How we calculate waiting times

The full YakNak methodology: where the car counts come from, how they become minutes, what the colours and dashes mean, and where we deliberately stay silent.

A border waiting time cannot be verified on the spot — you only ever see your own queue. So it matters whether the site explains where its number came from. Below is our methodology in full, weak points included.

What is measured and what is calculated

One thing is measured at the border: how many vehicles are standing in the waiting area of a specific crossing, in a specific direction, for a specific vehicle type. That is a physical quantity and it can be counted.

Nobody measures the waiting time. It is calculated. Our formula:

time = minimum processing time + cars in the queue × average throughput rate

We hold the minimum processing time for cars at 15 minutes. That means the page shows 15 minutes even when the queue is empty, never zero: the procedure itself, on two sides of the border, is never faster than that.

The average rate is the second constant, specific to each crossing, direction and vehicle type. For cars leaving Ukraine ours ranges from 4 to 10 minutes per vehicle. We calibrate it against observation history: how long a queue of known length actually took to clear at that crossing.

Where the car counts come from

We combine four sources, and each has its own weakness.

The State Border Guard Service of Ukraine. Official counts of vehicles in the waiting area. The most reliable source by content, but it updates on its own schedule and at some crossings can freeze for several hours.

eCherha (echerha.gov.ua). The state electronic queue. It shows registered vehicles, which is not always the same as vehicles physically present.

Roadside cameras on the approaches. They give a real-time picture but depend on weather, light and a clean lens. At night and in fog recognition quality drops.

Traveller reports. People standing in the queue right now. The fastest source and the most subjective: "I have been here three hours" describes one car, not the crossing.

When sources disagree, we do not average them. A plausibility check runs first: if a new value would require the queue to clear faster than the crossing can physically process vehicles, it is discarded as a recognition error rather than published. That removes jumps of the "it was 800, now it is 3" kind.

Refresh rate and the age of the data

Data arrives from the sources roughly every ten minutes. The page additionally keeps a short two-minute cache so that it does not hit storage on every view.

Under every figure we print when it was obtained. This is not a formality: "updated 8 minutes ago" and "updated 3 hours ago" are two different levels of trust in the same number. When a source is silent we print a dash. A dash means "we do not know", not "there is no queue".

What the colours mean

We colour the waiting time in three states:

The thresholds are identical at every crossing. That is deliberate: a colour has to mean the same thing at a small crossing and at a large one.

Where the typical week comes from

The "when the queue is shortest" block on a crossing page is not a forecast. It is history.

We take the last 60 days of observations and compute the median queue length separately for each of the 168 "day of week × hour" cells. The median rather than the mean: a single day with a 300-car queue after an accident must not distort the picture of an ordinary Tuesday.

For the block to appear, a crossing needs at least 60 of those 168 cells filled — otherwise a "typical week" would rest on three random measurements. Right now 28 of our 37 crossings have that much history. For the rest we say the data is insufficient and name no specific hour.

Two details that are easiest to get wrong:

• the unit in this block is cars in the queue, not minutes. This number cannot be compared with a waiting time;
• the week starts on Monday.

What the figure does not include

The waiting time is the time up to the barrier. It does not include:

• the checks themselves: passport, customs, and an inspection where required. Allow another 15–30 minutes, more if you are carrying goods or driving a car that is not yours;
• the queue on the EU or Moldovan side, where that is a separate queue;
• sudden events: a change in the crossing's operating hours, a technical failure, a road closure.

The figure is therefore approximate by nature. It is good at ratios — "three times longer here than at the neighbouring crossing" — and less good at the exact minute.

Where we deliberately stay silent

Trucks. The same formula produces nonsense for lorries: with 177 vehicles in the queue it calculates a wait of several days. Until the model for freight is recalculated, we publish nothing about it.

Rail crossings. They have no car, coach or pedestrian lane, so there is no vehicle queue to report.

Crossings without telemetry. A few Moldovan crossings still supply no data. Their pages remain — with the description, opening hours and nearest alternatives — but a dash stands where an invented number would go.

How to check us and correct us

The most useful feedback is a discrepancy with reality: you stood at a crossing and saw something other than what the site showed. Write to us through the contact form — it is the fastest way to find a broken source or a mis-calibrated crossing.

Corrections to data are made as soon as they are verified. Corrections to article text are dated: every article shows when it was last updated.

The practical use of these figures is covered separately, in How to choose a border crossing.

Page updated 2026-08-14 · Editor: Anna Wiśniewska