Timur Maystrenko, Digital Director at Method Media, explains which approach to use when calculating real reach in digital advertising.
Digital advertising still tends to operate with reach measured in cookies and devices, while business needs to know how many real people the advertising touched and with what frequency. With third-party cookies being phased out and multi-device consumption growing, classic metrics increasingly distort the picture. This article offers an example of a simple practical calculation that brings us closer to real human reach and a more accurate estimate of contact frequency in brand campaigns.
Why bother "converting cookies into people"?
Digital advertising most often operates with reach measured in cookies and devices, while business always expects an answer to a different question: how many real people did we touch, and how often did we do it per month, per year or over another period.
The problem is that a cookie is not a person but a technical identifier. As privacy regulation tightens and third-party cookies are abandoned, the industry increasingly recognises that classic cookie metrics give an ever more distorted picture.
This article is based on a large number of calculations and advertising campaigns carried out, and we would like to walk through an example of a simple practical calculation that allows you to:
- Get a realistic order of figures "in people", with possible corrections for the category, a specific client or the toolkit.
- Estimate annual frequency without mechanically multiplying by 12 or by any number of months the campaign runs.
- And, of course, discuss how reach is counted for clients and which main problems the industry faces today.
Why cookie reach is almost always overstated
1. Cookie deletion and identifier instability.
Research shows that a significant share of users regularly delete cookies, and browsers limit their lifespan. As a result, the same person may "appear" as a new user several times a year.
This means that part of the unique cookie reach consists of identifiers that have appeared again — the same people.
2. Multiple devices and multiple browsers.
The average digital user has several devices and browsers. According to Global Web Index, the number of connected devices per person in developed digital markets consistently exceeds three.
In advertising contacts this means one person easily turns into three to five "unique" cookies.
3. Audience overlap between channels.
Nielsen emphasises that without cross-channel deduplication it is impossible to count total reach and frequency correctly: the same person is counted many times across different environments.
4. Noise and non-human traffic.
Even in high-quality buying there is a share of IVT (invalid traffic). According to ANA, the average level of invalid traffic in digital campaigns is around 8–9%.
All these factors make the use of correction coefficients inevitable if we want to get closer to real human reach. Which agencies successfully apply.
Stable ID — better, but not always available
Login-based identifiers (email, phone, hashed ID, CRM onboarding) allow a more accurate estimate of real reach and frequency. However the industry today lives in a hybrid reality where cookies, MAID and various ID solutions are used simultaneously.
IAB Tech Lab shows that the market is in a transition stage and unified standards for person-level measurement do not yet exist. So in practice one has to work with cookies and build approximate conversion models.
A practical model: how we count "people"
In our practice we use the following basic formula:
Reach in people = (Reach in cookies / K1) x K2.
Where:
- K1 = 2.5 — the multi-device correction.
- K2 = 70% — the correction for audience overlap, noise, identification loss and IVT.
This is a basic formula that, thanks to its simplicity, can be universal for calculating most advertising campaigns.
In this article we do not consider how such an approach affects complex econometric or predictive models, where users are calculated with a deeper approach that includes many small parameters and specifics of the toolkit and formats.
Let us walk through the logic.
Why 2.5 devices?
Research shows that a user owns on average three to four internet-connected devices. However not all of them participate equally actively in advertising contacts, and some cannot participate at all.
So the coefficient of 2.5 is a conservative estimate of real multi-device contact. For certain categories (household appliances, for example) this coefficient should be increased to 3+, since in that case the use of "follow-up" mechanics will increase the number of devices.
Why 70%?
This coefficient accounts for:
- overlap between channels (Nielsen);
- technical impressions and IVT (ANA);
- identification losses.
Depending on the quality of the media split and buying, we would consider a range of 60–80%, where 70% is the basic balanced benchmark. With good media planning, in our experience, losses can be minimised to 10–12%.
Calculation example
Initial data:
- Impressions: 100 million.
- Average monthly frequency: 4.
- Cookie reach: 25 million.
Calculating reach in people.
25,000,000 / 2.5 x 70% = 7,000,000 people.
Calculating annual frequency.
100,000,000 / 7,000,000 ≈ 14 contacts per year per person.
Why multiplying monthly frequency by 12 is a mistake
The logic of "monthly frequency x 12" assumes that:
- the audience is stable;
- identifiers do not change;
- deduplication is not needed.
But data on cookie deletion, multi-device behaviour, the need for deduplication and the level of IVT show that these assumptions are far from reality.
So we prefer the following approach:
- first bring reach to people, then calculate annual frequency to check real values and evaluate the approach.
Frequent mistakes in calculating reach and frequency.
Besides mechanically multiplying monthly frequency by 12, other typical mistakes occur in practice:
- 1. Adding up reach across channels. When reach from different systems is simply summed without accounting for audience overlap. As a result the total reach can be overstated by tens of percent.
- 2. Ignoring multi-device. A cookie or MAID is taken for a person, although one user may come into contact with advertising from several devices and browsers.
- 3. Using monthly cookie reach as annual reach. A common situation where monthly reach in cookies is mistakenly perceived as unique reach for the whole campaign period.
- 4. No correction for noise and IVT. Even in high-quality buying there are technical impressions and invalid traffic. Ignoring them leads to a systematic overstatement of reach.
- 5. A universal coefficient without regard to category. The same coefficient is applied to FMCG, pharma and household appliances, although real multi-device behaviour differs across these categories.
Where this approach applies
Suited to: FMCG, pharma, household appliances, mass services, brand campaigns.
Not suited to: e-commerce, performance and lead generation, retargeting, CRM marketing.
Conclusion
Cookies are gradually losing value as a unit for measuring a person, and unified standards for person-level measurement do not yet exist on the market. Despite claims of an imminent "end of cookies", the industry continues to live in a hybrid reality where cookie metrics will be used in media planning for a long time.
In these conditions, clear and logical approximate models are especially important: they allow an honest conversation about the scale of advertising campaigns and media pressure without creating an illusion of precision.
The proposed formula does not claim to be universal, but it gives a predictable and explainable result that can be adapted to the category, the market and the quality of the media split.
A short checklist to avoid mistakes when calculating reach:
- Do not equate a cookie or a device with a person.
- Account for multi-device and multiple browsers.
- Do not sum channel reach without deduplication.
- Do not multiply monthly frequency by 12 without checking the logic.
- Allow for correction on noise, IVT and identification losses.
- Adapt coefficients to the category and the tool.
In the coming years we will most likely see the development of hybrid measurement models where cookies, MAID, Stable ID and modelling are used together. Until unified standards appear, it is exactly such practical approaches that allow maintaining control over the quality of media planning and the interpretation of results.