Searching (and mistyping)

Every search box in Summit Hub forgives a small typo. You do not have to spell a client's surname, a supplier code or a job name exactly: get close and the row you wanted will still be in the list.

What "close enough" means

• Type at least four characters. Below that, search still works, but it looks for exactly what you typed. Four characters is the point where guessing helps more than it gets in the way: with two or three letters, almost everything looks like almost everything else.
• One or two wrong letters is fine. Smthers finds *Jonathan Smithers*, kitchn finds *Kitchen renovation*, TIMBR-4X2 finds *TIMBER-4X2*. So do a missing letter, a doubled letter, and the wrong case.
• It is looking for one word, not the whole line. Searching kitchen will find *Kitchen renovation - 42 Ann St*; you never have to type the whole name.

What you typed always wins

Every search box groups its results before it sorts them:

1. rows that are what you typed,
2. rows that start with what you typed,
3. rows that contain what you typed,
4. rows that merely look like what you typed.

Only after that does the list's own order apply. So searching a price-book code like TIMBER-4X2 puts TIMBER-4X2 first even though AAA-TIMBER would normally sort above it, and a near-miss guess like TIMBER-4X3 sits at the bottom where you can ignore it. This is also why a search can return more rows than you expected: the extra ones are guesses, and they are always last.

The list's own order is not replaced, only applied *inside* each group. On the To-Do list, searching timber still shows the urgent to-dos before the low ones, but only among the to-dos that actually say "timber". A high-priority *Order timbre* is a guess, so it drops below them rather than leading the list. Clear the box and the order goes straight back to what it was.

Where it applies

Every box that filters the rows already on screen, plus these five, which ask the server and so can reach rows you have not scrolled to:

• Price Book: searches the code, description and category.
• Pipeline search: searches deal names across every pipeline.
• Client Book: searches client names, emails and phone numbers. Your chosen sort (name, spend, outstanding, activity) still applies inside each group above; only the guesses are moved to the end.
• The contact pickers: "Link a client to a portal login", and the pipeline's "Existing contact" picker. Both search contact names and emails.
• The To-Do deal picker: the + Link deal button on a to-do. It searches deal names across both HubSpot-managed and Summit-managed deals, and offers each deal once even when it exists in both.

Things worth knowing

• The To-Do deal picker shows a slice, and says so. It lists twenty deals at a time. When there are more, it tells you how many it is showing and offers Load more, so twenty deals no longer looks like all the deals you have. Keep pressing it and you will eventually reach a ceiling at around five hundred deals: the note stays, the button goes, and typing a few more letters is the way to what you are after. If the lookup itself fails, it says why (*Couldn't load deals* when there is nothing more specific to report), with a Try again button, instead of the old "No deals found", which could not tell you apart from an organisation with nothing to show. Try again picks up where it stopped, so a failure part-way down a long list does not send you back to the first twenty.
• % and _ are just characters in the contact pickers and the To-Do deal picker. They used to behave as wildcards in both (a leftover from how those searches were written) and no longer do. Every other box already treated them literally.
• The contact pickers stop offering more results once guesses appear. That page is already the end of the exact matches; refine the search rather than paging on. The Client Book's own paging is unaffected.
• Searching a deal's properties groups by the property's name, not its value. A property called what you typed comes first; one that merely *holds* that text is still listed, at the end. Its values are editable on the spot, so they are deliberately left out of the grouping, otherwise rows would move around while you were typing into one.