Most positioning is written in a conference room. The leadership team agrees on three pillars, a tagline and a category name, and the website gets rewritten around them. Six months later the pipeline looks the same, because the buyer never used those words to describe their problem in the first place.
Good positioning is a research output, not a workshop output. This post covers how to run customer and ICP research that actually changes your positioning, and how AI can take on the legwork without replacing the part where you listen.
What positioning research has to answer
April Dunford's framing is a useful checklist. Positioning needs clear answers to five questions:
- Competitive alternatives. What would the buyer do if you did not exist? Often it is an agency, a spreadsheet or nothing, not a named competitor.
- Unique attributes. What do you have that those alternatives lack?
- Value. What do those attributes make possible for the buyer?
- Best-fit customers. Who cares most about that value?
- Market category. What frame makes the value obvious?
You cannot answer the first and fourth from inside the building. Those two need customers.
Where the evidence already lives
Before you schedule a single interview, collect what customers have already told you:
- Sales call notes and recordings. Listen for the words buyers use for the problem, and for the alternatives they mention.
- Closed-lost reasons. These are often more honest than closed-won reasons.
- Support tickets and onboarding questions. They show which expectations your marketing set that the product did not meet.
- Search queries. Google Search Console shows the phrases that already bring people to your site. That is the buyer's vocabulary, unedited.
- AI answers. Ask ChatGPT, Gemini and Perplexity your category question and see who they recommend and how they describe the category. For some buyers, that answer is where they first meet you.
- Public conversation. Reviews of your competitors and threads in communities where your buyers talk.
What AI does well in this work
The slow part of customer research has always been the reading. Pulling themes out of dozens of call notes, tagging objections and grouping phrasing takes days. This is where AI earns its place:
- Synthesis at volume. Clustering notes, tickets and reviews into recurring problems and phrases.
- Vocabulary mining. Extracting the exact words buyers use, so your copy can use them back.
- Competitive scanning. Summarising how competitors position themselves and where their messaging overlaps with yours.
- Draft hypotheses. Turning themes into testable positioning statements for a person to challenge.
In Spryxa, source material matters. You can point the crew at Google Drive so it reads your existing decks, research and notes rather than inventing context. The Brand Crew needs a brand voice reference to work at all, and says so in its limitations.
What AI cannot do for you
AI summarises what is in front of it. It cannot run a good interview, notice the pause before a buyer answers a pricing question, or tell you which customer segment you should choose to serve. Three things stay with a person:
- The interviews. Talk to recent buyers, recent losses and one or two customers who churned. Ask about the moment they decided to look for a solution, what they tried first, and what almost stopped them.
- The segment choice. Research will show you several possible ICPs. Picking one to lead with is a strategy call with revenue consequences.
- The claims. Any positioning line that makes a claim about results has to be checked against what you can prove. In Spryxa, a person reviews claims in every draft before it goes live.
From research to an ICP you can use
An ICP that says "mid-market B2B companies" is not an ICP. A usable one describes a situation, not just a firmographic:
- Trigger: what just happened that makes them look now. A new VP hire, a missed pipeline quarter, an agency contract up for renewal.
- Current alternative: what they do today and what it costs them.
- Decision owner: who signs, and who has to be convinced.
- Deal-breakers: the requirements that end the conversation if you miss them.
Write two or three of these and rank them. The Outreach Crew builds prospect shortlists against your ICP, so a sharper ICP means a shortlist you will actually approve.
From ICP to the page
Research only pays off when it changes something a buyer sees. The first place to apply it is usually the homepage hero and the pricing page. Rewrite them in the buyer's words, name the alternative you replace, and make the next step fit the buyer's stage.
Then measure. The Content Crew drafts the new copy in your voice, the Website Crew stages the change for your review, and the Measurement Crew compares the page against its own before-state once it ships.
Test the positioning before you rebuild the site
You do not need to rewrite every page to find out whether a new position lands. Cheaper tests exist. Try the new framing in outbound first lines and see whether replies change in tone. Run it as ad copy against the current message with a small, approved budget. Use it on sales calls and listen for whether buyers repeat it back. If the new framing gets a clearer reaction in two of those three places, it has earned a place on the homepage.
Keep it running
Positioning drifts as the market moves. Rerun the light version of this research every quarter: new call notes, new search queries, a fresh check of AI answers for your category. It takes hours, not weeks, once the habit is in place.
Check how your site describes you today. The audit shows how your pages read, what you rank for, and whether AI assistants cite you. Run your free Spryxa audit.