How do I test different positioning angles quickly?

An angle changes one component

A positioning angle is one changed component of the position, not a fresh slogan. April Dunford defines it as how your product leads at something a defined set of customers cares about, and she separates it from messaging, a tagline, a brand story, and a vision. The order is fixed: the alternative if you did not exist, the capabilities it lacks, the value those create, the customers who care a lot, and the category that makes the value obvious.

That alternative includes the spreadsheet, the manual workaround, and no decision. In enterprise software, Dunford writes, about a quarter of deals are lost to no decision, so the status quo is an alternative. Pain, speed, and cost are one value until the alternative, the buyer, or the category changes.

SEO Software’s angle generator separates the frame, the stable promise, and the hook. Five hooks on one frame are still one angle. It scores the hook by click rate, conversion, and whether the right people continue. Draft hooks there. Do not let it choose the premise. A fill-in-the-blank statement is dangerous, Dunford writes, because it assumes one known answer per blank.

A glance, a restatement, and a live pitch are different tests

A five-second glance, a hidden-sentence restatement, and a live pitch answer three different questions. Only the last two can test a premise. Nielsen Norman Group calls a five-second test a gut reaction to visual style. Five seconds is too short to read copy. Do not warn them that the screen will vanish. The warning primes them to overanalyze or to prepare to memorize.

Use at least ten people who match the buyer, not colleagues. Let them read the premise once, hide it, and ask what you do, who it is for, and what it replaces. TMR’s lesson puts that same sample on a five-second flash. That flash is the glance above, not this restatement.

SurveyMonkey keeps the methods apart. Message testing scores comprehension, recall, stated intent, differentiation, and credibility. Copy testing scores a finished piece. Concept testing asks whether the idea is worth pursuing before the words exist. Positioning-statement testing scores the internal statement: market, frame, and buyer. It is not a homepage-line test. A live split scores behavior after launch.

Low comprehension voids the other scores. Show each person one message, because a lineup adds order effects. A side-by-side read is only an early gut check. A defendable score needs at least 100 people per message. New themes often dry up between nine and 17 interviews, so fewer people is a qualitative read. Versions that are too alike make people invent a preference, so change little besides the premise.

Write three to five premises, each with one job

Write three to five premises before anyone sees a page, each changing one component. If the proof or the ask changes with the headline, you cannot read the premise. Hold the offer, the proof, and the ask. When the premise is the buyer, only the list changes. Otherwise, freeze the audience too.

Premise	A yes	A no
Alternative, including no decision	They name what you replace	They cite a rival you never meet
Value	They repeat the outcome in their own work	They understand, and it is not a priority
Buyer	The right role takes the step	A different role would, and this one will not
Category	They place you without a lecture	They ask what the word means

Outbound Wiki uses three tests: pain for the problem, a solution frame for the category, and a founder story for whether you earn the talk. Hold the audience, the ask, and the subject line. Once one earns meetings you can explain, the first outbound month starts at ten or more segment combinations, not a later project. Some markets book within a few weeks, others need 20 or 25 variants, and both count only if you can explain the meetings. Drop rivals the buyer never meets. When claims sound alike, test the alternative, as in positioning in a crowded market.

Log each variant as audience, premise, trigger, exact wording, reply, and meeting. Keep the trigger separate from the premise. Review which angle drew replies from the intended audience, which produced meetings, and what people repeated, questioned, or ignored. A fit passes when the buyer recognizes the situation, understands the value without a translation, and you can state what is distinct.

Use the channel that can falsify the bet

Choose the instrument that can say no with buyers you can reach.

Instrument	Smallest honest read	It must not decide
Glance at a hero	A few matched strangers, unwarned	Whether the argument landed
Restatement	At least ten matched people	Whether they will buy
Sales conversation	Three qualified buyers after a prepared pitch	A clumsy telling, or a market-wide score
Outbound, one premise per batch	A pattern in the audience you froze	A raw reply rate
Page or ad pair	Comparable traffic on both	The company name
Monadic survey	At least 100 per message, for a score	A decision due this week

Dunford’s episode says a homepage test is noisy because it also tests the messaging, the page design, the call to action, and who arrived. One or two prepared people pitch it on a normal first call with pre-qualified buyers. If the telling is clumsy, rewrite the pitch before you drop the premise. Her show notes say to talk to best-fit prospects, then write the note before the page changes. TMR says the same reaction from three qualified buyers is enough to act. Rob Kaminski says not to A/B test a position: execute for 6 to 18 months and judge it over 3 to 7 years. One scheduling contrast he cites has held for more than 11 years. That keeps a survivor. It does not choose the first sentence. On that thread, judge calls by confusion, the wrong comparison, and a flat “so what”, and do not swap the line every quarter.

Do not kill a premise on reply rate, and do not scale the send in week two. A cutoff ignores the list, the folder, and the ask. After buyers understand it, TMR says keep it for at least three months. Change it at once if they read it as conflicting with a capability you have, as when a cost frame makes quality sound like a trade-off.

1. Freeze the offer, the proof, and the ask. Freeze the audience unless the premise is the buyer.
2. Restate three to five premises with matched strangers.
3. Give each survivor its own people. One premise per person.
4. Stop when the same reaction shows up, not when a timer ends. Clicks over a few days cannot choose the position.
5. Add a page pair only with comparable traffic. Change the homepage after the note exists.

Name the failure, then move one piece

Sort each response by the kind of failure, and change only that piece.

What you hear	Failure	Next edit
They cannot say what you do	Words	Rewrite and restate. Keep the premise.
They name the wrong rival	Frame	Change the alternative or the category.
It is not a priority	Value or buyer	Change who, or which outcome.
They step forward, then stall	Downstream	Trust, the sales path, or the product.
Clicks rise, steps do not	Hook	Keep it off the page and out of the pitch.
The right buyers repeat it and take the ask	Winner	Write the note, then the page.

Benny’s guide treats “that is interesting” as a polite no. Use strangers, not friends, family, or other founders. A procurement lead and a technical lead can hear a different emphasis. The core stays one. Define the ICP before you score, and validate it before you add volume. The note lists the alternative, the value, the buyer, the category, and the claims you will not make. The homepage translates it. A campaign waits until the sentence can lose. Count the live test as in what to measure in early GTM tests.

A model can draft the set. It cannot be the sample.

AI can draft a premise set per ICP and cluster what real buyers said. It cannot be the sample, and it cannot pick the winner. Demand three to five premises that each change one component. Reject synonyms, a benefit any rival could claim, and a predicted click score. Draft and restate a separate set per beachhead. One sentence for every ICP is the line least able to lose. You still label comprehension, relevance, or a downstream stall.

Leave the sprint on the shelf in these cases

Skip the sprint with no matched buyers, with an offer that is still two products, or before product, platform, and company are chosen. Also skip a click test that would rename the company, and a contrast that is only an adjective a rival can copy. SurveyMonkey flags a misread claim in healthcare, financial services, and packaged goods. Those are examples, not the whole list. For a factual claim, put the evidence in the sentence whatever the category. Before a restatement or a survey of people in the EU, say what you will write down and why.

I would run the reads before I touch the page

I would freeze the offer, restate three to five single-component premises with matched buyers, and put each survivor through its own calls of the same kind before the homepage moves.

I am Piet Baudoin, one person in the eenmanszaak Poldermarketing. I work fully remote, in Dutch and English, freelance or a few days a week. For a freshly funded startup anywhere, this is an all-in GTM solution from the first message to the first customers. I am AI-native and strong in AI, content, automation, and workflows. I build and run the work. Google Ads and Meta Ads are newer for me. I set them up and review them. I do not scale a large paid program.

The free growth scan shows how the current screen reads. Positioning and messaging is where the premise is chosen. How I work is the shape of an engagement.