Change Sets

Turn recommendations into controlled website changes.

A recommendation is a sentence. A Change Set is a record: what exists now, what will exist instead, why, what it is worth, what could go wrong, whether it passed validation, who approved it, when it shipped, how to reverse it, and what it did. It is the central object in OMNISTRIQ.

Anatomy

Everything a reviewer needs, in one place.

The example on this page uses illustrative demo data. The structure is the product.

CHANGE SET · cs_4e1b7aProduct example

Opportunity

Category page underperforming despite strong commercial intent.

Ready for Review

Evidence

18,420Impressions
8.4Average position
1.1%CTR
StrongHistoric conversion signal

Proposed change · title tag

/collections/mens-running-shoes · <title>
removed: Men's Running Shoes | Example
+added: Men's Running Shoes for Road & Trail | Example

Reason

Queries containing “road” and “trail” account for most of the page’s impressions but are absent from the title. The page already ranks on page one; a title that names both intents is designed to lift click-through without changing what the page sells.

Risk LowScope 1 pageReversible YesConfidence High

Pre-flight

  • Passed: Markup valid — 1 of 1 fields
  • Passed: No conflicting manual edit since analysis
  • Passed: Rollback point captured — rb_9c02d
  1. Draft
  2. Validating
  3. Ready for Review
  4. Approved
  5. Publishing
  6. Published
  7. Measuring
  8. Verdict
Illustrative demo data
Current state
The exact value as it exists on the live site today — the title, the meta description, the heading, the schema block, the internal link — captured at analysis time, not typed from memory.
Proposed state
The exact replacement value. Not “improve the title”, but the title. Rendered as a diff so the reviewer sees what is removed and what is added.
Evidence
The queries, impressions, positions, click-through, competitor pages and commercial signals the opportunity was derived from. Each figure links to its source.
Reason
One plain paragraph explaining the mechanism: why this change is expected to move this metric on this page.
Commercial relevance
Where the data exists, an estimate of what the page or query set is worth — conversion history, product value, orders, revenue — so the queue is ordered by money, not volume.
Risk
Scope (how many pages), blast radius (what else references the field), reversibility, and what a rollback restores.
Confidence
A five-segment meter with a word: Low, Medium, High. Never a percentage, because a percentage implies a precision the model does not have.
Validation
Pre-flight checks that must pass before approval is possible: markup validity, no conflicting manual edit since analysis, rollback point captured.
Approval
A person reviews the diff and approves. The system is designed so that approval is impossible until the diff has been opened.
Deployment
A two-step commit: approve, then confirm the target environment, page count and rollback token in words. Native publishing is planned for WordPress, WooCommerce and Shopify.
Rollback
The previous state is captured before publishing and restorable afterwards. The rollback token is visible before approval, not after.
Measurement
A defined window, a stated noise band, and one of three verdicts: Positive Effect, Negative Effect, No Measurable Effect. The verdict lives in the same record as the change.

States

The state machine is always visible.

A reviewer should never wonder where a change is. Each state has a distinct geometric mark and a word, so it is distinguishable in greyscale and by a screen reader — colour is never the only signal.

History is immutable. A rejected or rolled-back Change Set stays in the list with its diff intact. An approved set cannot be edited; reject it and generate a new one.

STATESProduct example
  1. DraftGenerated, not yet validated.
  2. ValidatingPre-flight checks running.
  3. Ready for ReviewChecks passed, awaiting a person.
  4. ApprovedA person approved it; queued for deployment.
  5. PublishingWriting to the live site.
  6. PublishedLive; measurement window not yet started.
  7. MeasuringInside the measurement window.
  8. VerdictPositive Effect, Negative Effect or No Measurable Effect.
Positive EffectNegative EffectNo Measurable Effect

Outcome

The verdict lives in the same record as the change.

From the moment measurement starts, the outcome block appears inside the Change Set. All three verdicts are rendered at the same weight. A change that did nothing is reported as a change that did nothing.

Positive Effectcs_4e1b7a · 28 days
CTR
1.1% → 1.9%
Average position
8.4 → 7.6
Clicks
203 → 350

Moved beyond the noise band, favourably. The change stays live.

Product example
Negative Effectcs_91d0c3 · 21 days
Average position
3.1 → 5.4
Clicks
1,240 → 890
Rollback
rb_2a77f available

Moved beyond the noise band, unfavourably. Rolling back restores the previous state.

Product example
No Measurable Effectcs_b330e8 · 28 days
Average position
6.2 → 6.1
Clicks
412 → 418
Noise band
±4% on this page set

Inside the noise band. Reported as exactly that. Roll back or keep — both are valid.

Product example

Know what to change.Ship it.Prove it worked.

Apply to become an OMNISTRIQ founding customer.

Apply for Early Access

Privacy

Cookie preferences

Analytics is off until you switch it on. Your choice is stored for 182 days and can be changed at any time from “Cookie preferences” in the footer. Full details in the cookie policy.