Almost every "YouTube competitor analysis" ends in the same place: a table of subscriber counts, sorted descending, with the biggest channel at the top. It is satisfying to look at and it answers a question nobody needed answered. You already knew who was biggest.
The questions worth asking are comparative in a different direction:
- Which channels in this niche are accelerating, and since when?
- Which of them is publishing more or less than it used to?
- Which video beat its own channel's normal, rather than beating a smaller channel's absolute numbers?
- What is normal for this niche at all — because "500,000 views" means something different in true crime than it does in woodworking?
All four are answerable from public data, and none of them are answerable from a single snapshot. They need the same thing: a record over time, per channel, against a baseline that is not another channel's raw scale.
Step 1 — fix the set, and treat it as a set
Write down the channels. Ten to thirty is a workable niche; five is too few for a median to mean anything and fifty is more than you will read.
The important discipline is that the set is stable. Every statement below is of the form "compared with the rest of this set", so quietly adding three large channels in month two makes month two incomparable with month one. Add channels deliberately, and when you do, know that your baseline moved.
Step 2 — sample, don't snapshot
You need a value and its history for every column you care about. That means sampling on a schedule and keeping the samples. Analytics4You samples channel metrics on a five-minute cadence, which is more than enough resolution for subscriber and view totals — public subscriber counts are rounded to three significant figures, so most of the time a faster cadence just records the same rounded number again.
Two consequences of that rounding are worth internalising, because they bite anyone building this by hand:
- A large channel's subscriber count barely moves. A channel at 2.41 M displays 2.41 M until it crosses 2.42 M. Growth-per-day for a big channel is therefore a step function read from coarse data, and short-window growth numbers on large channels are noise.
- View counts are not rounded the same way, so channel-level view totals are the more sensitive growth signal at the top of a niche.
Step 3 — the columns that carry information
A niche comparison table earns its space with five columns, and each one exists to answer a specific question:
| column | the question it answers |
|---|---|
| subscribers | scale, for context only — never as a ranking |
| growth per day | is this channel accelerating, and since when |
| views per hour vs typical | is the current upload outperforming this channel's own recent pace |
| uploads per week | has the publishing rhythm changed |
| best recent upload | which idea worked here lately |
Note what the third column is doing. "Views per hour" alone is a scale number — a big channel wins it every time. Compared against what this channel itself normally does at this point in a video's life, it becomes a like-for-like signal you can read across a niche of wildly different sizes. That comparison is the whole subject of outlier scoring.
Step 4 — a baseline the niche supplies itself
The natural instinct is to compare a channel against an industry benchmark. There is no such benchmark for your niche that you can verify, and adopting a number you cannot verify is how a comparison becomes fiction.
The baseline that is verifiable is the median of the set you chose. For every column, compute the median across the tracked channels, and show it beside each channel's value. It has three properties that make it the right choice:
- It is computed from data you observed yourself, so it can be re-derived.
- It is robust to one outlier channel — the median does not lurch because a ten-million-subscriber channel joined the set.
- It moves with the niche. A niche-wide slowdown shows up as the median falling, which is information rather than everyone "underperforming".
Read the table as distance from the median, not as a leaderboard. "Two of the seven channels are publishing above the niche median and both accelerated in the same fortnight" is an observation. "Channel A is winning" is not.
Step 5 — read the rhythm, not just the totals
Publishing rhythm is the most under-used public signal in this category, and it is legible from the upload record alone:
- Cadence — uploads per week, over a window long enough to smooth a holiday.
- Consistency — whether the gaps between uploads cluster or scatter.
- Change points — the week a channel went from one upload to three, or stopped for a month.
A change point in cadence is usually a decision: a team hire, a format switch, a sponsor commitment, a burnout. You cannot see which, but you can see when, and "when" is enough to go and look at what the channel published on either side of the line.
One caution, and it is a real one: gaps between uploads are only trustworthy at day-scale resolution. YouTube shows an upload's age relatively — "3 weeks ago" — and a timestamp derived from that text inherits its coarseness. Analytics4You derives an upload's timestamp when it first sees the video and never rewrites it, precisely so the precision of that timestamp is knowable afterwards. Day-scale rhythm survives this; hour-of-day claims do not, which is a guide of its own.
Four patterns worth acting on
- The quiet accelerator. Growth per day above the niche median for several weeks while subscribers stay unremarkable. Something is working before it is visible in scale.
- The format switch. Cadence drops, average video length changes, packaging style changes in the same fortnight. Whether it works shows up two months later in growth.
- The niche-wide dip. Every channel's views-per-hour falls in the same week. That is almost never everyone failing simultaneously; it is seasonality or a platform-side change, and it is the single most common cause of a creator concluding their own channel is broken.
- The lone outperformer. One video well above its own channel's normal while the niche median is flat. That is the one to actually watch — and the packaging it used is readable from the outside.
What this analysis cannot tell you
- Nothing about audience overlap. You cannot see whether their viewers are your viewers.
- Nothing about the funnel. Impressions, click-through rate, average view duration and revenue are private to each channel, and no external tool has them. A number presented as a competitor's CTR is a model's guess.
- Nothing about cause. Every pattern above is a correlation with a date on it. The interpretation is yours, and it should be held loosely.
- Nothing about channels you did not track. The median is the median of your set. It is not "the industry".
How often to actually look
Weekly, for the table. Daily reading of a comparison table produces confident conclusions from noise — most of these columns need a week to move by more than their own measurement error.
The exception is the change stream: new uploads, retitles, thumbnail swaps and running tests are short-lived and worth seeing the day they happen. That split — a fast stream of events, a slow table of positions — is why Analytics4You separates the feed from the watchlist table, and it is a good way to organise the work even if you do it by hand.