Decide what to test next

The best test idea is rarely “the thing we can change fastest.” It is the most important uncertain customer decision you can isolate well enough to learn from. Use this playbook when you have several ideas and need to decide what deserves traffic first.

Start with evidence sources

Good hypotheses can come from:

a mismatch between ad and landing page
experiment results
Jurni page analytics
recurring customer/support objections
campaign performance differences by audience or creative
user research or customer calls
obvious purchase-path confusion
new approved offers or product positioning
strong external references that suggest a pattern worth testing

An idea does not need perfect evidence. It should have a clear reason for existing.

Step 1: write observations without proposing solutions

Examples: Observation: “Visitors reach the product section, but many do not continue into Add to Cart.” Observation: “Our top Meta ad leads with convenience while the landing page hero leads with ingredients.” Observation: “Customers repeatedly ask whether subscription can be skipped or canceled.” Avoid jumping directly to “make the button orange.” First identify the decision that seems difficult.

Step 2: map the observation to a customer decision

Ask:

What does the visitor need to understand, believe, choose, or do at this moment?

Common high-value decisions include:

“Is this page relevant to why I clicked?”
“Is this product for someone like me?”
“Do I understand what this product does?”
“Do I believe the main benefit?”
“Is this meaningfully different from my current option?”
“Do I understand the offer?”
“Do I know which variant / plan to choose?”
“Do I trust this enough to purchase?”

These are usually stronger testing territories than isolated visual preferences.

Step 3: create a one-line hypothesis

Use:

If we [change one thing], then [audience] will be more likely to [behavior] because [customer reason].

Examples:

If we show the creator's exact recommendation in the hero, creator traffic will be more likely to continue because the page will feel consistent with the content they clicked.

If we explain the subscription's approved cancellation flexibility beside the plan selector, more product-section visitors will choose a purchase option because a common commitment objection will be addressed at the decision point.

The “because” forces you to state the behavioral logic.

Step 4: score test ideas with a lightweight framework

Score each idea from 1–3 on:

Factor	1	2	3
Decision impact	minor	meaningful	central to purchase
Evidence	mostly opinion	some supporting evidence	repeated / strong evidence
Clarity	several variables	somewhat isolated	one clear variable
Reach	few visitors encounter it	many encounter it	nearly everyone in target traffic sees it
Effort	large / risky	moderate	easy to build and QA

Do not treat the total as scientific truth. The scoring forces useful tradeoffs. A high-effort test can still deserve priority when the customer decision is important enough.

Step 5: choose the smallest useful change

Suppose the observation is poor ad-to-page congruence. You could:

rewrite the hero
redesign the hero
change product imagery
add creator proof
reorder the whole page

Start with the smallest change that directly tests the hypothesis. If a hero message change answers the question, do not redesign six sections at once.

Prompt: turn a page review into ranked tests

Use the observations below to propose experiment hypotheses.
Do not edit the page.

CAMPAIGN CONTEXT
Traffic: [source]
Audience: [audience]
Primary action: [action]
Primary business metric: [metric]

OBSERVATIONS
[paste observations / analytics / customer objections]

For each test idea provide:
1. observation
2. customer decision
3. hypothesis
4. one variable to change
5. what stays constant
6. primary metric
7. expected learning if it wins
8. expected learning if it loses

Then rank the ideas by:
- customer decision impact
- strength of evidence
- isolation / interpretability
- reach
- implementation effort

Do not estimate a conversion lift percentage.

High-value testing territories

Message match

Test when the campaign promise and page entry point feel disconnected. Ideas:

exact creative angle versus broad brand value prop
creator hook versus brand headline
campaign product image versus generic hero image

Value proposition

Test when visitors may not understand why the product matters. Ideas:

outcome framing versus mechanism framing
one primary benefit versus several competing benefits
audience-specific language versus category-general copy

Proof

Test when a claim may be understood but not believed. Ideas:

proof near the claim versus later
customer proof versus founder proof for a specific objection
product evidence above product section versus below

Offer clarity

Test when the real offer exists but is poorly understood. Ideas:

savings-led framing versus convenience-led framing
GWP qualification near hero versus only near cart
subscription value explained beside the plan choice

Purchase decision

Test when the visitor reaches the product but choice is difficult. Ideas:

simplified option hierarchy
clearer one-time/subscription explanation
earlier clarification of what is included

Do not change the actual offer or product economics unless that is intentionally the experiment.

Page depth

Test when you are unsure how much education this audience needs. Ideas:

short retargeting path versus full educational path
product introduced earlier versus after education
condensed FAQ versus extensive objections section

Low-quality test patterns

Be cautious with:

button color with no visibility problem
arbitrary headline changes with no campaign logic
“make it more premium” variants with many simultaneous changes
moving sections because a competitor does it
testing a new price, offer, copy, and layout at the same time

These can still produce different numbers, but they often teach less.

What good looks like

A useful test backlog is not a list of designs. Each item should state:

the observed problem
the customer decision
the hypothesis
the intended change
the success metric
the learning you hope to get

That backlog becomes more valuable over time because new results can strengthen, weaken, or replace old assumptions.

Page Building Guides

Prompting Guides

Templates & Reusable Sections

Experimentation Guides

Decide what to test next

Decide what to test next

Start with evidence sources

Step 1: write observations without proposing solutions

Step 2: map the observation to a customer decision

Step 3: create a one-line hypothesis

Step 4: score test ideas with a lightweight framework

Step 5: choose the smallest useful change

Prompt: turn a page review into ranked tests

High-value testing territories

Message match

Value proposition

Proof

Offer clarity

Purchase decision

Page depth

Low-quality test patterns

What good looks like

Next playbooks

​Decide what to test next

​Start with evidence sources

​Step 1: write observations without proposing solutions

​Step 2: map the observation to a customer decision

​Step 3: create a one-line hypothesis

​Step 4: score test ideas with a lightweight framework

​Step 5: choose the smallest useful change

​Prompt: turn a page review into ranked tests

​High-value testing territories

​Message match

​Value proposition

​Proof

​Offer clarity

​Purchase decision

​Page depth

​Low-quality test patterns

​What good looks like

​Next playbooks

Decide what to test next

Start with evidence sources

Step 1: write observations without proposing solutions

Step 2: map the observation to a customer decision

Step 3: create a one-line hypothesis

Step 4: score test ideas with a lightweight framework

Step 5: choose the smallest useful change

Prompt: turn a page review into ranked tests

High-value testing territories

Message match

Value proposition

Proof

Offer clarity

Purchase decision

Page depth

Low-quality test patterns

What good looks like

Next playbooks