How to vet YouTube creators: views per subscriber and more

To vet YouTube creators before you pay them, look past the subscriber count and check four things: views per subscriber (median recent views divided by subscribers), engagement rate (likes plus comments divided by views), upload recency, and whether the subscriber count is hidden. A channel with 48,200 subscribers whose recent videos get a median of 11,600 views has a views-per-subscriber ratio of 24%; a channel with 312,000 subscribers and a median of 7,900 views has 2.5%, and the smaller channel will usually deliver more viewers per dollar. All four inputs are visible on the channel page and exposed by the YouTube Data API, which reports subscriber counts rounded down to three significant figures [1]. This post works three illustrative channels through the checks, proposes starting thresholds, and explains where those thresholds break.

Why subscriber count is a weak signal

A subscriber count is a running total of everyone who ever clicked subscribe and has not unsubscribed or been removed. It does not decay when people stop watching. A channel that grew fast in 2021 on a topic it no longer covers can carry 300,000 subscribers and reach a few thousand of them per video today. For a sponsor, what matters is how many people will watch the specific video your product appears in, in the first few weeks after it goes live. Subscriber count predicts that poorly; recent view counts predict it much better.

There is also a measurement detail. The public subscriber number on YouTube is abbreviated: below 1,000 it is exact, but 12,345 subscribers shows as 12.3K and 1,234,567 shows as 1.23M [3]. The API's subscriberCount is likewise rounded down to three significant figures [1]. For vetting this does not matter much, because you are looking for differences of several times, not of a few percent, but it means you should never quote a creator's subscriber count back to them as if it were exact.

How to calculate views per subscriber

Views per subscriber is the median view count of the channel's recent long-form videos divided by its subscriber count. Three choices in that sentence matter.

• Median, not mean. One video that went viral will pull an average up by several times. The median of the last 10 videos tells you what a typical video does, which is what you are buying.
• Recent, but not too recent. Exclude videos published in the last 7 days, because they are still accumulating views, and stop at about 10 to 15 videos or 90 days back, whichever comes first.
• Long-form and Shorts separately. Shorts views are counted differently and behave differently, so a channel that mixes both should be measured on the format your sponsorship will use.

One more timing caveat as of October 2026. YouTube's API documentation says that starting on 24 August 2026, views on all formats count from the moment a video begins to play, including autoplay and hover-to-play [2]. If you compare a video published before that date with one published after, the later one may show more views for the same real audience. When you build a median, prefer videos on the same side of that date.

Engagement rate is (likes + comments) ÷ views for the same videos. It tells you whether viewers are paying attention. Dislike counts are not available: the API stopped returning them to anyone except the video owner on 13 December 2021 [2]. Upload recency is the number of days since the last public upload, plus the number of uploads in the last 90 days. A channel that has not posted in two months may be on a break, changing topics or winding down, and the audience that sees your sponsored video will be smaller than the back catalogue suggests.

Three illustrative channels, worked step by step

The three channels below are invented for this post, but the arithmetic is the arithmetic you would run on real ones. Each is a channel a SaaS brand might consider for a sponsored tutorial.

Channel A: a mid-size tutorial channel

Channel A has 48,200 subscribers. Its last 10 long-form videos that are at least 7 days old have a median of 11,600 views. Views per subscriber = 11,600 ÷ 48,200 = 0.241, or 24%. The median video has 420 likes and 64 comments, so engagement = (420 + 64) ÷ 11,600 = 484 ÷ 11,600 = 4.2%. The last upload was 6 days ago and there were 10 uploads in the last 90 days. Every check passes.

Channel B: large, but the audience has moved on

Channel B has 312,000 subscribers, more than six times Channel A. The median of its last 10 videos is 7,900 views. Views per subscriber = 7,900 ÷ 312,000 = 0.025, or 2.5%. The median video has 150 likes and 22 comments: (150 + 22) ÷ 7,900 = 172 ÷ 7,900 = 2.2%. The last upload was 71 days ago, with 2 uploads in the last 90 days. Channel B looks bigger on every directory listing, and reaches fewer people per video than Channel A.

Channel C: small, with a hidden subscriber count

Channel C hides its subscriber count. The API reports this as hiddenSubscriberCount = true [1], and the channel page shows no number. You cannot compute views per subscriber, so measure reach directly: the median of the last 10 videos is 3,400 views. Engagement = (190 likes + 41 comments) ÷ 3,400 = 231 ÷ 3,400 = 6.8%. The last upload was 3 days ago and the channel posts weekly, 13 uploads in 90 days. Hidden counts are a choice some creators make, not a warning sign in themselves; judge the channel on views, engagement and consistency.

What does a creator cost per viewer and per click?

Vetting is about price as much as quality. Suppose each creator quotes a flat fee for one sponsored tutorial: Channel A $500, Channel B $1,200, Channel C $150. These quotes are invented for the example. Assume the sponsored video gets the channel's median views, and assume 1% of viewers click the link in the description. The 1% is an assumption for the example, not a benchmark; your own numbers will differ and are what you should use once you have them.

Check the arithmetic: $500 ÷ 11,600 × 1,000 = $43.10. $1,200 ÷ 7,900 × 1,000 = $151.90. $150 ÷ 3,400 × 1,000 = $44.12. Clicks are 116, 79 and 34; cost per click is $500 ÷ 116 = $4.31, $1,200 ÷ 79 = $15.19 and $150 ÷ 34 = $4.41. Channel B costs three and a half times as much per viewer as either of the others. If its creator will only quote by subscriber count, offer a performance structure instead: a smaller fixed fee plus commission on the customers the video sends.

That last point is where a partner program helps. A creator who has a tracked link and a personal coupon code earns on every customer they refer for as long as your recurring commission lasts, which changes the conversation from "what is your rate card" to "how much of this do you want up front". In RelayWonder, a content campaign lets you post a brief with a per-piece bounty, accept or counter a creator's quote, and then pay the agreed bounty into the ledger once the video is delivered and checked, with the creator's tracking link and coupon attached to that specific video.

Starting thresholds, and where they break

The thresholds below are starting points we suggest for B2B and SaaS sponsorships. They are not published industry benchmarks, and you should move them once you have results of your own.

Where they break:

• Niche B2B channels. A channel about accounting software for dental practices may have 4,000 subscribers and 900 views per video. Its ratio is fine, its absolute reach is small, and its audience may be exactly your buyers. Judge cost per relevant viewer, not size.
• Evergreen search traffic. Some tutorial channels get most views from search over months, not from subscribers in the first week. Their median recent views understate long-run reach. Look at a few videos 6 to 12 months old to see whether views keep climbing.
• Shorts-heavy channels. Shorts can carry very high view counts and low engagement. Measure the format your sponsorship will use.
• Sponsored videos inside the sample. If half of the last 10 videos are sponsored, check whether those underperform the unsponsored ones. The API exposes paidProductPlacementDetails.hasPaidProductPlacement, which creators set when a video includes paid placement [2].
• The August 2026 view-counting change. Medians that mix videos from before and after 24 August 2026 are not like-for-like [2].
• Engagement that looks too good. An engagement rate far above the channel's own history, or comments that are generic and repetitive, deserves a read-through before you pay.

What the YouTube Data API exposes, and what it costs

If you vet more than a handful of channels, you will want to pull the numbers rather than copy them by hand. The fields you need are all in two resources [1][2].

• channels: statistics.subscriberCount (rounded down to three significant figures), statistics.hiddenSubscriberCount (true or false), statistics.viewCount, statistics.videoCount (public videos only), and contentDetails.relatedPlaylists.uploads, the playlist that holds the channel's uploads.
• videos: statistics.viewCount, statistics.likeCount and statistics.commentCount; snippet.publishedAt (when the video became public); contentDetails.duration (an ISO 8601 duration such as PT15M33S); and paidProductPlacementDetails.hasPaidProductPlacement.
• Not available: dislikeCount, which is returned only to the video owner since 13 December 2021, and favoriteCount, which has been deprecated since 2015 and always returns 0.

Quota is generous for vetting and tight for discovery. As of October 2026, a project gets 10,000 units a day for most endpoints, and channels.list, playlistItems.list and videos.list cost 1 unit per call. search.list and videos.insert have their own buckets: 100 search.list calls a day, at 1 unit each [4]. Vetting one channel takes about three calls: one channels.list for the statistics and uploads playlist, one playlistItems.list for up to 50 recent video IDs, and one videos.list for their statistics. That is about 3 units, so 200 channels cost about 600 units, 6% of the daily allowance. Finding channels through search is the bottleneck, at 100 searches a day.

Read the developer policies before you build a tool on top of this. YouTube's policies say stored API data must be deleted or refreshed after 30 days, and that API clients must not use API data to create new or derived metrics, with a separate permission process from June 2026 for audited analytics developers [5]. A one-off calculation in your own spreadsheet to decide whom to sponsor is a different thing from a product that stores and ranks channels; if you are building the second, follow the policy and its audit route.

Beyond numbers: audience fit and disclosure

Numbers tell you how many people will see the video. They do not tell you whether those people buy your kind of product. Before you pay, watch two or three recent videos end to end. Read the comments for signs of who the viewers are: job titles, tools they mention, questions about pricing. A channel whose comments ask "does this work for a team of ten" is a better fit for a team tool than one whose comments are about the presenter's keyboard.

Check how the creator handles sponsors. The FTC says a creator should disclose a paid relationship in the video itself, not only in the description, and that a platform's built-in disclosure tool should not be relied on alone [6]. A creator who already does this well is lower risk for you. Put the requirement in your brief: a spoken and on-screen disclosure, plus a plain-language line next to the link in the description.

A vetting procedure you can run this week

1. List 20 candidate channels from searches your buyers would run and from the channels your existing customers mention.
2. For each, record subscribers (or "hidden"), the median views of the last 10 long-form videos at least 7 days old, likes and comments on the median video, days since last upload and uploads in 90 days.
3. Compute views per subscriber and engagement rate. Drop channels that fail two or more of the starting thresholds, unless they are niche channels whose audience is exactly your buyers.
4. Watch two videos from each remaining channel and read 50 comments. Note audience fit in one line.
5. Ask for a quote. Compute cost per 1,000 views from the quote and the median views. Offer commission on referred customers alongside a smaller fee where the fixed quote is high.
6. Write the brief with disclosure requirements, the tracked link and the coupon code, and agree how delivery will be checked before money moves.

RelayWonder's Hunter runs the numeric part of this for YouTube results it finds, showing views per subscriber, engagement rate, upload recency and hidden subscriber counts next to each channel, so you spend your time on the watching and the conversation rather than on the spreadsheet. The judgement about fit is still yours.