looks under H, finds Haddad and then Ito, and sends Hart to the walk-in queue.
A later step depended on an order, and the drawing had no place that said where the order was made. Some steps deserve their own symbol for that reason. Multi-document marks a step that handles a set of documents (not one); Internal Storage is a working store, usually in memory; Sort orders a collection; and Collate combines or organizes sets into new ones. Like the other documentation symbols they are plain one-in, one-out steps on the Grid: they say what kind of step it is at a glance, and so what the next step may count on.
What each box promises the next step
- Sort promises that the whole set leaves in a known order. Its divided diamond is one step with one exit, not a branch. It gets its own step when a later step depends on that order, as the desk packs do; sort inline inside a generic Process and you hide the guarantee, which is how the late sign-ups ended up after the sort. Now "Add latecomers" sits visibly above it, and the title names the key, because "sorted" alone does not say by what. If nothing later cares about the order, there is no promise to draw.
- Multi-document promises several documents. "Badge sheets" is hundreds of separate documents, one print file per sheet. A single document, however many records it holds, is a plain Document, which Data, documents & manual input covers. Use Multi-document for one and the reader looks for separate documents that are not there.
- Internal Storage promises only that this is the flow's working copy, data kept at hand while the work runs. "Desk batches" holds each desk's print files and check-in list while the packs are built; once the packs exist nobody needs it. Data kept to be read later, like the registrations the desks look up on the day, is Stored Data. Choose by what the flow means to do with the data. Neither symbol guarantees anything: where data lives and how long it really lasts depend on the system behind the box. Stored data, displays & manual steps covers Stored Data.
- Collate (the hourglass) promises new sets put together from what it receives: here, one pack per desk, made from that desk's print files and check-in list. It is not a Merge. On the Grid, a Merge rejoins control paths and says nothing about data (Merge & path convergence covers it); Collate combines data sets on one path. On the Grid it has a single input, so two sets cannot arrive on two edges; they reach it together, here through the working store.
The same promises, nightly
Open the Nightly Batch Report challenge and name the sort and collate steps explicitly instead of folding them into a generic Process: the Collate builds its report from sorted sections, so the order has to be visible before it. The brief asks for a sort by department and date, so put both keys in the Sort's title. It also asks to hold intermediate results in Internal Storage, and its Multi-document comes last, as the multi-part report the Collate produces.