How our numbers are worked out

What is measured, and what is modelled

Two different kinds of number appear on this site and they deserve different levels of trust.

QR visits are recorded events. A tracking link can record a request and timestamp, but Nakovia does not verify that each request is a unique person or a person who saw the sign.

Impressions are modelled. There is no exposure sensor on a Nakovia placement. An impression figure is calculated from a stored traffic estimate or planning assumption, visibility inputs, time, and model limits. It is not a direct count of people who saw an ad.

The formula

A monthly impression figure starts with the listing’s stored daily-traffic value. New listings currently receive a location-type default, and an AI reasonableness check may compare that value with typical ranges. This is not a traffic measurement or independent traffic study. From there:

• Estimated vehicles per day or another stored planning assumption for the location type.
• Multiplied by how clearly the sign is seen: the sight line rating, the speed of the road, how far the sign sits back from the curb, and which way it faces. Each of these is a published factor between 0 and 1, never above 1.
• Multiplied by the share of the day the sign can actually be read. An unlit sign counts only the traffic that passes between sunrise and sunset where it stands, worked out from the latitude of its state and the day of the year, averaged across 24 days spread over a full year, and weighted by that location type's own hourly traffic curve. A lit sign counts the whole day.
• Sunrise and sunset use the Nakovia internal solar-position calculation. Repository fixtures at selected US locations use a 6-minute tolerance. This is an internal approximation, not an independently audited solar-time service.
• Spread over a month of 30.4375 days, which is the real average month length rather than a round 30.
• Then capped by the physical ceiling described in the next section.

The ceiling: what the street could physically produce

A chain of multipliers has no idea what is possible. Stack a strong visibility rating on a slow road on a corner right at the curb and a naive model will credit a spot with more viewings than there are people on the street. So the last step of every calculation is a hard limit built only from things you can count.

The limit is: vehicles per hour, times the hours the sign is legible, times Nakovia's 1.55 people-per-vehicle model assumption, times a 45 percent maximum notice-rate assumption.

Those values are model assumptions used to cap output. They are not listing measurements, audited industry standards, unique-person counts, or proof that anyone noticed the sign.

If a listing calculates above that ceiling, the ceiling is what gets published. No exceptions, and no override in the admin tools.

Why the night is quiet

Traffic is not spread evenly across a day, and it is not shaped the same way at every kind of spot. A construction hoarding is finished by six in the evening. A school gate has two sharp peaks and almost nothing in between. A venue with a late licence has an audience at midnight that a residential street does not.

So each location type carries its own 24-hour curve, and the live counter on a running ad ticks at the speed that curve says, using the clock at the sign rather than the clock on the server. A New Jersey fence should not be sitting at its 4am minimum while it is 10pm on the road outside.

The curves only ever move traffic between hours. They never create or remove any: each one is normalised so a day still totals the same amount however it is shaped.

Cost per thousand

Cost per thousand prices the whole campaign against the whole campaign, not this week against this month. One-time costs, the banner, the install kit and the installation, are spread across the run they bought, exactly as an agency amortises a production cost. Every monthly rent cycle the run will incur is counted, rounded up, because a part month still bills and the honest direction to round a cost is the expensive one.

This figure divides campaign cost by Nakovia's modeled impression estimate. It is not a measured media cost and should not be used as proof that another advertising channel is more or less effective.

How often the numbers update

Published impression figures are derived at render time from the listing’s stored traffic assumption and visibility inputs. Campaign progress also uses the current time.

A given past day always calculates to the same number. Yesterday does not quietly rewrite itself between two page loads.

QR event counts update when tracking requests arrive. A listing’s model inputs change only when the relevant stored data changes.

What this cannot tell you

Impressions are modeled opportunities to see, not proof that anybody looked.

Impressions are not unique people. A residential street is largely the same neighbours driving past twice a day, so a month of impressions on a quiet street represents far fewer individual people than the same figure on a commercial strip. Reach and frequency are not separated here, and until they are, treat a residential figure as exposures rather than as an audience.

The daylight window follows real sunrise and sunset, but it is still an annual average rather than a per-day almanac, because the figure it feeds is monthly. A December afternoon is darker than the average says and a June evening is brighter. It is honest at monthly resolution and approximate on any particular day.

The position used is the centre of the listing's state, not the listing's own coordinates. Across the width of a state that is worth a few minutes of sunset. It is deliberate: every page that shows an impression figure selects different columns, and a calculation that read coordinates would publish one number on a page that fetched them and a different number on a page that did not.

A listing with no state on file falls back to a flat 6am to 8pm window, 14 hours, which is the continental average and is more generous than the real figure almost everywhere.

Traffic values are estimates or planning assumptions and are not independently surveyed by Nakovia. The ceiling limits the model output but does not validate the underlying traffic value.

The weather does not appear anywhere in this model.

If a number looks wrong

Tell us which listing and figure looks wrong. We can review the stored inputs and model output. A correction depends on what the review finds.