New Impact Model Score

Not every change matters. Now you know which ones do.

Your site changes hundreds of times a month. Most of it is noise. The Impact Model Score reads every tracked change, filters the clutter, and surfaces only the edits worth your attention — scored 0 to 100, the moment they happen.

4,552 changes on one site in 60 days. Six were worth a look. The score finds those six.

app.pagepulse.dev/changes SignalProof
Ranked by impact Min impact ≥ 40 · 6 of 2,817 shown
88critical
H1 & intro removed/hydroseeding · risk · removal
−22%
74high
Title rewrite + FAQ added/lawn-care · 612 clicks/mo · pos 7.1
+18%
54notable
H1 changed + 2 internal links/services · 340 clicks/mo · pos 9.4
+4%
12low
Nav links changed across 410 pagestemplate change · collapsed from 410 rows
Jun 6
2,811 recurring & fanned-out changes hidden · show everything
Scored instantlyNo performance data needed
4,552Changes tracked on one site in 60 days
410Pages a single nav edit fanned out across — one real change, hundreds of rows
6Changes that actually deserved a human’s attention
The signal-to-noise problem

Change tracking gives you everything.
Even the stuff that doesn’t matter.

One nav edit fires across 400 pages. A testimonial rotator re-flags every crawl. A typo fix on a page nobody visits looks the same as a title rewrite on your money page. Without a way to rank them, a complete change log is just a longer to-do list.

The Impact Model Score

One number, 0 to 100, for every change you make

The Impact Model Score weighs three things at once: how big an SEO lever the change is, how much the page actually matters, and whether the change is real or just noise. High scores rise to the top. Noise sinks out of sight — but is never deleted.

1

Kill the noise

Sitewide edits that fan out across hundreds of pages collapse into one event. Rotating modules and cosmetic reflows are recognized and set aside.

One nav change = one row, not 410.
2

Weigh the lever

A full title rewrite moves rankings far more than an alt-text tweak. Every change is weighted by its real SEO pull — and removals of key elements are flagged as risk.

Title > H1 > body > cosmetic.
3

Weigh the page

A perfect edit on a zero-traffic page isn’t urgent. The score factors in the page’s real clicks, conversions, and striking-distance ranking opportunity.

Changes on pages that earn, rise.
4

Score it

The three combine into a single 0–100 score with a plain-English reason. Sort your whole change history by what matters, instantly.

“Title rewrite on a 612-click page.”
Noise0–14

Hidden by default. One click to reveal.

Low15–39

Minor. Quietly logged.

Notable40–64

Worth a look. In the feed.

High65–84

Real lever. Alert-eligible.

Critical85–100

Act now — especially risk removals.

Under the hood

How a raw change becomes a score

Every diff runs the same pipeline: normalized to kill noise, weighed for signal and page value, combined into an instant score that powers the filter — then re-checked against real performance 30, 60, and 90 days later.

Raw change stream
4,552 diffs / 2 months — real data from one site
L0 · Normalize

Kill the noise first

biggest win

Collapse sitewide fan-out, suppress recurring modules, map DOM → semantic field. One nav change becomes one row, not 410.

meaningful candidates only
L1 · Change Signal

How big a lever?

element × magnitude × risk

L2 · Page Value

Does the page matter?

traffic · conversions · opportunity

L3 · Impact Potential Score

0–100, instant — no performance data needed

signal × page-value floor × noise gate

THE FILTER

One dropdown · kills the noise today

L4 · Windows + Quiet Period

Wait for a clean attribution gap

Group edits and hold until seasonality and Google updates clear, so a lift isn't miscredited.

L5 · Realized Impact + Confidence

Counterfactual vs actual, with a confidence band

Synthetic control compares the page to similar unchanged pages · aligned signals · confounder guard.

ATTRIBUTION

30 / 60 / 90-day "did it work?"

Two scores, never one
Impact Potential (Signal)
When
Instant, at detection
Inputs
The change + the page only
Answers
"Worth my attention?"
Realized + Confidence (Proof)
When
30 / 60 / 90 days later
Inputs
+ performance vs counterfactual
Answers
"Did it contribute? How sure?"
Signal vs Proof

Two lenses on the same changes. One flip of a switch.

Every change carries two scores. Signal tells you what might matter, the instant it happens. Proof tells you what actually moved the needle, once the data is in. Toggle between them anywhere you see your changes.

SIGNAL

What might matter — now.

Instant · scored the moment a change is detected

Signal is your triage view. It ranks changes by Impact Potential using the change and the page alone — no waiting for performance data. Perfect for catching a risky edit the day it ships, or spotting the high-leverage change your team just made.

  • Available immediately — zero lag
  • Surfaces risky removals even on quiet pages
  • One dropdown hides the noise for good
PROOF

What actually worked — proven.

Earned · measured against a counterfactual at 30 / 60 / 90 days

Proof re-ranks the exact same changes by measured outcome. It compares each change to what your page would have done anyway — using similar, unchanged pages as a control — so seasonality and Google updates don’t get miscredited. You get confidence, not a coincidence.

  • Confidence bands, not guesswork
  • Quiet-period rules keep attribution honest
  • Client-ready: “here’s what we changed, here’s what it did”
See it decide

Three changes. Three very different scores.

The same model that buries a 410-page nav edit will pin a form removal on a page barely anyone visits. Here’s why.

Critical · risk
88/100
Contact form removed from a 12-click page

Low traffic, but a removed conversion element is a business event. Risk changes bypass the page-value filter and get pinned — so a rogue CMS edit never slips past you.

High · opportunity
74/100
Title rewrite + FAQ on a 612-click money page

The strongest on-page lever, applied to a page that already earns and sits in striking distance. This is exactly the change you want at the top of the list.

Low · template
12/100
Nav links changed across 410 pages

One template edit, not 410 changes. It collapses into a single low-score row so it never buries the handful of edits that count.

Right where you already work

It lives on your performance chart

No new place to check. The Impact Model Score plugs straight into the Performance view you already use — change markers on your GSC and GA4 timeline are re-weighted by impact.

app.pagepulse.dev/performance
Performance · impact-weighted changes
Min impact: 40 Signal
A tall bar means impact happened — not that the crawler was busy.
Apr 1May 1Jun 1Jun 30

One dropdown cuts the clutter

Set your noise floor from the top-bar Min impact dropdown. Flip Signal to Proof to switch from “what changed” to “what worked.” Your Tags, Source, and date filters carry over untouched.

  • Impact-weighted change markers on the timeline
  • One min-impact dropdown to cut the clutter
  • Works with the GSC + GA4 data you already connect
Built to be trusted

We never claim a change “caused” a result. We earn confidence — and show our work.

Page Pulse is deliberately conservative. It under-credits by default, and requires a quiet period before attributing a lift. A tool that occasionally says “not enough evidence yet” is a tool you can put in front of a client.

FAQ

How the score works, answered

A count treats every edit equally — so a nav change fanned across 400 pages looks like 400 events, and a title rewrite on your best page looks like one. The score ranks changes by how much they can actually move performance, weighting the SEO lever, the page’s real value, and whether the change is genuine or noise. You go from “here’s everything” to “here’s what matters.”
Signal is instant. It scores a change’s potential impact from the change and the page alone — available the moment an edit is detected, no waiting. Proof is earned. It measures the change’s actual outcome against a counterfactual over 30, 60, and 90 days. Signal is for triage; Proof is for reporting.
When the same change is detected across many pages in a short window, it’s collapsed into a single template event and scored once — with a discount, because site-wide changes are harder to attribute to any one page. You still see every affected page, but your feed shows one row instead of hundreds.
No. Page value can quiet down a routine edit on a quiet page, but it can never fully silence a risk change. Removing a form, an H1, a canonical, or major content gets flagged as risk and surfaces regardless of how much traffic the page gets.
Today you filter it, not tune it: set your minimum-impact level from the top-bar dropdown to decide how much surfaces. The scoring model itself runs on conservative defaults out of the box. Deeper controls — like adjusting the quiet period or alert sensitivity — are on the roadmap for teams that want to go further.

Stop scrolling change logs. Start seeing what matters.

Track up to 50 pages on one site, free. Connect your data and Page Pulse starts scoring your changes today.

No credit card required · Free plan available · MCP included on every plan