Four seconds is the median visit to our website

Four seconds

In the seven days to 28 September 2026, 22 people visited nightfill.co. Between them they produced 23 sessions and 23 page views. Our own tracker records active time: time the tab is actually in front of someone, with the pointer or the scroll moving. It logged 241 seconds of it across 22 loads.

Four minutes and one second. That is the total human attention our website received in a week.

The average visit lasted 11 seconds. The median lasted 4. The 90th percentile was 23 seconds, and the longest single visit all week was 81.

The distance between 11 and 4 is the part worth reading twice. The average is carried by two or three people who stayed. Half of everyone who opened the page gave it four seconds or less, which is long enough to read one headline and decide.

Where the seconds went

Our homepage has eight sections: the problem, how it works, proof, our own numbers, pricing, FAQ, the founder, and contact. The tracker attributes time to whichever section is on screen. Of the 241 active seconds, it could attribute 74 to a named section. This is the entire week:

• The problem: 42 seconds across 15 views (2.8 seconds per view)
• Contact: 7 seconds across 8 views (0.9)
• How it works: 7 seconds across 10 views (0.7)
• Proof: 4 seconds across 8 views (0.5)
• Our own numbers: 4 seconds across 8 views (0.5)
• Pricing: 4 seconds across 8 views (0.5)
• FAQ: 3 seconds across 8 views (0.4)
• Founder: 3 seconds across 8 views (0.4)

The first section took 42 of the 74 attributed seconds. The other seven sections, added together, held 32 seconds of attention across the whole week.

Our pricing section, with the three plans, the free pilot and the entire commercial offer, the thing the page exists to sell, collected four seconds. Eight people had it on screen and not one of them stayed with it for a full second.

The scroll number that lies

Our dashboard said 30.4% of loads reached the bottom of the page. That reads well until you look at how the figure is produced. Every menu link on the site is an instant jump, and a single scroll event marks as reached every milestone below the new position. Someone who clicks Pricing publishes a load that hit 100% without having read anything.

The number for people who actually scrolled down is 13%, and it is a floor rather than a final figure: 4 of those milestone arrivals predate the version of the tracker that records where a load came from, and those cannot be classified in either direction.

If your analytics tells you people reach the bottom of your booking page, check first whether your anchor menu is inflating that percentage. It is an easy mistake and an expensive one, because it convinces you the problem is at the bottom of the page when it is at the top.

What this means in a 40-bed hostel

Your page is not our page, but the shape repeats: a lot of people in and straight out, and a few who stay. If your median looks anything like ours, you are writing for four seconds, not for a brochure.

Three concrete decisions that fall out of this:

The reason to book direct goes in the first screen, in words. Not in a badge, not in a rotating banner, not halfway down. If your advantage is 10% under the OTA rate, or breakfast included only here, or free cancellation until 24 hours out, it has to be readable without moving a finger. The section on our site that gets nearly six times more attention per view than any other is the first one. The same will be true on yours.

The price has to be visible before the scroll. Our pricing table got 0.5 seconds per view. If someone has to scroll to find out what a bed in a six-share costs, half your visitors have already left. A range is enough: from USD 14 a bed, from USD 48 for a private.

The gallery, the FAQ and your story are for the 13%. Do not remove them. For the people who stay, that is what closes the booking. But do not put them where everyone else's four seconds have to pass.

Two counters, two numbers

A useful note if you are setting up your own measurement. In that same week, Google Analytics counted 12 sessions on nightfill.co. Our own tracker counted 23. Neither is lying. They count different things, with different session rules and different exposure to blockers. Pick one as the source of truth for a given decision and stay with it. Comparing one dashboard's number this week with another dashboard's number next week is like measuring occupancy from the PMS one month and the extranet the next.

The test this argues for, and how to judge it

The problem section holds 2.8 seconds per view. The proof section holds 0.5. The conclusion is not that proof does not matter; it is that almost nobody reaches it. The obvious test, on our page and on yours, is to move one line of proof, a measured figure rather than an adjective, up into the first section, and leave everything else where it is.

Judge it on seconds per view over the same seven-day window, not on a feeling about the new layout. If the first section holds its 2.8 seconds per view and proof rises above 0.5, the move worked. If the first section drops and proof does not move, put it back. Write down both numbers before you change anything, because you cannot reconstruct them afterwards. We will publish ours here when the window closes, in whichever direction it goes.

The limits of this data

It is 22 visitors. This is not a study; it is one small week at one small company, and any of these figures could move a long way with ten more people. Nineteen of the 23 loads were desktop and 21 of 23 arrived with no referrer, so this describes people who already knew about us, not search traffic.

And the most honest limit: of the 241 active seconds, only 74 were attributed to a section. The other 167 happened between sections, in transitions, or before the tracker knew where the visitor was. The table above describes less than a third of the week's attention. It is the third we have, and it is more than we had a month ago, but we are not presenting it as the full map.

The point that does survive the sample size is the simplest one: the median was four seconds, and almost no site is built as though it knew that.

If this sounds like your property, see the three Nightfill tiers and start a free 14-day pilot on your own data: view pricing.