Searches

Searches and the activity log

What a saved search does on each run, what every filter changes, and how to tell a broken search from a quiet market.

What a search is

A saved search is a standing question you have asked of Upwork: which postings are worth your attention. It runs on its own schedule, independently of every other search you have, and each one keeps its own filters, its own draft mode and its own daily cap.

Creating one costs 5 credits. Everything after that is free except drafting: editing filters, pausing, resuming, reading the log and switching draft mode all cost nothing.

Having several narrow searches is usually better than one broad one. A narrow search gives you a daily cap and a draft mode that suit that kind of work, and its log is readable, which a firehose is not.

What each filter actually does

FilterEffect
KeywordsSent to Upwork as the query. This is the widest lever you have: everything else only narrows what these return.
Excluded keywordsApplied after the fetch, on our side. A posting containing any of these is dropped before it is scored, so exclusions cost nothing.
SkillsPassed to Upwork, and also given to the scorer as a hint of the kind of work you want. Not a requirement the posting must satisfy in full.
Minimum hourly rate / fixed budgetFilters the feed, and sets the floor a suggested bid is written against.
Client minimum spend and hiresFilters on the client's history rather than the posting. The strongest quality lever there is, and the easiest one to set so high nothing gets through.
Payment verifiedOnly postings from clients who have verified billing with Upwork. Worth leaving on.
Experience levelUpwork's own entry / intermediate / expert tag on the posting.
Poll intervalHow often this search looks. Minimum 60 seconds.
Max drafts per dayA spend cap, not an apply cap. Once this many proposals have been drafted today, further matches wait rather than draft.

Fast for the searches that matter, slower for the rest

Clients tend to read the first proposals that arrive, so a tight search you really want is worth checking every minute. A broad search you only skim loses little on a longer interval, and a quieter log is easier to read.

Reading the activity log

Every completed check writes a row to the search's Activity page, including the checks that found nothing — which are the ones the page mostly exists to explain. Rows are kept for 30 days.

Each row carries a one-line outcome:

OutcomeMeaning
DraftedMatched new postings and started writing proposals for them.
MatchedMatched new postings and alerted you. This search is set to alert only, so nothing was drafted and no credits were spent.
Daily cap hitMatched new postings, but this search had already used its drafts for the day. The rest are waiting in Matches.
Empty feedUpwork returned no postings at all for these filters. If every run says this, the filters are probably too narrow.
Nothing newUpwork returned postings, but you had already seen every one of them.
All filtered outNew postings arrived, but each one was dropped before scoring: too old for the poll window, or restricted to a country you cannot work from.
None relevantNew postings arrived, but none of them scored high enough against your profile to be worth a proposal.
Out of creditsPostings matched, but your balance was too low to draft. They are waiting, and drafting resumes as soon as you top up.
Run failedThis run did not finish. Nothing was charged.

Underneath each row are the counts it came from: how many postings Upwork returned, how many were new to your account, how many matched, and how many were drafted. Reading the outcome first and the numbers second is the fast way round; a row of five numbers on its own tells you very little.

Diagnosing a quiet search

Nearly every “is this thing working?” question resolves to one of four patterns in the log.

Most runs sayWhat to change
Empty feedThe filters are narrower than the market. Drop the client minimum spend first, then widen keywords. Client filters exclude far more than people expect.
None relevantPostings arrive but score under 0.50. The keywords are catching adjacent work. Check your Upwork profile is synced too, since it is half of what the score is judged against.
Nothing newNormal for a narrow search in a slow category: the same postings keep coming back. Lengthen the poll interval and stop watching it.
Run failedThe Upwork connection needs attention. Reconnect from Integrations; nothing was charged for the failed runs.

A paused search writes no rows at all. If the log simply stops on a date, check whether the search was switched off then.