docs
Finder

YouTube discovery

How Finder finds YouTube channels for your product, what the two numbers on each row mean, and which channels it leaves out and why.

Finder searches YouTube for channels that already talk about your category. You get the channel, the videos that matched, and the reason each one is there.

YouTube and website discovery run at the same time. YouTube gets up to 270 seconds for provider work inside the scan's five-minute boundary. Any channel and evidence finished before that cutoff is kept. Requests still running at the cutoff are aborted, and they cannot add a late row or spend more of the scan budget.

What a scan searches for

Each scan runs these searches, in your brand's country and language, in this order:

  1. Four for each competitor: "Name review", "Name vs", "Name alternatives" and, later, "Name tutorial". Each ends with the word your keywords share most. Let's say your keywords are about affiliate software: the search is "Impact review affiliate", so YouTube returns reviews of the platform, not of a power tool.
  2. "Best X" and "X review" for up to three keywords that name a kind of product, like "saas affiliate platform".
  3. Your product keywords as you typed them.
  4. The competitor tutorials.
  5. Keywords that name a feature rather than a product, like "stripe integration", last.

The order comes from counting who each search found on real scans. People who'd promote software make videos that review a product or compare two. A feature phrase mostly finds developers and how-to channels. Each search asks YouTube for 50 videos, the most one search can return. A scan that can't afford every search stops partway down the list, so a weekly scan spends its searches on the reviews and comparisons.

How many searches a scan gets

YouTube measures its daily allowance in units. A search costs 100 units. Reading a video's or a channel's details costs 1. Until Google raises our daily allowance, each scan gets a set number:

ScanUnitsSearches
Your first scan4,500Up to 40
The weekly scan1,500Up to 10
A search you start1,500Up to 10

500 units of each are kept for reading details, so a scan never runs out halfway through a channel.

A competitor's name on its own isn't a match either. A video counts as naming Impact when the title or description also uses one of your category words, names impact.com, or puts a product word right beside the name, like "Impact review" or "Review of Impact". "Impact Driver Review 2026" is left out. Website titles follow the same rule.

A channel that turns up in five searches is still one row. The row lists every search that found it.

What you see on each row

Every YouTube row carries the videos that earned it. Each video shows:

  • Its title and link.
  • Its view count and the date it went up.
  • Which of your competitors it names, in the title or the description.
  • What the channel already promotes, when the video says so.
  • Its duration, when YouTube supplied one.

You never see a score without at least one video behind it.

No score on YouTube channels yet

For now, a YouTube row shows no fit score. Google has to approve an app before it can score channels from YouTube's numbers, like subscribers and views. We've asked, and we're waiting to hear.

Until then, each YouTube row shows:

  • YouTube's own counts: subscribers, total views and videos.
  • The videos that found them, with links back to YouTube.
  • Where YouTube placed them, like "3rd in YouTube for 'affiliate software'".

The list still puts the closest matches first. It sorts on the videos alone: where they ranked, and whether they name your competitors or a commission rate. It never uses a channel's size or how often it posts. Once Google says yes, the score comes back.

The two numbers

NumberSmall or lowMediumLarge or high
SubscribersUnder 10k10k to 100k100k or more
Average viewsUnder 10k10k to 50k50k or more

Average views is the mean view count of the channel's last ten uploads. It tells you what a new video is likely to get, which the subscriber count doesn't.

Some channels hide their subscriber count. Those rows say "unknown". We don't estimate it.

Sometimes we can't read a channel's last ten uploads. Then the average covers only the videos that matched, and the row says so. A missing duration, subscriber count or upload cadence stays unknown. Finder does not turn missing data into zero.

How active the channel is

We count how many of the last ten uploads landed per month, measured up to today. A channel that stopped posting a year ago reads as quiet, even if it used to post daily.

Uploads a monthBand
Under 1Quiet
1 to 4Steady
4 or moreBusy

The row shows uploads a month as a third number, with its band, and the score's activity part is read from it: busy scores 20 of 20, steady 14, quiet 8, stopped 3. Finder stores that normalized cadence beside the evidence. A later re-score uses the same fact even after the raw provider response has expired. Older rows without cadence use the most recent evidence date instead.

What they already promote

An affiliate link is a link with a code in it that tells the seller who sent the buyer. Finder reads each video's title and description for signs the channel already uses them:

  • A rate in their own words, like "earn 30% lifetime".
  • A link with a referral code, like ?ref= or ?via=. The row names the site it points to.
  • A link through an affiliate network.
  • A line saying they earn from links.

We only repeat what the video says. We never guess a rate. A discount like "30% off" is shown as a discount code, not as a commission.

Channels we leave out

Five kinds of channel match your keywords and still aren't worth an email. Finder hides them. Each one is still saved under Hidden with the reason and what Finder saw, and you can bring any of them back.

  1. Off topic. YouTube's search is loose. A search for one product can return videos about something else with a similar name. We keep a video when it names a competitor, or its title has two words from one of your keywords.
  2. A competitor's own channel. You won't recruit them.
  3. Content farms. A content farm is a channel that posts thousands of videos nobody watches. We remove a channel above 3,000 videos when its matched videos average under 500 views. Between 2,000 and 3,000 videos, it is removed only when the average is under 250. One exception: a channel that already links to your competitor as an affiliate stays in. That's your best kind of prospect. We also remove a tutorial mill: a channel with 300 or more uploads where every video that matched is at least two weeks old and has under 50 views. They post a quick how-to for every app, and nobody watches them. A new video isn't judged on its views, and neither is a small channel.
  4. The wrong audience. This one only runs if you sell affiliate software, which we take from your keywords: at least a third of them mention affiliates, referrals or commissions. Channels that teach people to earn from affiliate links match your keywords, but their viewers want to join programs, not run one. We compare their matched titles. When more of them teach earning than running a program, the channel goes. That includes tutorials for a network's own affiliates, like "How to create affiliate links on Impact", even though they name a competitor of yours.
  5. A company's own channel. You'd be emailing a marketing team. When the company is a big platform, like a payments company's developer channel, the reason says platform. Finder reads the links in the channel's about text first, and a plain link to one of your competitors, or to a big platform whose name is in the channel's title, settles it on the spot. No call. No cost. For the rest, an AI model reads the channel's name, the first line of its about text, the sites it links to and its video titles, and says whether a company, an agency, a directory or a person owns it. A link with a referral code in it is never held against a channel, because that's an affiliate's link and the opposite sign. A channel the AI hides goes under Hidden with the AI's own few words on why. You can bring it back.

Shorts and long videos

We keep every video's length. Finder marks anything of 60 seconds or less as a likely Short, so you can filter them.

Country

Many channels don't state a country. When they don't, the row says the country is unknown. We don't guess from the language or the name.

Find channels like one you know

Paste a channel link, a handle like @name, or a video link. Finder reads that channel's last twenty video titles and searches on two things:

  • Phrases the titles share.
  • The products its recent reviews cover. Another channel that reviewed the same product is the closest match YouTube has.

The channel you pasted never appears in its own results. Channels it features on its page are added when it has any. This search spends one search credit. If it fails, the credit comes back.

When you add a known channel or import an exclusion, Finder resolves a handle through YouTube's official channel lookup and stores the channel ID. Two forms of the same handle therefore point to one saved channel. If the lookup cannot resolve a handle, Finder leaves it unresolved and never claims it will exclude a channel.

When a scan fails

YouTube gives every app 100 searches a day, shared by every brand Finder scans, and Finder counts them before each search rather than finding out afterwards. When the day's 100 are gone, the scan still runs its website half, and the row says YouTube wasn't searched today instead of showing you half a list as if it were the whole one. Nothing is sent. The count resets at midnight UTC. Weekly scans are spread across the week, so one morning never uses the day's searches on its own.

If YouTube itself refuses a request because its own daily quota is spent, the scan fails and tells you why. That quota resets at midnight Pacific time. The next scan runs as normal.

Each scan also has its own request budget. If it runs short, Finder skips the last-ten-uploads step first and keeps the channels it already found.

Optional work that is not needed to rank the first review pool runs later. It never keeps the scan open past its terminal state.

What we keep

YouTube lets an app keep its data for 30 days unless it refreshes it. A weekly scan refreshes every channel it finds again. A channel not found for 30 days loses everything YouTube gave us: its name, country, about text, subscribers, average views, upload cadence, and each video's title, date and views. Its score goes too. We built it from those numbers. What stays is the channel's link and anything you did to the row, like saving it or adding a note. The row says YouTube's details were removed. Find the channel again in a later scan and it all comes back.