94% of B2B buyers had ranked their shortlist in order of preference before they first spoke to a seller, according to 6sense's 2025 Buyer Experience Report, a survey of nearly 4,000 buyers by a company that sells intent data. Gartner's survey of 646 buyers in August and September 2025, reported by Digital Commerce 360, found 45% had used AI tools during a recent purchase. The site argues to people deciding mostly without you, so the argument must be settled before anyone draws a hero section.
This guide is a method for writing that case down: four inputs, positioning, audience-to-page mapping, page briefs, the content inventory and the sign-off. Page types are covered in The B2B SaaS Website Playbook: Page Map, Message Hierarchy and the 4-Week Build, and design choices in B2B Website Design in 2026: The 9 Decisions That Decide Whether It Books Demos. This post is the document that comes before both.
What the strategy document contains, and in what order
Nielsen Norman Group's Jen Cardello separates information architecture, meaning what content exists, how it relates and what it is called, from navigation, its visible tip. The architecture is documented in spreadsheets and diagrams, not wireframes. In her example, a team picks a navigation pattern because it looks good, then finds the content needs more levels than the pattern supports. A page map drawn first in a design tool is navigation pretending to be strategy.
So the strategy lives in a document. We would structure it in seven sections, each built from the one before.
| Section | Question it answers | Main input | Signed by |
|---|---|---|---|
| 1. Positioning | Alternatives, customer, category | Interviews, win/loss, sales | Founder or CEO |
| 2. Audiences | Who reads the site; what each asks | Interviews, win/loss | Marketing and sales leads |
| 3. Demand map | What people search; which pages earn visits | Search Console, GA4 | Marketing lead |
| 4. Page map | Which pages exist, how they nest | Sections 1 to 3, card sort | Marketing lead |
| 5. Page briefs | Job, audience, action and proof per page | Sections 1 to 4 | Each page's owner |
| 6. Content inventory | Keep, update, cut or write | Crawl, analytics | Content owner |
| 7. Measurement | What counts as success per page | Page briefs | Marketing operations |
Source: our recommended structure
The order matters. A page map written before positioning tends to mirror the org chart; a brief written before the inventory promises proof nobody has. Tracking for section seven is in The Analytics Stack for a Four-Week Site Build.
The four inputs, and what each cannot tell you
Customer interviews. NN/g's User Interviews 101 (2023) calls interviews attitudinal: people report what they remember, with poor recall and a pull towards the answer that sounds good. Ask about specific past events instead. For a website, that is the last purchase in the category: the trigger, the alternatives compared, the search terms, who else had to agree and which page got forwarded internally.
NN/g's Maria Rosala suggests starting with five or six interviews and stopping when new ones stop changing the themes. She cites one study where 20 to 30 interviews uncovered 90% to 95% of customer needs and another where saturation came at 12. A narrow question such as "how did you buy" saturates sooner.
Win/loss. Forrester's July 2025 best-practice report treats win/loss as tracking win rates, analysing why deals are won or lost and changing go-to-market strategy, with direct buyer feedback as the key step. Lost deals name the alternative you lost to, the first input positioning needs. The CRM's "reason lost" field is the seller's view; a call with the buyer is the buyer's.
Search demand. Search Console's Performance report shows queries and pages with clicks, impressions, click-through rate and average position, defaulting to three months; widen the range. Google's October 2022 deep dive sets out two limits: anonymised queries, those not issued by more than a few dozen users over two to three months, are excluded, and the interface exports at most 1,000 rows, against up to 50,000 a day through the API. Niche B2B problem language can sit below that threshold, so interview language belongs beside it.
Analytics. The GA4 Landing page report lists the first page of each session with sessions, users and key events. It shows where entrances really start, which is often not the homepage. Every page with real entrances needs a place in the new map or a redirect.
| Input | Tells you | Cannot tell you | Enough for a page map |
|---|---|---|---|
| Interviews | Buyers' words, triggers, alternatives | What people actually do | 5 or 6 per audience, until themes repeat |
| Win/loss | Which alternative won, and why | Why some buyers never arrived | Recent wins and losses, both |
| Search Console | Category and problem language; pages that rank | Anonymised queries; who searched | Longest range available, via the API if large |
| GA4 landing pages | Real entry points; paths to key events | Why visitors left | Twelve months, by channel |
Source: NN/g, 2021 and 2023; Forrester, 2025; Google, 2022; the last column beyond NN/g's guidance is our recommendation
Positioning before pages
April Dunford's Quickstart Guide to Positioning (2021) sets out five components in a fixed order: competitive alternatives, differentiated capabilities, value for customers, target customer segmentation, and market category. Step one is finding out what customers see as the alternatives, not researching competitors online. Best-fit customers are the ones who care a lot about the unique value.
In a March 2026 guest post on Lenny's Newsletter she puts the heads of marketing, sales, product and customer success, the founder or CEO and a couple of experienced account executives in the room, and calls disagreement about what to position against the root of most weak positioning. In Positioning Prep Part 1 (January 2026) she sets three decisions first: position for customers, not investors; decide whether it is a product or the company; and write for the deal champion, who builds the shortlist and recommends to the economic buyer.
Each component then decides something concrete in the page map.
| Component | What it decides on the site |
|---|---|
| Competitive alternatives | Comparison pages, and what "why change" argues against, including spreadsheets and doing nothing |
| Differentiated capabilities | How many capability pages exist: one per capability the alternatives lack |
| Value for customers | The headline and outcome on each capability page |
| Target customer segmentation | Whether segment, role or industry pages exist, and their labels |
| Market category | Homepage headline, title tags and target queries |
Source: Dunford, 2021; the mapping is ours
The Playbook's message hierarchy (category, problem, outcome, proof) is how each page speaks. Positioning decides which pages exist at all.
Map audiences to pages, not pages to audiences
6sense found buyers initiated 79% of first contacts in 2025, and 94% use large language models during a purchase, most often to compare vendors. Gartner found 67% favour a rep-free experience. The audience list is a list of people with questions they will answer elsewhere if the site does not.
Build it from the interviews: who had to agree, and what each asked. Write one row per question, with the page that answers it. A question with no page is a gap; a page that answers no one is a candidate to cut; a page answering five audiences probably needs splitting.
| Audience | Question from interviews | Page that answers it | Primary action there |
|---|---|---|---|
| Deal champion | How is this different from what we use now? | Homepage, comparison page | View pricing or product tour |
| Economic buyer | What does it cost and what does it return? | Pricing | Talk to sales |
| Technical evaluator | Does it fit our stack? | Integrations | Read documentation |
| Security and procurement | How is data handled, and by whom? | Security and trust | Download the security pack |
| Day-to-day user | What changes in my week? | Capability pages | Start a trial or tour |
Source: illustrative, not a client project
Grouping the pages. Pages on paper still need groups and labels. NN/g's 2024 card sorting guide recommends an open sort, where participants make their own groups, for generating a structure: at least 15 participants for a qualitative sort, 30 to 50 for a quantitative one, and 30 to 50 cards. For validating categories it advises tree testing over a closed sort. Fifteen buyers in a week is ambitious, so a sort with sales and customer success is a usable proxy; label it as one, and tree-test with real buyers before navigation is designed.
One job, one primary action, one owner per page
Section five turns every row of the page map into a short brief with these fields.
URL and parent. Where the page sits, which decides its breadcrumb and internal links.
Audience. One main audience from the matrix, at most one secondary.
Job. One sentence with a verb: "convince a security reviewer that data handling meets their standard". If two pages have the same job, merge them.
Question answered. Copied from the matrix, in the buyer's words.
Primary action. The one thing this audience can do next at this stage. With 67% of buyers favouring a rep-free path, early-stage pages usually need a self-serve action; Decision 2 in B2B Website Design in 2026 covers how to choose it.
Proof needed. The named customer, standard or number the page must carry, and whether it exists yet.
Target query and owner. The query the page should earn, if any, and who writes and approves it by when.
The page map that survives is usually shorter than the first draft, which is the point.
The content inventory decides the schedule
NN/g's Content Inventory and Auditing 101 (Anna Kaley, 2020) defines the inventory as a list of every piece of content, with its title, URL, owner, topic, format, date and metadata, and the audit as the judgement on quality: high, mediocre or low, then keep, update or remove. Its example timebox for an initial effort is six weeks.
A one-week strategy cannot audit a large site in full, so inventory everything and audit only what the new page map needs. We would add five columns: Search Console clicks and top query, GA4 entrances and key events, the destination page, the decision (keep, update, merge, remove or write new) and a due date. The destination column becomes the redirect map the Playbook's launch week relies on.
The sign-off document, and how we would run Strategy Week
The strategy is finished when each section has a named approver's signature and changes after sign-off go into a change log, costed in days. A page map approved by a committee in a meeting is not signed; it is postponed.
Our four-week marketing-site process starts with a Strategy Week. The day-by-day below is how we would run this method inside it: a recommendation, not a published schedule.
| Day | Work | Output |
|---|---|---|
| Before | Book interviews; get Search Console and GA4 access | Calendar, data access |
| 1 | Prep decisions; start the inventory crawl | Decisions note |
| 2 | Interviews, win/loss calls, sales session | Questions per audience |
| 3 | Positioning session; demand map | Sections 1 to 3 |
| 4 | Matrix, page map, proxy card sort, briefs | Sections 4 and 5 |
| 5 | Inventory decisions, measurement, review | Signed document |
Source: our recommendation; the four-week process with a Strategy Week first is Octo Agency's
How to use this
Write the seven sections in order, and open no design file until section five exists. Feed positioning from buyers and lost deals, not the investor deck. Cut every page without a question behind it, give the rest one job and one action, then set the launch date from the count of content still to write. Our marketing websites service page explains how we scope the four weeks that follow.
