How we work

How we improve this site, one measured change at a time

Every Monday Gina reads the numbers, checks the changes already running, and stages at most one new one. A founder approves what goes live.

What you can check before you decide

No invented numbers or borrowed testimonials. These are the parts you can verify in a demo.

  • Nothing goes out without your yesEvery drafted reply, proposal, post or payment waits in your approval inbox until you say send.
  • Runs on your Claude accountGina works through your Claude account, so her work draws on a plan you already pay for. Today that means a Claude Pro or Max plan; more AI accounts are coming.
  • Works with the tools you already useEmail, calendar and documents connect today; your CRM connects on request. There's no new system to move into.
  • See the real product firstIn a demo we run the real Gina console on your week.

The weekly loop

The weekly improvement loop Gina reads search and analytics data, reads running experiments, decides or waits at the sample floor, picks at most one new page, writes the experiment record, passes a validator, stages the change, a founder approves, and the page goes live and is read again next Monday. 1. Read search + site data 2. Read running experiments 3. Enough data yet? 4. Decide: win, loss, or wait Guardrail hit? Revert it 5. Pick one page to change 6. Write it down first 7. Stage it for review 8. A founder says yesthen it goes live Next Mondayread it again
The left column handles what is already running. The right column adds at most one new change per week, using the same data. Both end at the next Monday read.

The rules

  1. One change per page. Copy, a call to action, an FAQ answer or links. Titles, descriptions and headlines wait for a founder's approval of the exact wording.
  2. Written down before it ships. The change, the one number that decides it, the numbers that must not get worse, how much data is enough, and the rule for calling it. A page can't carry an experiment without that record.
  3. Compared with itself. Each page is read against its own earlier weeks, not against a control group, and every readout says that's weaker evidence.
  4. Decided only at the floor. Until there is enough data the answer is "inconclusive", however good it looks.
  5. Never live without a person. Gina stages and reports. A founder approves what goes live.

What is running, from the real record

This section is built from the experiment ledger the weekly loop works from, so it is the actual process, not a description of it. Last read: 2026-10-27. The loop adds a read every Monday; the page updates with it.

EXP-001 running

Contextual 'No Gina yet?' CTA after Step 1 on the Facebook group watcher guide

Page: /guides/facebook-group-watcher-meta-muse/

What we think
Readers of this guide stall at Step 1 because the watcher needs a Gina they do not have; a CTA at that point, promising setup on a call, will produce booking clicks where the generic hero and footer 'Book a call' buttons produced none.
Before
Only the generic hero and footer 'Book a call' buttons; nothing at the point in the steps where a reader without a Gina gets stuck.
After
Inline note after Step 1: 'No Gina yet? The watcher needs one to send its finds to. Book a call and we'll connect Muse to your Gina and tune the first watcher with you.' Link /?offer=fb-watcher#get-started, so a resulting CRM lead reads 'Asked about: the Facebook group watcher'. No title, meta, H1 or H2 changed.
The number that decides it
cta click. Baseline before the change: 0.
Must not get worse
engagement rate >= 0.6, organic ctr >= 0.03
How we call it
Read weekly; decide only once the page has >= 60 sessions in the read window. Win: >= 2 cta_click on this page (or any CRM lead 'Asked about: the Facebook group watcher') with both guardrails holding. Loss: 0 cta_click at the floor. Otherwise inconclusive. A guardrail breach reverts the change.

Latest read (2026-10-27): inconclusive. Day one. Far below the sample floor, so no read. Search Console hasn't reported yet.

Log

  1. Proposed. Readers of the guide stall at Step 1 because the watcher needs a Gina they don't have.
  2. Staged. Built and checked on the staging site, in a browser at desktop and phone width.
  3. Approved. A founder approved it for production.
  4. Live. On production. The clock starts.
  5. Read. 4 of 60 sessions: inconclusive. Day one. Far below the sample floor, so no read. Search Console hasn't reported yet.
The raw record
{
  "id": "EXP-001",
  "title": "Contextual 'No Gina yet?' CTA after Step 1 on the Facebook group watcher guide",
  "page": "/guides/facebook-group-watcher-meta-muse/",
  "source_file": "src/guides/facebook-group-watcher-meta-muse.md",
  "status": "running",
  "risk_tier": "low",
  "owner": "a founder",
  "hypothesis": "Readers of this guide stall at Step 1 because the watcher needs a Gina they do not have; a CTA at that point, promising setup on a call, will produce booking clicks where the generic hero and footer 'Book a call' buttons produced none.",
  "change": {
    "kind": "cta",
    "before": "Only the generic hero and footer 'Book a call' buttons; nothing at the point in the steps where a reader without a Gina gets stuck.",
    "after": "Inline note after Step 1: 'No Gina yet? The watcher needs one to send its finds to. Book a call and we'll connect Muse to your Gina and tune the first watcher with you.' Link /?offer=fb-watcher#get-started, so a resulting CRM lead reads 'Asked about: the Facebook group watcher'. No title, meta, H1 or H2 changed.",
    "shipped": "2026-10-26"
  },
  "primary_metric": "cta_click",
  "guardrails": [
    {
      "metric": "engagement_rate",
      "op": ">=",
      "value": 0.6
    },
    {
      "metric": "organic_ctr",
      "op": ">=",
      "value": 0.03
    }
  ],
  "sample_floor": {
    "metric": "sessions",
    "min": 60
  },
  "baseline": {
    "window": {
      "start": "2026-09-28",
      "end": "2026-10-25"
    },
    "values": {
      "sessions": 13,
      "cta_click": 0,
      "form_start": 0,
      "generate_lead": 0,
      "engagement_rate": 0.769,
      "organic_clicks": 5,
      "organic_ctr": 0.0455,
      "position": 5.52
    },
    "note": "GA4 sessions/engagement 2026-09-28..10-25; Search Console clicks/CTR/position 2026-09-26..10-23 (2-3 day lag). cta_click baseline undercounts: offer links were not matched by the old click selector, fixed in the same PR."
  },
  "read_window": {
    "start": "2026-10-27",
    "end": "2026-12-21"
  },
  "decision_rule": "Read weekly; decide only once the page has >= 60 sessions in the read window. Win: >= 2 cta_click on this page (or any CRM lead 'Asked about: the Facebook group watcher') with both guardrails holding. Loss: 0 cta_click at the floor. Otherwise inconclusive. A guardrail breach reverts the change.",
  "approval": {
    "state": "approved",
    "requested_from": "a founder",
    "production_approved": "a founder 2026-10-26, sha 17356cb live 03:26 UTC"
  },
  "result": {
    "label": "pending"
  },
  "sources": [
    "GA4 property runReport (sessions, engagementRate, eventCount by eventName x pagePath), pulled 2026-10-26",
    "Search Console talktogina.ai top pages, pulled 2026-10-26"
  ],
  "limitations": [
    "Staged before/after on one page, not a randomised split: seasonality and traffic mix can move the number.",
    "About 3 sessions a week: the floor may take two months; most weekly reads will be inconclusive.",
    "The tracking fix ships with the change, so part of any cta_click rise may be measurement, not behaviour; the offer=fb-watcher parameter isolates clicks on the new link."
  ],
  "log": [
    {
      "date": "2026-10-26",
      "event": "Proposed",
      "note": "Readers of the guide stall at Step 1 because the watcher needs a Gina they don't have."
    },
    {
      "date": "2026-10-26",
      "event": "Staged",
      "note": "Built and checked on the staging site, in a browser at desktop and phone width."
    },
    {
      "date": "2026-10-26",
      "event": "Approved",
      "note": "A founder approved it for production."
    },
    {
      "date": "2026-10-26",
      "event": "Live",
      "note": "On production. The clock starts."
    }
  ],
  "reads": [
    {
      "date": "2026-10-27",
      "window": {
        "start": "2026-10-26",
        "end": "2026-10-26"
      },
      "sample": "4 of 60 sessions",
      "observed": {
        "sessions": 4,
        "engagement_rate": 0.25,
        "cta_click": 0,
        "form_start": 0,
        "generate_lead": 0,
        "app_link_click": 0
      },
      "label": "inconclusive",
      "note": "Day one. Far below the sample floor, so no read. Search Console hasn't reported yet."
    }
  ]
}

EXP-002 running

Services title/meta answers the 'call gina' query

Page: /services/

What we think
Searchers typing 'call gina' see /services/ at position 3.9 but don't click because the title starts with 'Services:'; leading with 'Call Gina' and saying what Gina does will raise organic CTR on this page.
Before
Title 'Services: more customers and every call answered | Talk To Gina'; meta 'What Gina runs for your business: get found, win and keep leads, convert more inquiries, measure what works, and every call answered.'
After
Title 'Call Gina: every call answered, follow-ups drafted | Talk To Gina'; meta 'Gina answers your business calls, drafts the follow-up and keeps your CRM current. Nothing goes out without your yes. See what Gina runs for you.' H1/H2 unchanged.
The number that decides it
organic ctr. Baseline before the change: 0.0087.
Must not get worse
position <= 8, engagement rate >= 0.5
How we call it
Read weekly once the page has >= 150 impressions in the read window. Win: organic CTR >= 2.5% with guardrails holding. Loss: CTR <= 0.9% (no better than baseline). Otherwise inconclusive. Google may rewrite titles; check the live snippet before reading.

Latest read (2026-10-27): inconclusive. Search Console lags 2 to 3 days, so there is no search data for this page yet.

Log

  1. Proposed. 'call gina' shows this page at position 3.9 with no clicks; the title starts with 'Services:'.
  2. Wording approved. A founder approved the exact title and description.
  3. Live. On production. The clock starts.
  4. Read. 0 of 150 search impressions: inconclusive. Search Console lags 2 to 3 days, so there is no search data for this page yet.
The raw record
{
  "id": "EXP-002",
  "title": "Services title/meta answers the 'call gina' query",
  "page": "/services/",
  "source_file": "src/services/index.njk",
  "status": "running",
  "risk_tier": "medium",
  "owner": "a founder",
  "hypothesis": "Searchers typing 'call gina' see /services/ at position 3.9 but don't click because the title starts with 'Services:'; leading with 'Call Gina' and saying what Gina does will raise organic CTR on this page.",
  "change": {
    "kind": "title_meta",
    "before": "Title 'Services: more customers and every call answered | Talk To Gina'; meta 'What Gina runs for your business: get found, win and keep leads, convert more inquiries, measure what works, and every call answered.'",
    "after": "Title 'Call Gina: every call answered, follow-ups drafted | Talk To Gina'; meta 'Gina answers your business calls, drafts the follow-up and keeps your CRM current. Nothing goes out without your yes. See what Gina runs for you.' H1/H2 unchanged.",
    "shipped": "2026-10-26"
  },
  "primary_metric": "organic_ctr",
  "guardrails": [
    {
      "metric": "position",
      "op": "<=",
      "value": 8
    },
    {
      "metric": "engagement_rate",
      "op": ">=",
      "value": 0.5
    }
  ],
  "sample_floor": {
    "metric": "organic_impressions",
    "min": 150
  },
  "baseline": {
    "window": {
      "start": "2026-09-26",
      "end": "2026-10-23"
    },
    "values": {
      "organic_impressions": 115,
      "organic_clicks": 1,
      "organic_ctr": 0.0087,
      "position": 5.55,
      "sessions": 19,
      "engagement_rate": 0.684
    },
    "note": "'call gina' alone on this page: 49 impressions, 0 clicks, position 3.9."
  },
  "read_window": {
    "start": "2026-10-27",
    "end": "2026-12-21"
  },
  "decision_rule": "Read weekly once the page has >= 150 impressions in the read window. Win: organic CTR >= 2.5% with guardrails holding. Loss: CTR <= 0.9% (no better than baseline). Otherwise inconclusive. Google may rewrite titles; check the live snippet before reading.",
  "approval": {
    "state": "approved",
    "requested_from": "a founder",
    "wording_approved_by": "a founder",
    "production_approved": "a founder 2026-10-26, sha 17356cb live 03:26 UTC"
  },
  "result": {
    "label": "pending"
  },
  "sources": [
    "Search Console talktogina.ai query x page and page totals 2026-09-26..2026-10-23, pulled 2026-10-26",
    "GA4 property sessions/engagement by pagePath 2026-09-28..2026-10-25"
  ],
  "limitations": [
    "Staged before/after, not randomised; Search Console lags 2-3 days and Google may show its own rewritten title.",
    "Read window begins at staging; the clock that matters starts at the prod ship date (change.shipped)."
  ],
  "log": [
    {
      "date": "2026-10-26",
      "event": "Proposed",
      "note": "'call gina' shows this page at position 3.9 with no clicks; the title starts with 'Services:'."
    },
    {
      "date": "2026-10-26",
      "event": "Wording approved",
      "note": "A founder approved the exact title and description."
    },
    {
      "date": "2026-10-26",
      "event": "Live",
      "note": "On production. The clock starts."
    }
  ],
  "reads": [
    {
      "date": "2026-10-27",
      "window": {
        "start": "2026-10-26",
        "end": "2026-10-26"
      },
      "sample": "0 of 150 search impressions",
      "observed": {
        "sessions": 0,
        "cta_click": 0
      },
      "label": "inconclusive",
      "note": "Search Console lags 2 to 3 days, so there is no search data for this page yet."
    }
  ]
}

EXP-003 running

Homepage title leads with 'Gina AI'

Page: /

What we think
'gina ai' is the most common brand query; a homepage title that starts with 'Gina AI' matches it and will raise homepage organic CTR without losing the signed-off category.
Before
Title 'Talk To Gina | Your AI Chief of Staff'.
After
Title 'Gina AI: your AI chief of staff | Talk To Gina'. Meta, H1 and eyebrow unchanged.
The number that decides it
organic ctr. Baseline before the change: 0.0204.
Must not get worse
position <= 8, engagement rate >= 0.35
How we call it
Read weekly once the page has >= 300 impressions in the read window. Win: organic CTR >= 3% with guardrails holding. Loss: CTR < 2%. Otherwise inconclusive.

Latest read (2026-10-27): inconclusive. Day one. Search Console hasn't reported yet; site visits are shown for context only.

Log

  1. Proposed. 'gina ai' is the most common brand search; the homepage title doesn't lead with it.
  2. Wording approved. A founder approved the exact title.
  3. Live. On production. The clock starts.
  4. Read. 0 of 300 search impressions: inconclusive. Day one. Search Console hasn't reported yet; site visits are shown for context only.
The raw record
{
  "id": "EXP-003",
  "title": "Homepage title leads with 'Gina AI'",
  "page": "/",
  "source_file": "src/index.njk",
  "status": "running",
  "risk_tier": "medium",
  "owner": "a founder",
  "hypothesis": "'gina ai' is the most common brand query; a homepage title that starts with 'Gina AI' matches it and will raise homepage organic CTR without losing the signed-off category.",
  "change": {
    "kind": "title_meta",
    "before": "Title 'Talk To Gina | Your AI Chief of Staff'.",
    "after": "Title 'Gina AI: your AI chief of staff | Talk To Gina'. Meta, H1 and eyebrow unchanged.",
    "shipped": "2026-10-26"
  },
  "primary_metric": "organic_ctr",
  "guardrails": [
    {
      "metric": "position",
      "op": "<=",
      "value": 8
    },
    {
      "metric": "engagement_rate",
      "op": ">=",
      "value": 0.35
    }
  ],
  "sample_floor": {
    "metric": "organic_impressions",
    "min": 300
  },
  "baseline": {
    "window": {
      "start": "2026-09-26",
      "end": "2026-10-23"
    },
    "values": {
      "organic_impressions": 343,
      "organic_clicks": 7,
      "organic_ctr": 0.0204,
      "position": 5.76,
      "sessions": 318,
      "engagement_rate": 0.431
    },
    "note": "Homepage queries: 'call gina' 54 imp, 'gina ai' 23 imp, 'gina to' 6, 'text gina' 4."
  },
  "read_window": {
    "start": "2026-10-27",
    "end": "2026-12-21"
  },
  "decision_rule": "Read weekly once the page has >= 300 impressions in the read window. Win: organic CTR >= 3% with guardrails holding. Loss: CTR < 2%. Otherwise inconclusive.",
  "approval": {
    "state": "approved",
    "requested_from": "a founder",
    "wording_approved_by": "a founder",
    "production_approved": "a founder 2026-10-26, sha 17356cb live 03:26 UTC"
  },
  "result": {
    "label": "pending"
  },
  "sources": [
    "Search Console talktogina.ai query x page and page totals 2026-09-26..2026-10-23, pulled 2026-10-26",
    "GA4 property sessions/engagement by pagePath 2026-09-28..2026-10-25"
  ],
  "limitations": [
    "Staged before/after, not randomised; Search Console lags 2-3 days and Google may show its own rewritten title.",
    "Read window begins at staging; the clock that matters starts at the prod ship date (change.shipped)."
  ],
  "log": [
    {
      "date": "2026-10-26",
      "event": "Proposed",
      "note": "'gina ai' is the most common brand search; the homepage title doesn't lead with it."
    },
    {
      "date": "2026-10-26",
      "event": "Wording approved",
      "note": "A founder approved the exact title."
    },
    {
      "date": "2026-10-26",
      "event": "Live",
      "note": "On production. The clock starts."
    }
  ],
  "reads": [
    {
      "date": "2026-10-27",
      "window": {
        "start": "2026-10-26",
        "end": "2026-10-26"
      },
      "sample": "0 of 300 search impressions",
      "observed": {
        "sessions": 17,
        "engagement_rate": 0.235,
        "cta_click": 0,
        "form_start": 0,
        "generate_lead": 0
      },
      "label": "inconclusive",
      "note": "Day one. Search Console hasn't reported yet; site visits are shown for context only."
    }
  ]
}

Started 26 October 2026. The guide gets a few visits a week, so expect its first real read in about two months. The two title tests read sooner, once search data arrives.

Common questions

Is this an A/B test?
No. This site doesn't have the traffic for a split test yet. Each page is compared with its own earlier weeks, which is weaker evidence, and every readout says so.
What does Gina change, and what stays with a person?
Gina can draft and stage body copy, calls to action, FAQ answers and links. Titles, descriptions and headlines change only with a founder's approval of the exact wording, and nothing reaches the live site without a founder's yes.
What happens when a change makes things worse?
Each experiment names guardrail numbers up front. If one is breached once there is enough data to read, the change is reverted.

Want Gina running this on your site?

Book a call. We'll look at your own numbers first.