Marketing from A to Z: the complete overview for 2026

Marketing is a system of decisions about who you sell to, what you sell them, at what price, why they should pick you, where they find out, how they become a customer, how they stay, and how you know which parts of it work. Advertising is one of the last steps. Without everything else it simply burns money faster.

This is a complete overview of the field as it stands in 2026. We wrote it because in meetings we kept missing a single document to point at when somebody asks what marketing actually covers and in what order it makes sense to tackle. Read it as a reference rather than a ten minute article: the contents in the margin are there so you can jump to whatever you are dealing with right now.

01If you have no time to read all of it

Eight sentences containing the rest.

02What fundamentally changed since 2024

Most marketing textbooks are now wrong about one thing. They assume a search engine stands between you and the customer, returning links, and that the job of marketing is to climb that list. But increasingly what stands between you and the customer is an assistant returning a finished answer, with links as a footnote underneath.

That does not mean everything before it gets thrown away. It means a new layer is added on top of a familiar foundation, and in that layer the contest is not for position but for citation.

The practical consequence is simple and uncomfortable at once. Value has moved from producing content, which is now cheap and can be done by a machine, to strategy, evidence, distribution and relationships, which are expensive and which a machine will not replace. Companies that used to hire somebody to write articles are discovering they can do that part themselves. Companies that never hired anybody to make decisions are discovering that this is suddenly the only part that counts.

The classic four Ps, product, price, place and promotion, still hold. For services they extend to seven with people, process and physical evidence. And physical evidence is interesting for digital products: it is not a brochure but the user interface, the documentation, how fast an e-mail gets answered and what an invoice looks like.

03Strategy and market

Without this block everything else is decoration. It is also the part that cannot be bought as a turnkey service, because the answers are inside your company and nobody will get them out of there for you. They can only be extracted properly, sorted and written down.

Segmentation means dividing the market into groups and picking one. It sounds obvious, and in practice almost nobody does it, because choosing hurts: it means saying who you do not sell to. So companies settle for „everybody who needs our products“, a sentence from which not one piece of copy or one campaign can be written.

For the ideal customer it is essential to distinguish the company from the person. In business to business the organisation buys, but a group of people decides, and they have different interests: the user wants it to work, the payer wants it cheap, and the technician does not want extra work. An offer that speaks to only one of them gets stuck on the others. The reverse definition is useful and underrated too: write down who you deliberately do not sell to. It saves more time than any tool.

The buying trigger is a question most companies cannot answer, and it decides the timing of everything else: what had to happen for the customer to be looking for a solution today? In manufacturing it tends to be an inspection, an accident, a new contract, a key person leaving or new legislation. Whoever knows the trigger knows where to be visible and at what moment.

Competition includes the zero option, that is the customer doing nothing, or doing it in a spreadsheet. It is usually the strongest competitor and most companies never count it. And a new source has appeared: ask an AI assistant for a recommendation in your field and see who it names. It is the cheapest competitive research there is and it takes five minutes.

Positioning is one sentence, not a paragraph

Positioning answers why a customer picks you rather than the alternative. It has to fit into one sentence and it has to be true enough that a competitor could not put it on their website.

This template works: for [segment] who [need], [brand] is the [category] that [main benefit], because [evidence]. The last part is what matters. Without evidence it is a wish, not positioning.

Positioning then produces the value proposition, in the customer's language rather than as a list of features. The formula that works: it does X, so you save Y, which Z confirms. A number always beats an adjective. „A fast website“ convinces nobody; „loads in under two seconds on mobile, measured in PageSpeed Insights“ does.

The last piece is the goal. Not an annual document nobody opens, but one headline metric and a quarterly plan. A split of effort that works in small companies: seventy per cent on what is proven, twenty on development, ten on experiments.

We go through this part with clients first, and in an audit it is the first area covered. Not because it is the most interesting, but because without it nothing else can be judged: you have no yardstick by which to say whether a website is any good.

04Brand

A brand is not a logo. It is the sum of what people think of you when you are not in the room, and now also of what a machine says about you when somebody asks.

The strategic part comes down to a few questions the owner has to be able to answer without preparation: why the company exists, where it is heading, what it decides by when that is inconvenient, and what it always honours. Out of that come the brand personality, the tone it speaks in, and the main claim with three pillars and evidence for each.

A new and underrated point: consistency of details. AI systems build an internal model of a company from what they find across its website, registries, directories and media. When the name, the field, the address or the description differ from source to source, the model is weaker and the company appears in answers less often. Using the same name everywhere has stopped being an aesthetic question and become a technical one.

Visual identity: the list every company should have finished

It does not have to run to a hundred pages. It has to be good enough that a new supplier or a new employee can open one file and get things right first time.

It covers the logo in every variant you genuinely need: primary, secondary, mark without text, monogram, favicon and a version for dark backgrounds. Plus clear space, minimum size and prohibited uses. Then a colour palette including semantic colours for success, error and warning, typography with a size scale and a licence, icons in one style, and a photographic style, that is rules for what gets photographed and how.

Two things that belong in an identity today and are usually missing from the manuals: motion, that is the speed and character of animation, and accessibility, that is contrast and legibility as part of the brand rather than an afterthought.

We build a basic identity, a logo and a consistent visual style in usable form, ourselves when you have none. When the brand needs to be exceptional rather than merely decent, we say so and hand it to somebody who does nothing else.

05Product, offer and price

Marketing is often handed the job of promoting something that cannot be promoted, because the offer itself is incomprehensible. So it pays to reach into this block before reaching for campaigns.

A portfolio is mostly about hierarchy: what is the flagship, what is the low risk entry product, and what is the tool for capturing a contact. The entry product is the most underrated thing in small companies. A customer who is going to spend two hundred thousand with you wants to spend twenty first and find out what working with you is like. Our own entry product is a marketing audit, for exactly that reason.

User experience is marketing, even when it is not called that. For digital products what decides is time to first value, that is how long it takes before a customer first sees that it is useful. Empty states, comprehensible error messages and readable documentation decide it. Documentation has a second benefit people forget: it is content AI assistants cite when somebody asks them about your product.

Price is a message, not just a number

Price can be set from costs, from the competition, or from value to the customer. Value based is almost always the right answer and almost nobody uses it, because it requires knowing what you save the customer.

Three things work from psychology: anchoring, that is the first price seen defining what looks expensive, three options with the recommended one in the middle, and rounding. Discounts cut both ways: they sell in the short term and say in the long term that the original price was not true.

A new point for 2026 concerns even those who keep prices private. A hidden price means an AI assistant cannot include you in a comparison. When somebody asks what things cost, the model names companies where it found out, and you will not be among them. It is not an argument for publishing prices at any cost, but it is a new item on the scales.

06Website and digital experience

A website is not a business card. It is where everything else leads, and the only thing you fully control. So we judge it as a sales tool rather than as graphics.

On structure, the homepage has to answer three questions within five seconds: who you are, who for, and what the visitor should do next. Beyond that a simple rule settles most arguments about structure: one page per problem. A company with five services needs five pages, not one page with five paragraphs. Customer and search engine then both have somewhere to aim.

The conversion layer is what happens next. Trust elements, that is references, figures and the faces behind the company, belong where doubt arises, not in the footer. Forms should be short, because the length of a form is directly proportional to resistance, and you can always ask more later.

On technical quality three things are in play today. Speed, because a slow site loses visitors before they can decide. Accessibility, which for many services in the EU has been a legal duty since 2025 rather than goodwill. And server side rendering of content, because AI crawlers cannot reliably read what is generated later in the browser.

This website is built to exactly those rules and you can check the numbers yourself in PageSpeed Insights. That is neither coincidence nor boasting: a site whose argument rests on speed cannot afford to be slow.

07Search visibility: SEO, AEO and GEO

The block that has changed most in two years and generates the most noise. These are not three competing disciplines but three layers of one.

AI assistants are estimated to handle roughly a quarter of searches this year and more than half by 2028. At the same time something most articles about GEO leave out holds true: the brands winning in generated answers are almost always the ones that mastered classic SEO first. Nobody skips the foundation.

That foundation has not changed. It is still about the site being indexable, having sensible addresses and headings, loading fast, carrying structured data, and not losing its links during a migration. A redirect map is the single most important document of a migration project, and its absence is the most common reason a company's traffic drops by tens of per cent after a redesign.

What changed is searching itself. An assistant does not search a whole sentence; it breaks it into sub-questions and assembles an answer from several sources. The practical consequence for content: cover a whole topic rather than one keyword, and answer the main question at the top of the page rather than a third of the way down.

What genuinely raises your chance of being cited in an AI answer

Nine things that work according to available studies and our own practice. The first two are necessary conditions; without them the rest is pointless.

This layer cannot be measured with classic tools. Sessions and click through rate say nothing about it, because often there is no click at all. It has to be tracked separately: which questions an assistant mentions you for, in what context, which competitors it names alongside you, and whether what it says about you is true. That last question tends to be an unpleasant surprise, because models draw on sources the company does not know about.

In our audit it is a separate area: we test ten to twenty real questions across four environments and document the result with screenshots. Most audits on the market do not cover this layer at all, because the methodologies they come from predate its existence.

08Content

Content is not a collection of articles, it is a system. The difference is that a system has three to five topics you want to be known for, a map onto the buying stages, and a cadence you can hold even in a bad month. Regularity beats intensity: six articles spread over six months achieve more than twenty written over a weekend and then nothing.

On formats one rule applies, and from 2026 it is worth applying hard: if anybody could generate your content in thirty seconds, it is not worth publishing. A rewritten Wikipedia definition convinces nobody and will not appear in AI answers, because the model knows that definition better than you do.

What has enormous value instead is your own data and your own experience. Figures from your own practice get cited in media, on social networks and in AI answers, because nobody else has them. Small companies have an advantage here that they usually waste: they sit on operational data they assume is boring, and which nobody else can publish.

Production now includes something that is a legal matter rather than merely a quality one: somebody specific has to stand behind content made with AI. Editorial review with a named responsible person is also the simplest way to avoid part of the obligations that arrive in August 2026. Details are in the chapter on law.

09Social networks and communities

The most common mistake with social is opening an account everywhere and doing none of them properly. One channel done regularly beats four done carelessly, and the choice of channel follows where your audience is, not what is popular.

Two things surprise almost every owner. First, a personal profile usually has several times the reach and credibility of a company page, because people follow people. Second, comments under other people's posts often bring more than your own posts, because you appear where the audience already is.

The biggest shift of this year is people leaving crowded public spaces for smaller groups where relevance matters more than reach. For a company that means membership of three good trade groups tends to be worth more than a page of your own with a thousand passive followers.

Anyone wanting to go further and build their own community has to accept it is a commitment measured in years: it needs rules, moderation, founding members and rituals, that is regular topics and meetings. A community then pays back threefold, because it is a source of content, of feedback and of evidence all at once.

10E-mail and your own database

The only channel nobody can switch off on you, where you pay nothing for reach and where delivery is not decided by an algorithm. Also the most underrated one.

The most interesting place in all of e-mail is transactional messages: order confirmations, invoices, access details, notifications. They have open rates newsletters can only dream of, and in most companies they are bare text with not a single next step in them. Yet they are read by somebody who is doing business with you right now.

The rest is craft. Segment at the point of collection, personalise by behaviour rather than by a name in the greeting, watch deliverability through SPF, DKIM and DMARC, and clean the database. And one question to ask yourself right now: when did you last write to existing customers without selling them anything?

11Paid advertising

Advertising comes once you know who you sell to, for what, and have somewhere to send the traffic. It amplifies what works. It only sinks a failing offer more expensively. That is not a platitude, it is the most common reason companies say advertising does not work for them.

What changed most in recent years: automated campaigns took over decisions about targeting and bidding. A marketer no longer wins by configuring an audience better. They win on creative and on the quality of the data input, that is which conversions get reported to the system and at what value. Testing dozens of versions is now routine, because AI can produce them cheaply.

On the economics, watch three numbers: cost of acquiring a customer against their lifetime value, where a healthy ratio is around one to three or better; incrementality, that is how much would have sold without the advertising; and the result of a switch off test. That last one is almost never run and is the cheapest way to learn the truth.

Campaigns here are run by Martina, who spent eleven years in performance marketing, most recently as head of digital advertising with a team of eight. We mention it once and for the avoidance of doubt: most of the advice above comes from practice, not from a course.

12Authority, PR and partnerships

In an era when anybody can produce content, trust is the main currency. And trust cannot be manufactured, only evidenced.

The cheapest evidence is reviews, and most companies still do not collect them systematically. They now have a second use: what public sources say about you shapes what an AI assistant says about you. A reference with a name, a company and a number beats ten anonymous stars.

A level above sits professional authority: publishing in trade media, speaking, original research, a methodology, membership of associations. In a small company the founder's personal brand is almost always stronger than the company's, and not using that is a mistake.

And the thing that gets forgotten entirely: partner distribution is usually faster for a small company than building reach of your own. One meeting with somebody who already serves your audience is often worth more than a month of writing posts.

13Sales and conversion

Marketing and sales are artificially separated in small companies. The customer does not perceive the divide; they travel one path from awareness to recommendation and drop out wherever nobody is looking after them.

The most expensive point on that path is usually the least watched: what happens to an enquiry after it is sent. The speed of the first reply decides more than most companies suspect, and no amount of better advertising substitutes for it. So in an audit we always send the client an enquiry of our own and measure how long a reply takes and how good it is. That single finding tends to have the biggest impact in the whole report, and to be the most uncomfortable.

That figure changes the order of priorities. Most of the deciding happens invisibly, before the customer contacts you, so you have to be persuasive where you cannot see anybody looking. In practice that means investing in the things that persuade in your absence: references with numbers, case studies, a price list, answers to objections, technical documents to download.

And conversion optimisation, that is testing headlines, forms and the pricing page, remains the highest returning activity in marketing for a simple reason: it needs no bigger budget, only attention.

14The customer after the purchase

Winning a customer costs more than keeping one. Yet most companies invest the other way round, because a new customer is visible and an existing one is taken for granted.

The first week decides. Onboarding and time to first value determine whether the customer starts using the product at all, and for services the same holds: what happens after signing shapes expectations for the whole relationship. Beyond that little is needed and almost nobody does it: send existing customers educational content, get in touch before renewal rather than once they start thinking about leaving, and ask for references systematically.

15AI across the whole of marketing

AI is not a channel. It is a layer running through everything else. 2026 is the point where experimenting stopped and deployment into live operation began.

The fundamental principle, on which practically every analysis this year agrees: AI speeds up execution, but a human leads strategy. Reverse the order and you will produce a lot of content nobody cares about and wonder why nothing is happening.

In practice AI is useful in four places today. In research, for competitor analysis, synthesising interviews and clustering topics. In production, for concepts, variants, translations, editing and tool prototypes. In personalisation, for recommendations and predictive scoring. And in operations, as an agent that watches mentions, prepares material or qualifies enquiries.

The rule that holds it together is simple: AI does the first version and the rough work, a human does the last version and the decisions. And beware one trap that is hard to see: AI gives you the market average, because that is what it was trained on. Use it to understand the field, not to choose your position. An average position is not a position.

It will not replace deciding where the company goes, your own experience and data, relationships, trust, taste and accountability for the result. That list has not got shorter in two years; the field around it has simply narrowed.

16Machine visibility and agentic buying

Agentic buying means the customer states an intent and some boundaries, an AI agent compares the options, picks one and possibly completes the purchase for them. It usually starts with routine, repeated purchases, that is where comparing is a chore.

The model settled during the year in a way that is good news for merchants: selling directly inside the chat did not catch on, and the prevailing approach is that the agent recommends and the purchase is completed on the merchant's own site. So you keep the customer relationship, the login and the data.

The condition is that the agent can read your offer at all. That means complete and accurate names, attributes, prices and availability, a catalogue exposed as a structured feed, delivery and returns terms in machine readable form rather than in prose, and the ability to measure which orders came via an assistant. For services the equivalent is a clearly described tariff, a public price or at least a range, and documentation reachable without logging in.

17Risks of AI and how to avoid them

Seven ways to come unstuck, in order of how often we see them.

Uniformity is the quietest and most insidious. When everybody uses the same tools with the same prompts, the same content appears and nobody remembers any of it. Differentiation has to come from your own experience and data, or it does not come at all.

Hallucinations are the most expensive. Never publish figures, quotations or legal claims without checking. From our own experience: while writing this website we published one claim about the advertising market without verifying it, and research subsequently disproved it. We rewrote the text and kept a note about it in the documentation, because that is more instructive than staying quiet.

Then loss of brand voice wherever the tone is not written down, overproduction, that is mistaking the ability to produce for the need to produce, and data: what you send to external tools, to whom, and under what contract. With personal data that is a processor relationship with all the duties attached.

The last two are existential. False authenticity, that is invented reviews and synthetic faces, damages trust faster than anything builds it, and it is now regulated. And dependence on a single platform: distribution built on one channel is always temporary, because you do not own that channel.

18Data and measurement

What you do not measure you do not manage. But measuring everything means measuring nothing, so this block is mostly about choosing.

The foundation is first party data: customers, contacts, in-product behaviour, orders. More valuable still, and far cheaper, is zero party data, that is what the customer tells you voluntarily when you ask. The quality of that data is a precondition of everything else, because an AI tool on top of dirty data produces confident nonsense.

Of all the metrics available, four groups matter to a small company: where people come from, how many convert, what that costs against what a customer brings in, and whether the brand is growing, that is searches for the company name and direct traffic.

Two things about evaluation are routinely overlooked. Last click systematically lies in favour of the bottom of the funnel, so advertising looks better and content worse than they are. And a test without a sufficient sample is just a more expensive guess, so half the „A/B tests“ in small companies decide on noise.

A rhythm that works: three numbers weekly that you actually look at, monthly evaluation of channels and a budget decision, quarterly strategy. The dashboard you open, not the one that looks good.

19Process, people and budget

Marketing in a small company usually fails on operations rather than on ideas. There is no brief, no calendar, nobody with the final word, and everything hangs on one overloaded person with no deputy.

The minimum that holds it together: a quarterly plan, a content calendar, a brief as a mandatory input for every task, and a clearly stated approver. Plus tools that do not overlap, because companies routinely pay for four things with the same function.

A budget rule for a small company: most of it into things that accumulate, that is content, search, community and your own database, and a minority into things that get consumed, that is advertising. Cumulative investments have the property of still working a year later at moments when you are not paying.

And if the answer to most of the previous paragraph is „nobody here can do that“, that is precisely the situation external marketing leadership exists for. A company that cannot support a full time marketer does not need an agency, it needs one person to set the order and make sure it happens.

20Law and ethics

This chapter is dull at the moment you read it and expensive at the moment you ignore it.

The basics are familiar: legal bases for processing personal data, records of consent, processor agreements with suppliers, a working unsubscribe link, information duties towards consumers, disclosure of paid partnerships and a ban on fake reviews. The new item that gets forgotten most often is a processor agreement with AI tools. When you send customer data into them, they are processors like any other.

AI and transparency, new from August 2026

From 2 August 2026 the transparency duties in article 50 of the EU AI Act apply. For marketing that means five specific things:

21The order of steps by company maturity

Most companies start at the end, that is with a logo and advertising, and wonder why it does not work. This is the order that makes sense, and the order we follow.

Phase one is the foundation, without which spending makes no sense. Who the customer is and what their problem is, the offer and the price, positioning and the main message, a basic visual identity, a working website with a clear action, and measurement. Six items, usually two to three months.

Phase two is visibility. Technical SEO and machine readability including access for AI crawlers, content pillars and the first pillar pages, one main social channel done consistently, e-mail collection with a welcome sequence, and the first reference with a measurable result.

Phase three is scaling. Only here do regular content with recycling, e-mail automation, paid advertising behind what already works organically, partnerships and PR make sense. Notice that advertising arrives as the fourteenth step out of twenty one.

Phase four is optimisation and defence. Testing and conversion optimisation, retention and expanding the value of existing customers, AI visibility as a discipline of its own with its own measurement, and possibly expansion into further segments.

22The sixteen most common mistakes

The list comes from what we actually see, not from a textbook. The first twelve always applied; the last four are new.

1. Starting with advertising before positioning is settled
2. Talking about features instead of the outcome for the customer
3. Trying to address everybody and addressing nobody
4. Inconsistency: three months of intensity, then six months of nothing
5. Changing strategy before it had time to work
6. Measuring only vanity metrics, typically follower counts
7. Creating content and not distributing it
8. Underrating the customer database you already have
9. Letting the brand look different every time
10. Selling before the customer has had time to build trust
11. Copying competitors instead of finding your own position
12. Doing everything yourself to exhaustion instead of picking two channels
13. Using AI for quantity rather than speed and producing content that differentiates nobody
14. Not knowing whether AI systems can see you at all and what they say about you
15. Keeping product data in a mess and being invisible to agents because of it
16. Deploying an AI chatbot without labelling it and without human escalation

23Checklist: twenty three questions

Answer them. Wherever you hesitate, you have a gap. It is the quickest way to turn this whole document into a task list, and it is a short version of what we go through in an audit.

Strategy and brand

1. Can I say in one sentence who I sell to and what problem I solve?
2. Do I know why a customer picks me rather than the alternative?
3. Do I know who I deliberately do not sell to?
4. Do I have a consistent look across everything a person sees from me?
5. Do I have a written tone of voice I can hand to a person or to an AI tool?
6. Would anybody recognise my content without the logo?

Website and visibility

7. Does a visitor know what to do within five seconds?
8. Have I accidentally blocked AI crawlers?
9. Are my prices and terms readable by a machine, not just by a human?
10. Do I know what an AI assistant says about me when somebody asks for a recommendation in my field?

Content and relationships

11. Do I have three topics I want to be known for?
12. Do I have a content calendar for next month?
13. Do I give distribution at least as much time as creation?
14. Am I collecting e-mail addresses, and do I have something to send the people who sign up?
15. Do I have at least three specific references with a result?
16. Do I contact existing customers even when I am not selling anything?

Numbers, law, AI

17. Do I know what it costs me to win one customer?
18. Do I know what one customer earns me over the whole relationship?
19. Do I look at the same three numbers every week?
20. Do I know what would happen if I switched advertising off for two weeks?
21. Do I have consents and processor agreements in order, AI tools included?
22. Do I label the AI chatbot and synthetic content where I have to?
23. Does every piece of published content have a specific person answerable for it?

24What to do with this now

This document has one flaw: it is complete. Reading it shows how many things you do not have, and that is paralysing.

Do not do everything. Take the checklist above, mark the three questions you hesitated over most, and spend the next quarter on those alone. Small company marketing does not lose by doing few things. It loses by doing twenty things halfway.

And if the answer to most of the questions is „I do not know“ and there is nobody to take it on, that is exactly the situation we offer marketing leadership for. Not as another single channel supplier, but as people who set the order, can do most of it themselves, and for the rest write the brief and watch the result.

Details about the audit, or simply tell us what is running today. We reply within 24 hours on working days.

25Sources

Material used for the trends section, as of July 2026:

• American Marketing Association, 2026 Future Trends in Marketing Report
• Gartner, Future of Marketing, predictions for agentic AI and authenticity
• Kantar, Marketing Trends 2026, micro-communities and the return on authentic engagement
• Deloitte Digital, Marketing Trends 2026, the move from experiments to full deployment of generative AI
• European Commission, guidance on the transparency duties in article 50 of the AI Act, applicable from 2 August 2026
• Princeton and the Allen Institute for AI, GEO: Generative Engine Optimization (KDD 2024), the measurable effect of statistics and citations on visibility in generated answers
• Google, guidance on optimising for AI search 2026, debunking myths about special files and tags
• 6sense, B2B Buyer Experience Report 2025
• Braze, Retail Customer Engagement Review, adoption of agentic shopping

The best sources are permanently elsewhere, though: your own analytics, competitors' public ad libraries, and conversations with your own customers. Those will tell you more than any trend report.