For Shopify brands running Meta ads

Find your next Meta ad test from storefront evidence.

autoprune watches the signals that changed this week, recommends the next test, and keeps every launch tied to a reason.

The workflow spine

The page should prove one clean operating loop.

1

Storefront signals

Pull claims, proof points, product context, objections, and customer language into one starting point.

2

Ranked angles

Turn the strongest signals into a short list of testable Meta angles with a clear reason.

3

Approved tests

Review video-forward cards, approve what deserves a launch, and reject weak or off-brand ideas.

4

Results memory

Log what happened after manual launch so the next batch starts from evidence instead of recall.

5

Next recommendation

Carry the learning forward into the next weekly decision: scale, refresh, prune, or test a new angle.

Why this exists

The unit of value is not one ad. It is the weekly test loop.

Creative ideas live in notes, Slack, Figma, and someone's memory.

Results live in Meta without the reason a test existed.

Product context changes faster than the testing plan.

The next batch starts from vibes instead of last week's learning.

Reactive weekly plan

The plan changes when the signal changes.

autoprune is the upstream decision layer before creative analytics: it watches what shifted, explains the recommendation, and keeps weak angles out of the batch.

Signal found

The product, claim, objection, or customer language that makes the test worth considering.

Angle ranked

A decision-ready test brief with format, placement, hook, rationale, proof source, and review state.

Learning captured

Manual launch marker, performance fields when available, and the qualitative note that should inform next week.

Week 22

This week's test batch

Built from what changed since the last run.
  1. Run

    Barrier routine proof

    Reviews and homepage copy both moved toward fewer steps.

    Brief one creator video around a simpler routine.
  2. Rewrite

    Ceramide proof check

    Ingredient proof is stronger, but the wording is too clinical.

    Soften the claim before it enters the batch.
  3. Hold

    Discount-first opener

    Last run and brand notes both point away from price-first creative.

    Keep it out until the offer needs a dedicated test.

Approve, edit, or skip before anything launches.

Proof format

Every future case study should read like a decision record.

autoprune does not have mature customer outcome proof yet. The page still needs the container for it: brand, signal, test decision, and learning, without inventing lift numbers before pilots earn them.

Brand

The product or offer being tested

Signal

The storefront, customer, or ad-learning input

Decision

Approve, edit, reject, launch, prune, or scale

Learning

What the next recommendation should remember

Best fit / not fit

Built for founder-operators with a real weekly testing habit.

Best fit

  • Shopify or ecommerce brand running Meta as a real weekly channel
  • Founder or operator close to product, creative, and growth decisions
  • One to three products or offers that need clearer test direction
  • Wants a repeatable testing loop before adding more ad-ops complexity

Not a fit

  • Pre-launch brand with no active testing habit
  • Looking for guaranteed performance lift
  • Only wants finished AI ad images
  • Wants full Meta launch automation today

Early product

Built close to ecommerce workflows, with human approval in the loop.

autoprune is early. It should not promise guaranteed ROAS, replace your media buyer, or launch spend without you.

The product is shaped around a practical founder workflow: turn product and brand context into a cleaner weekly testing system, then keep the learning visible.

Start the loop

Run your first weekly test batch.

No Meta login required for the first run. Start with brand and product context, then review the cards before anything launches.

Prefer a founder conversation first? Leave an email and Catalina will follow up with a quick fit check.