How to run a Facebook Marketplace sourcing desk

A Facebook Marketplace sourcing routine is four surfaces working together: searches that check written filters, lanes that tell you what to open next, Snapshot that says whether the ask is cheap, and Saved listings that you keep checking after the first pass. A Marketplace feed is none of those. It is a scroll.

Stand up the first search on a no-card Day Pass. One search, 24 hours, then add lanes only if the queue is real. Pass facts: Facebook Marketplace Day Pass.

Deals get missed when detection, judgment, and follow-up live in the same tab. The operator who is still refreshing a search messages later than the one who is not.

Why a Marketplace feed is not a routine

A feed mixes new posts, old stock, relists, and ads in one stream. A routine separates those jobs so each listing gets one decision.

Facebook Marketplace does not publish a saved-search SLA. Timing varies by account and day. Native notifications can arrive minutes after a listing is live, if they arrive at all. That is fine for a personal shopper. It is a bad operating system for several cities and a written ceiling.

A sourcing routine answers four questions in order:

1. What am I searching for?
2. What needs a decision right now?
3. Is this ask cheap, fair, or bait against the local tape?
4. Which listings should I keep checking after I leave the queue?

If you cannot point to a surface for each question, you are browsing.

What a search is for

A search is one set of saved filters: a Marketplace category, the filters that define a pass, and the cities it covers. Crawlbench checks busy cities every 10 minutes on Starter, every 5 on Growth, every 2 on Pro and every minute on Scale, as pricing shows on 18 October 2026.

Write the filters before you name the search. Example that a partner can defend: 2016–2021 Toyota Tacoma, max $18,500, Dallas–Fort Worth plus Austin, no salvage, no "parts only," no dealer credit language. If that sentence needs three "or" clauses, split it. Two tight searches beat one vague net.

Treat the search list as inventory, not bookmarks. Active uses one of your plan's searches. Paused keeps filters without checking. Draft is never live. Alerts off means checks are still running but no alerts go out. Needs attention is the pile you open first.

On Crawlbench, each search opens to its overview, the listings that pass, its filters, Snapshot, and its activity. Saved filter edits re-score collected history in the background. They do not hit Marketplace again.

Every search on a plan is checked at that plan's speed.

For vehicles, start from vehicle monitoring and keep make/model tighter than keyword hope. Baseline stock already on Marketplace fills your listings when you activate. New-listing alerts stay quieter: they fire for listings Facebook posted recently. Fill the queue. Do not train yourself to ignore the channel.

How queue lanes keep triage honest

Queue lanes are team-wide piles for Unseen, Saved, and Handled. Every search's listings sit in one list, with the search name on the card. There is no search dropdown on those piles.

Use them as tickets, not as folders you feel guilty about:

• Unseen is work. Nobody on the team has opened or glanced at the listing.
• Saved parks a listing you like but are not ready to message.
• Handled is follow-through. Someone reached the seller.

The page for each search stays the place for filter pressure, price bands, and who handles what. Lanes answer a different question: what does the whole team owe right now?

Run the same two-minute triage on Unseen. Read the reasons first. Kill parts cars, typo prices, and cities you will not drive today. Message only what survives. If three junk rows share a reason, edit the filters.

When to open Snapshot before you message

Market Snapshot is the price picture for one search: the median, the typical range, how many listings, days to clear, price-cut pressure, and whether the median moved this week. Use it to judge cheap, fair, or bait before you burn a drive. These are asking prices, not sold prices.

Read it in this order:

1. Sample. A thin window is directional. Do not treat five listings as a metro.
2. Median and typical range. Good deals often sit near or below the low end of the range where most similar listings sit.
3. Days to clear. Lower days means message sooner.
4. Price cuts. A hotter cut rate is a reason to save the listing, not to skip a true under.
5. Year or model ladder. On a multi-model search, pick one model before you trust a year median.

Every paid plan can show a price comparison on each listing: Great price, Good price, At market, Above market, Overpriced, Verify first, Thin market, or No comps yet. Snapshot is the layer above that: the whole price picture. It is plan-gated on Pro and Scale. An active Day Pass includes the same surface with a 7-day window. Starter and Growth do not. Details live in the Market Snapshot guide.

Concrete case. A 2018 Tacoma shows up in your listings at $16,900 in Dallas on a Tuesday morning. Snapshot for that search shows a $19,400 median and a middle band from $17,800 to $21,200. Volume is thick enough that one typo would not move the median. That row is a message, not a research project. A $9,400 ask in the same lane with "parting out" in the body is verify, then skip.

Saved listings and price drops

Saved is your team's list of Marketplace listings that Crawlbench re-checks on purpose for ask drops.

Add a row by pasting a Marketplace URL or saving a listing from its card. The listing does not have to pass a search.

Each row carries a movement track: ask when you added it, today's ask, an optional target, and the cohort median when sample is thick enough. Lane filters are Moved down, Target hit, Raised, and Flat. A target must sit below today's ask. Leave it blank when any drop is enough. Targets do not spend a check.

Use it when the listing is not a "message in the next two minutes" event: a fairly priced 4Runner you will buy at $16,500, a firm seller you expect to cut this weekend, or a URL a teammate found that no search would catch.

Price-cut alerts fire on your usual channels (email, Telegram, Slack, Discord). Two things put a listing in scope: any listing that passes a search, and any saved listing. Crawlbench re-checks each saved listing on purpose, so a cut reaches you even after the listing drops out of the feed. A listing that both passes and is saved still only alerts once. Cut alerts are not limited to recent posts.

Caps follow the owning account, pooled across that owner's teams, and they differ by plan. A saved listing stops being checked when it sells, leaves Marketplace, you stop it, or 60 days pass. See pricing for the live caps.

Owner, Admin, and Member can all add, retarget, and stop. Alerts go to every member with the channel on.

How should you run the four surfaces on a weekday morning?

Start in Unseen, then Snapshot, then Saved, before you open Facebook as a browser.

1. Open Unseen. Clear hard kills. Message the under-median rows that fit today's cities.
2. Move "interesting, not today" into Saved only if a human still owes a decision.
3. Save firm-but-fair listings with one written target below today's ask.
4. Open Snapshot on the two searches that produced the most Unseen. If the median jumped and your price cap did not, fix one filter dimension, not four.
5. Scan Handled for quiet sellers. A cut on a saved listing is often the second chance.

Loud queues are usually a filter problem. Same junk phrase: add an excluded term. Same city you never drive: drop the city. Same yearless wrecks: raise the year floor. After the first-day baseline fill, a lull is often steady-state new posts. Check coverage before you widen your filters into junk.

Crawlbench detects and filters. You send from your own Facebook account: Crawlbench never messages sellers, so your account stays safe.

Find with searches. Decide in lanes. Price with Snapshot. Follow with Saved. If one of those four is missing, you are back in the refresh button.