Glossary

Short for pay per click. Used in Google Ads, Meta Ads, Sklik and on marketplace platforms. The advantage is paying only for expressed interest; the drawback is that click prices rise with competition and in some sectors now exceed CZK 100.

The most common mistake with PPC is optimising for the click rather than what happens after it. A click on its own has no value.

Calculated as total cost divided by number of conversions. The usefulness of the figure stands or falls on what counts as a conversion — an order for one e-shop, an add-to-cart for another, giving two incomparable numbers.

CPA alone says nothing until it is set against margin. A cost of CZK 800 is excellent on a product with CZK 3,000 margin and ruinous on one with CZK 500.

A ROAS of 4 means every koruna spent returned four in revenue. The metric is popular because it is easy to calculate, but it works with revenue rather than profit — which is its main weakness.

Across a range with varying margins, optimising for ROAS leads to selling most of whatever earns least. That is why we pass margin into measurement and steer campaigns by it.

It is usually the cheapest part of a campaign because the audience already knows the brand. That is precisely why its results are easily overrated — some of those people would have bought anyway and the campaign takes credit for a settled outcome.

It is sensible to set a frequency cap and a time window. Following someone for a month with ads for a product they have already bought does the brand more harm than good.

A quick indicator of whether message and targeting fit each other. A high CTR earns nothing on its own — an ad promising something the site does not deliver often has excellent click-through and no revenue.

It can only be compared within one format and placement. Between search and display the figures differ by an order of magnitude and the comparison is meaningless.

It can be an order, a submitted inquiry, a phone call or a document download. What matters is that the definition matches what actually brings the business money.

For service businesses it pays to separate an inquiry, a qualified inquiry and a signed contract. The gap between the first and last figure is routinely tenfold, and a campaign steered by the wrong one spends on uninteresting leads.

A customer usually encounters a brand several times — sees a spot, later clicks a search ad, and finally arrives directly. The attribution model decides which of those steps gets credited.

Last-click attribution systematically overvalues search and undervalues channels at the start of the journey. So budget for brand campaigns cannot be steered by it alone.

As cookies are restricted and scripts blocked, a share of browser-measured conversions is lost. Server-side tracking reduces that loss because the merchant’s own system sends the data.

It is not a way around consent. Consent to processing is still required and server-side tracking changes nothing about that — it only improves data quality for visitors who granted it.

Two figures competing for the same budget. Broad reach with low frequency fails to make the message stick; high frequency on a narrow audience starts to irritate after a few exposures.

Brand campaigns typically aim for an effective frequency of three to five contacts. The exact figure varies with message complexity and existing brand awareness.

Short for out-of-home. It covers billboards, citylights, transit formats and digital screens. Its strengths are reach and visibility; its weaknesses are targeting precision and measurability.

Frequency decides with OOH. One site for a month is wasted spend; twenty sites over three weeks has a measurable effect on branded search.

The basic unit in which media space is traded. An impression does not mean anyone noticed the ad, which is why digital adds viewability — whether the placement entered the browser window at all.

Comparing CPM across channels is treacherous. A thousand impressions on television, on a billboard and in a mobile app are three contacts of very different intensity, and the same number means something different in each.

One hundred GRPs means the campaign reached the target group once over in total — say fifty percent of people twice, or a hundred percent once. The number alone therefore does not say how many people saw the ad.

That is exactly why a plan is never built on GRPs alone. Alongside them you track net reach — how many distinct people the campaign met at least once — and average frequency.

Instead of negotiating with a specific outlet, you buy an audience across thousands of sites. The system evaluates in real time how valuable a given user is to the advertiser and bids accordingly.

Its weakness is transparency. Several intermediaries sit between advertiser and publisher and each takes a cut, so noticeably less of the budget reaches the publisher than was spent. That is why a plan should also include direct buys with the outlets you actually want.

It is measured in money or in reach. It is useful compared against market share: a brand with a smaller voice than its share usually loses share over time, and the reverse also holds.

The figure is meaningful only relative to competitors, never on its own. In a thinly contested category a fraction of a budget buys a loud voice that elsewhere would go unnoticed.

In marketing this is the most important number that most often never reaches the campaigns. Without it, optimisation targets revenue and the campaign pushes the least profitable goods.

For e-shops we recommend passing margin into conversion tracking. It is extra work at setup and a difference in profitability within a single quarter.

This is not a slogan or a list of banned words. It is a set of decisions by which you can tell whether a text sounds like the brand even without a logo on it.

A tone of voice guide is only useful when it contains good and bad examples from real situations. Abstract adjectives like “friendly but professional” help nobody.

Positioning is not what a brand claims about itself but what people think of it. Work on it therefore begins by establishing the current state, not by writing aspirations.

The hardest part is not finding what makes the brand distinctive, but agreeing what it gives up. A brand that wants to be for everyone is for no one.

It settles when the shared umbrella should be visible and when it gets in a sub-brand’s way. The most common failure in company groups is that each division gradually starts communicating its own way.

OMNIA GROUP solves this by keeping the wordmark always monochrome and differentiating only through the entity’s colour rule. The group stays recognisable without the subsidiaries blurring together.

It is established by surveying two groups — those exposed to the campaign and a control group that was not. The difference in answers is attributed to the campaign.

It is one of the few ways to evidence the effect of brand advertising, where conversions cannot be measured. The measurement must be set up before launch, though; you cannot assemble a control group retrospectively.

The most common mistake in paid campaigns is sending all traffic to the home page. Someone who clicked a specific offer then has to find it again — and mostly will not.

A landing page has to deliver on the ad’s promise within seconds and offer one clear action. The more options it presents, the fewer people take the one that matters.

In practice it means sufficient colour contrast, keyboard operability, labelled form fields, text alternatives for images and a meaningful heading hierarchy.

Since June 2025 the European Accessibility Act has made these recommendations obligations for many commercial websites. As it happens, these are exactly the properties search engines reward too.

It comprises three indicators: largest contentful paint, interaction to next paint, and cumulative layout shift — content jumping around as it loads.

The effect on search rankings is indirect, but the effect on conversions is direct and measurable. With paid traffic, every extra second feeds straight into cost per acquisition.

Marketing and analytics cookies may only be set after consent is given. Pre-ticked boxes and banners without a reject option breach the rules and are fineable.

Refusing must be as easy as accepting. From a measurement standpoint that means accepting that some data simply will not exist — and not trying to work around it.