Seven days. Seven metrics. One verdict. That is the entire promise of the Maps Visibility System: we make your business more visible on Google Maps, we measure the result against your own baseline, and if the numbers do not move, you do not pay.
This is the definitive article for that system. It is the master SOP our team and our AI agents follow every time a business enters a trial. It covers the full pipeline: intake, access verification, baseline measurement, the trial, the automated analysis, and the verdict. At the bottom you will find the QA checklist we run before any result ships, plus a skill file your own AI agent can parse and execute.
The offer, pricing, and qualification criteria live on the Maps Visibility System page linked above. This page documents how the machine works. We publish it because we would rather show the machine than pitch it.
“Prescription before diagnosis is malpractice. If you see a marketer making recommendations not based on your data, run.” — Dennis Yu
Why the system is measurement-first
Most local SEO engagements fail at the same spot: nobody agreed up front what “working” means. The agency points at rankings. The owner points at the phone. Six months later they argue about attribution instead of results.
The Maps Visibility System removes that argument before the work starts. Every trial has a defined start date, a defined end date, a baseline computed from the business’s own history, and verdict thresholds written down before anyone touches the profile. The business either beats its own baseline or it does not.
The system at a glance
The pipeline has six stages, and every stage writes its state back to one shared tracker so any team member or agent can see exactly where a business stands.
The stack is deliberately plain. A Tally qualification form feeds a Google Sheets master tracker. Pipedream workflows verify Google Business Profile access, pull performance history from Google’s Business Profile Performance API, and store it in BigQuery. A scheduled dispatcher refreshes the data daily for every active business. When a trial ends, an AI analyst compares trial performance to baseline and drafts the report. A human reviews every verdict before it ships.
Boring plumbing is the point. Boring plumbing is what makes the guarantee enforceable.
Stage 1: Intake and qualification
Every business enters through the qualification form. The form captures the essentials: business name, website, vertical, Google Business Profile link, and contact details. Submission creates a row in the master tracker, and that row becomes the single source of truth for the engagement.
Qualification matters more than volume. The system fits established local service businesses with a real reputation to amplify, the profile described on the offer page. A brand-new business with no reviews has nothing for the system to amplify yet, and we say so rather than take the money. Those owners get a build-first plan through the free Quick Audit instead.
The vertical field is not optional. The vertical decides the trial’s primary metric, as you will see in Stage 5. Intake without a vertical is an incomplete intake.
Stage 2: Access verification
Nothing starts until we can see the Google Business Profile from the inside. When the tracker marks a business audit-complete, a Pipedream workflow confirms that our agency account holds manager access to the profile.
The verification step does one thing worth copying: it fuzzy-matches the business name from the intake form against every location our agency account can see. Owners type “Joe’s HVAC” when the profile says “Joe’s Heating & Air Conditioning LLC.” The matcher scores similarity, auto-accepts matches at 80% or better, corrects the tracker to the official profile name, and writes the profile’s location ID into the tracker. Weaker matches get a “did you mean” flag for human review instead of a guess.
From this point forward, every downstream system reads that location ID, never the name. Names are for humans. IDs are for pipelines.
Stage 3: Baseline pull
The moment access is verified, the system pulls daily performance history for the profile from the Business Profile Performance API into BigQuery. Seven metrics, one row per metric per day:
- Business impressions, desktop Maps
- Business impressions, desktop Search
- Business impressions, mobile Maps
- Business impressions, mobile Search
- Call clicks
- Website clicks
- Direction requests
Two data rules keep this honest. First, every pull ends yesterday, never today, because Google reports partial zeros for the current day and partial data poisons averages. Second, the daily refresh re-pulls a rolling seven-day window rather than just yesterday, because Google restates GBP performance data three to five days after the fact. Pull only yesterday and your recent numbers go quietly wrong.
Every write to BigQuery is an upsert keyed on location, date, and metric. Re-pulling a window never duplicates a row, which means the pipeline recovers cleanly from any outage. Design for re-runs. Outages are not hypothetical.
Stage 4: The trial
The trial has a defined start and end date recorded in the tracker. During the trial we do the visibility work: Google Business Profile optimization, category and service tuning, media and posting cadence, and the keyword-targeted improvements the audit identified. The specifics vary by business and market. The measurement window around them never does.
While the trial runs, the daily dispatcher keeps pulling fresh metrics for every active business. Nobody waits until the end to discover the data broke. A pipeline that fails during the trial fails loudly, while there is still time to fix it.
Stage 5: Automated analysis
When the trial end date lands in the tracker, the analysis fires automatically. Most agencies hand-wave this stage. We wrote it down.
The baseline. The system computes a baseline from the three periods of equal length immediately before the trial started. We measure a 7-day trial against the 21 days before it, expressed as a per-period average. Comparing one week against an average of three smooths out the lumpy weeks every local service business has.
The comparison. For each of the seven metrics, a BigQuery query produces a trial total, a baseline per-period average, and the percent difference between them.
The vertical lens. Raw percentages mean nothing until you know which metric pays the bills. The analyst applies a vertical playbook:
| Vertical type | Primary metric | Secondary | De-emphasized |
|---|---|---|---|
| Home services (HVAC, plumbing, roofing, electrical) | Call clicks | Website clicks | Direction requests (mostly vendors and drivers) |
| Destination businesses (juice bars, gyms, retail) | Direction requests and call clicks | Website clicks |
The same engine runs both playbooks. Our Pure Green Maps Visibility System deployment measures juice bar locations on direction requests and calls, while calls come first for an HVAC contractor. One pipeline, many verticals.
The verdict thresholds. Written down so nobody bends them after the fact:
- Clear win: primary metric up more than 5% against baseline.
- Visibility win: primary metric stable (within roughly -10% to 0%) while Maps impressions or secondary metrics rise more than 15%. Exposure grew. Conversion work comes next.
- Failure: primary metric down more than 10%, or everything flat. The trial did not earn continuation.
One honesty rule sits on top: small swings on small volume get called “stable,” never spun as wins or mourned as losses. A business that gets twelve calls a week can swing 15% on pure noise.
Stage 6: Verdict and report
The analysis produces a report with a fixed structure: business context, the full metrics table with summary rows for total Maps impressions and total actions, before-and-after map scans for the top keywords, an analysis section that leads with the primary metric, and an action section that states the verdict plainly.
A failure verdict ends the engagement with no invoice. The report still tells the owner what to do next: keep collecting legitimate five-star reviews, strengthen the profile with photos and accurate information, and answer every inbound call, because those are the signals the system amplifies when the business returns.
A win moves the engagement into the ongoing Maps Visibility System, where the same measurement loop keeps running month over month. A human reviews every report before delivery. The automation drafts. A person stays accountable.
Why we publish this
Every task in our operation has a definitive article like this one, and every execution produces a meta article documenting the specific run, the system Dennis Yu describes in his meta-article process. The definitive article is the recipe. The meta articles are the receipts. As executions accumulate, the lessons feed back into this SOP and it sharpens.
Publishing the SOP also means you can hand this page to your own AI agent. The skill file below is written to be parsed. For the full picture of how our engagements fit together, start with how our system works.
QA checklist
Run this before any trial result ships. Every item must pass.
- Tracker row complete: business name, vertical, GBP link, website, trial start, trial end, location ID present.
- Location ID verified against the live profile, never assumed from the business name.
- Baseline window equals exactly three times the trial length, ending the day before trial start.
- BigQuery holds a row for every day in both windows. No gaps from pipeline outages.
- Data freshness confirmed: the most recent pull covers the full trial window, with the 3-to-5-day restatement window re-pulled after trial end.
- All seven metrics present for both windows.
- Vertical playbook applied. Primary metric matches the business’s vertical.
- Verdict matches the written thresholds. No adjective inflation on low-volume swings.
- Metrics table totals recompute correctly (Maps impressions sum, total actions sum).
- Map scan screenshots attached for the top target keywords, before and after.
- A human reviewed the report before delivery.
- Tracker updated with verdict, report link, and status so the next agent sees true state.
Skill file
---
name: mvs-trial-analysis
description: Run the Maps Visibility System trial-versus-baseline analysis
for a local service business. Trigger when a trial end date is set for
a tracked business. Requires GBP daily metrics in a warehouse and a
tracker row with vertical, trial_start, trial_end, and location_id.
---
## Inputs
- location_id (GBP locations/... resource name, never the business name)
- vertical (home_services | destination)
- trial_start, trial_end (dates)
- Warehouse table: one row per location_id, date, metric, value
## Steps
1. duration = trial_end - trial_start + 1 (days)
2. baseline_end = trial_start - 1 day
baseline_start = baseline_end - (3 * duration) + 1
3. Pull 7 metrics for both windows: impressions (desktop maps, desktop
search, mobile maps, mobile search), call clicks, website clicks,
direction requests. Exclude today. Confirm the restatement window
(last 3-5 days) was re-pulled after trial_end.
4. Per metric: trial_total, baseline_avg = baseline_total / 3,
lift_pct = (trial_total - baseline_avg) / baseline_avg.
5. Apply vertical playbook:
- home_services: primary = call clicks, secondary = website clicks,
de-emphasize direction requests.
- destination: primary = direction requests + call clicks.
6. Verdict:
- CLEAR_WIN if primary lift > +5%
- VISIBILITY_WIN if primary in [-10%, 0%] AND (maps impressions
or secondary lift > +15%)
- FAILURE if primary lift < -10% or all metrics flat
- Low volume rule: lift within 5% of zero on primary counts as stable.
7. Produce report: context, metrics table (with Total Maps Impressions
and Total Actions summary rows), keyword map-scan placeholders,
analysis (lead with primary metric), action section per verdict.
8. FAILURE action text: end engagement, no charge, prescribe reviews,
profile strength, and call answering.
9. WIN action text: recommend continuation into ongoing optimization.
For a visibility win, state the conversion focus for the next period.
10. Human review gate before delivery. Update tracker: verdict,
report link, status.
## QA
Run the 12-point checklist in the parent definitive article. All items
must pass before the report ships.
Your market has a verdict waiting. Request your free Quick Audit and we will run the diagnosis before anyone prescribes a thing.