KB-017: Marketing Audit Scoring Logic
Purpose: Explain how the GHL Prospecting marketing audit report calculates its overall score and each section score. Each formula is checked against real reports rather than taken from the vendor's description.
Who This Is For: Anyone who reads an audit report, quotes from one, or writes about the audit (outreach, blog, video, social).
Companion doc: KB-015: GHL Prospecting Tool covers what the report contains and how the tool runs. This doc covers only how the numbers are made.
Quick Answer
- The overall score is the plain average of the sections the report shows, rounded to the nearest whole number. The report calls it "weighted". It is not.
- A hidden section is left out of the average. It is not counted as zero. Listings is hidden on SC's reports (report customization, since 2026-09-29). With the AI Search Visibility section added (seen from 2026-11-11), the overall is the average of seven sections, each worth one seventh.
- AI Search Visibility counts toward the overall like any other section. It is a readiness check: it does not ask any AI assistant anything. See AI Search Visibility.
- Hiding a section does not change the other sections. Each section score is calculated on its own.
- Only Website Performance matches the vendor's published formula to the decimal. The Local SEO Heatmaps score does not match the published formula. Treat it as directional.
Verification Status Labels
| Label | Meaning |
|---|---|
| Reproduced | The formula, applied to the inputs printed in the report, gives the printed score on every report tested |
| Stated by vendor | Taken from the vendor's grading article, and consistent with the reports, but the reports cannot test it (for example, every input scored full marks) |
| Inferred | Not published by the vendor. The simplest formula that fits the printed numbers |
| Does not reproduce | The vendor's formula gives a different number from the report |
The test reports are SC's own audit, run on 2026-09-23 (Listings shown) and 2026-09-29 (Listings hidden). See Sources.
Overall Score
Formula: the average of the visible section scores, rounded. Status: Reproduced.
| Report | Sections shown | Section scores | Average | Printed |
|---|---|---|---|---|
| 2026-09-23 | 7 (with Listings) | 100, 60, 100, 5, 37, 68, 90 | 460 / 7 = 65.7 | 66% |
| 2026-09-29 | 6 (Listings hidden) | 100, 60, 100, 37, 87, 72 | 456 / 6 = 76.0 | 76% |
| 2026-11-11 | 7 (Listings hidden, AI Search Visibility added) | 75, 60, 100, 0, 68, 71, 0 | 374 / 7 = 53.4 | 53% |
On the 2026-11-11 run, the average without the AI section would be 303 / 6 = 50.5. The 53% printed confirms the AI section is inside the average at equal weight.
The 2026-09-23 result shows the overall is rounded, not truncated. Truncating 65.7 would print 65.
The vendor's article is out of date on this. It describes "five equally weighted sections" at 20% each: Business, Reviews, SEO, Website, Listings. The report actually has eight: Business Details, Techno Stack, Google Business Profile, Listings, Reputation, Website Performance, AI Search Visibility, Local SEO Heatmaps. With Listings hidden, seven are visible and averaged. Techno Stack, Google Business Profile and AI Search Visibility are not mentioned in the grading article at all.
What hiding Listings did
| With Listings | Listings hidden | |
|---|---|---|
| Sections in the average | 7 | 6 |
| Weight of each section | 1/7, about 14.3% | 1/6, about 16.7% |
| Effect on SC's overall | 66% | 76% |
Almost the whole 10-point rise comes from removing the 5% Listings score. Averaging the 2026-09-23 scores without Listings gives 455 / 6 = 75.8. The real measurement changes between the runs, Website Performance up from 68 to 87 and Local SEO Heatmaps down from 90 to 72, net out to about zero. The four sections whose inputs did not change (Business Details, Techno Stack, Google Business Profile, Reputation) printed the same scores both times.
Consequences:
- Hiding Listings raises the overall for most prospects. Small businesses usually score very low on directory listings, so removing it lifts the average.
- Do not compare overall scores across the 2026-09-29 change or the 2026-11-11 change. A report from before 2026-09-29 averaged seven sections (or six before Listings was enabled on 2026-09-09, per KB-015); from 2026-09-29 it averaged six; from 2026-11-11 it averages seven again, with AI Search Visibility in place of Listings. Compare section by section instead.
- Each remaining section now moves the overall more. A weak section, such as Reputation for a business with no Facebook page, now costs more.
- The Listings data is still collected. Listings was hidden in the report layout only. The Prospect AI agent still runs the Listings audit, so the section can be shown again without re-running anything.
Section by Section
Section scores are truncated (rounded down) where the tests can show it. See Website Performance.
Business Details (printed as "Business Critical Information")
Status: Stated by vendor. SC scores full marks on every item, so the weights cannot be tested on SC's report.
| Item | Weight | Where it appears in the report |
|---|---|---|
| GBP claim status | 20 | Not shown on this page |
| Review reply percentage | 20, banded: 0-10% = 0, 11-49% = 10, 50-89% = 15, 90-100% = 20 | Shown ("Review Replies") |
| WordPress or LeadConnector website | 20 | Shown ("Website Hosting") |
| Text-enabled business number | 15 | Shown |
| Chat widget | 15 | Shown |
| Operational hours | 5 | Shown on the Google Business Profile page, not here |
| Business photos | 5 | Shown on the Google Business Profile page, not here |
Watch for:
- The page shows 4 checks, but the score uses 7. A business can pass all four visible checks and still score below 100.
- The hosting item rewards the vendor's own platform. Only WordPress or LeadConnector (GHL) sites earn those 20 points, so a fast, well-built Wix or Squarespace site loses a fifth of the section.
- Review replies are scored here, not under Reputation.
Techno Stack
Status: Inferred. The vendor article has no Techno Stack section.
Formula: 5 checks at 20% each: Google Ads campaigns, Meta Pixel, Google Ads conversion pixel, Google Tag Manager, Google Analytics.
Check: SC shows 3 of 5 detected on both runs, and printed 60% both times.
Watch for: detection fails on JavaScript-rendered sites and behind tag managers (KB-015). A missing mark means not detected, not not present. "Google Ads campaigns" also scores a business down for not advertising, which is a sales prompt, not a fault.
Google Business Profile
Status: Inferred. The vendor article has no GBP section.
Formula: 5 checks at 20% each: address, photos, website, operational hours, phone number.
Check: SC shows 5 of 5 and printed 100%. Neither run can test partial credit.
Reputation (printed as "Analyzing Your Reputation Game")
Status: partly reproduced.
Formula (vendor): 75% Google reviews + 25% Facebook reviews. A perfect rating is marked down when the review count is very low. The report labels this "Low Review Volume" when there are fewer than 10 reviews.
Check against SC: 8 Google reviews, all 5 stars; 0 Facebook reviews.
- Facebook caps the score. With no Facebook reviews, that 25% is zero, so the section cannot exceed 75%.
- Low volume halves the Google score. The printed 37% fits a Google score of about 50: 0.75 × 50 = 37.5, truncated to 37. The size of the low-volume penalty is inferred. The vendor does not publish it.
Watch for: a business with no Facebook page is scored zero on it, not treated as not-applicable. A small business with a perfect Google record can still land in the "Bad" band. Read the review count and rating, not the percentage.
Website Performance (with "Website Health")
Status: Reproduced on both runs.
Formula (vendor):
| Input (as printed) | Weight |
|---|---|
| Mobile page speed (the Mobile bar) | 25% |
| Desktop page speed (the Desktop bar) | 25% |
| Overall SEO Score | 30% |
| Interactivity | 10% |
| Large Content | 10% |
Checks:
| Run | Mobile | Desktop | SEO | Interactivity | Large Content | Calculated | Printed |
|---|---|---|---|---|---|---|---|
| 2026-09-23 | 23 | 77 | 100 | 48 | 91 | 5.75 + 19.25 + 30 + 4.8 + 9.1 = 68.9 | 68% |
| 2026-09-29 | 51 | 100 | 100 | 97 | 100 | 12.75 + 25 + 30 + 9.7 + 10 = 87.45 | 87% |
The 2026-09-23 run shows section scores are truncated: 68.9 printed as 68.
Watch for:
- The "Page Speed" bar under Core Web Vitals is not scored separately. It matched the Desktop bar on both runs.
- The Website Health counts ("Should Fix", "Passed Test") are not part of the formula. That page only repeats the section score.
- Speed scores move between runs of the same site, so treat them as directional.
Local SEO Heatmaps (with "Local Rank Tracker")
Status: Does not reproduce.
Formula (vendor): a 3×3 grid of 9 map points for one search term. Ranks below 3 score nothing.
| Point | Rank 1 | Rank 2 | Rank 3 |
|---|---|---|---|
| Centre point | 20 | 10 | 5 |
| Each of 8 outer points | 10 | 5 | 3 |
Checks:
| Run | Grid, read from the map | Vendor formula gives | Printed |
|---|---|---|---|
| 2026-09-23 | Centre 1. Outer: six at rank 1, two at 10+ | 20 + 60 = 80 | 90% |
| 2026-09-29 | Centre 2. Outer: four at rank 1, one at rank 2, three at 10+ | 10 + 40 + 5 = 55 | 72% |
The printed score is higher than the published formula on both runs, by different amounts. No simple adjustment fits both, such as partial credit for 10+ points or equal weight per point. The real formula is not known.
Watch for:
- Never quote the heatmap percentage.
- The grid follows one search term, chosen by the tool: "marketing automation near me" on SC's reports.
- Where the grid is centred: resolved 2026-10-09 ✅. It follows the prospect's own address. This entry previously recorded that both SC reports centred on Southbury (06488), which is also the Prospect AI agent's anchor postal code, and that whether the agent set the centre was not verified — a live risk, because a grid centred on the wrong town makes the section meaningless. Settled by reading the heatmap data for two generated prospect reports: an Illinois prospect's 3x3 grid centres at
42.2157, -88.3280, which is Crystal Lake and 779 miles from Southbury, and every competitor it lists trades in Crystal Lake or Lake in the Hills; a Thomaston, CT prospect's grid centres at41.6573, -73.0867, which is Thomaston, listing Cheshire, Northfield and Plainville businesses. The agent's anchor does not set the centre. The Southbury reading on SC's own reports was simply SC's own geography.- What this licenses, and what it does not. The ranked competitor table may be described to a prospect as who Google ranks around them — the touch-2 template now does exactly that. A wrong business match still voids the section, and that check is unchanged. Re-verify if the agent's location configuration ever changes.
AI Search Visibility
Status: overall contribution Reproduced; internal formula Does not reproduce. Source: SC's own audit of 2026-11-11 and the vendor's help article, "AI Search Visibility in Prospecting Reports" (modified 2026-10-27).
What it is: a readiness check. It scores whether AI assistants (ChatGPT, Claude, Gemini, Perplexity) can find, read and act on a business. It does not query any AI assistant: there is no mention rate, no record of what an assistant says, and no competitor list. The vendor's own note: the score "does not guarantee that an AI assistant will recommend the business".
Output: a score from 0 to 100, labelled "Ready" from 70 ("Recommended: 70 or above"), a "What's hurting it most" list, and four area scores.
| Area | What it checks | SC, 2026-11-11 |
|---|---|---|
| Local presence | Name, address and phone consistency across Google Business Profile, business directories and reviews. A business without a website can still earn credit here | 52 (directories Weak: 3 inconsistent, 28 missing; GBP Strong) |
| Crawler access | robots.txt rules for GPTBot (ChatGPT), ClaudeBot (Claude), PerplexityBot (Perplexity) and Google-Extended (Gemini) | 100 (all allowed) |
| AI readability | Share of the page that depends on JavaScript; LocalBusiness structured data, field by field (name, phone, address, hours, services, reviews); a preview image for shared links (Open Graph); an llms.txt file | 63 (JS 82% Weak; structured data 1 of 6, Partial; preview image Strong; llms.txt Strong) |
| Agent readiness | Whether an assistant could send an inquiry, book, find the phone number and start a call, from buttons, links, forms and field labels on the one page tested | 45 (inquiry and booking Not ready; phone and call Ready) |
The section score is not a plain average of the four areas. 52, 100, 63 and 45 average 65, but the report printed 71. How the areas combine is not known. Quote area findings, not the arithmetic.
Watch for:
- It is a Labs feature. It appeared on SC's reports by 2026-11-11, with general release about 2026-12-01. Whether reports from the
/auditwidget and the sub-account carry it is not yet verified. - Agent readiness tests one page. A business with a contact page elsewhere can still show "Not ready". See the Prospecting Audit Accuracy SOP.
Listings (hidden on SC reports since 2026-09-29)
Status: Inferred. The vendor article lists what the section considers (accuracy, how important each source is, overall visibility) but gives no weights.
Formula that fits: the average of the name, address and phone match rates across 31 directories.
Check: on 2026-09-23 SC showed name 6%, address 3% and phone 6%. Their average is 5, which matches the printed 5%. Those rates come from 2, 1 and 2 of 31 directories: 3 partial matches and 28 no-matches. The vendor's "source importance" weighting, with Google counting more, is not visible in these numbers.
Built-In Biases to Keep in Mind
| Bias | Section | Effect |
|---|---|---|
| Rewards the vendor's own hosting | Business Details | A non-WordPress, non-LeadConnector site loses 20 points regardless of quality |
| Facebook absence counts as zero | Reputation | Maximum 75% for a business with no Facebook page |
| Low review volume penalty | Reputation | A perfect rating on fewer than 10 reviews scores about half |
| Not advertising counts as a gap | Techno Stack | A business that chooses not to run Google Ads loses 20 points |
| Hidden inputs | Business Details | 3 of 7 scored items are not shown on the section page |
| Plain average | Overall | Showing or hiding a section moves the overall with nothing about the business changing |
Open Questions
- Heatmap formula. What does the report actually use? The next step is to collect heatmap grids and printed scores from several prospect reports and fit against them.
- Low-volume penalty. Is the Google review score halved below 10 reviews, or does it scale with the count? Test this on a prospect report with 10 to 20 reviews.
- Partial credit on Techno Stack and GBP. Are the checks really equal weight? Test on a report with a mixed result.
Heatmap grid centre.Closed 2026-10-09 — the grid centres on the business address, not the agent's anchor. Evidence in the Local SEO Heatmaps section above.- AI Search Visibility formula. How do the four area scores combine into the section score? Collect area and section scores from several prospect reports and fit against them.
- Hiding versus switching off. Does switching off an audit on the agent drop the section from the average in the same way as hiding it in the report layout? The 2026-09-09 history in KB-015 suggests it does, but nobody has checked a report from before and after that change side by side.
How to Re-verify
Do this when the vendor updates the report, the article, or the report layout:
- Run SC's own audit and save it to the Drive enablement folder (see Sources).
- Recompute the overall from the printed section scores. It should equal the rounded average of the visible sections.
- Recompute Website Performance from its five printed inputs. It should equal the weighted sum, truncated.
- Compare every other section with the previous SC run. If the inputs are unchanged, the section score should be unchanged.
- Update this doc's status labels and worked examples. Note any change in the Document History.
Sources
| Source | Location |
|---|---|
| Vendor grading article, "Understanding the Grades in Your Marketing Audit Report" (modified 2026-03-13) | sc-docs-storage/03-sales/enablement/2026-09-ref-ghl-marketing-audit-grading-article.pdf |
SC's own audit, 2026-09-29, Listings hidden (11 pages; the tool's own filename was report_2026-09-29_en-US-Symphony-Core.pdf) | sc-docs-storage/03-sales/enablement/2026-09-ref-symphony-core-own-marketing-audit-report-listings-hidden.pdf |
| SC's own audit, 2026-09-23, Listings shown (16 pages) | sc-docs-storage/03-sales/enablement/2026-09-ref-symphony-core-own-marketing-audit-report.pdf |
| SC's own audit, 2026-11-11, AI Search Visibility added (12 pages) | sc-docs-storage/_symphony-core-documents-inbox/marketing-audit/report_2026-11-11_en-US.pdf (to be filed beside the other runs) |
| Vendor help article, "AI Search Visibility in Prospecting Reports" (modified 2026-10-27) | https://help.gohighlevel.com/support/solutions/articles/155000009302-ai-search-visibility-in-prospecting-reports; PDF capture in the same inbox folder |
Related Documents
- KB-015: GHL Prospecting Tool: report contents, section names, reading the bands
- Prospecting Audit Accuracy SOP: which findings to trust before outreach
- SC GHL Prospect AI Agent Configuration Reference: which audits the agent runs
- Marketing Audit Production PRD: the SC-authored audit that replaces section percentages with qualitative status
Document History
| Version | Date | Changes | Author |
|---|---|---|---|
| 1.2 | 2026-11-12 | AI Search Visibility section added. Seven visible sections now make up the overall, each worth one seventh, reproduced on SC's 2026-11-11 run (374 / 7 = 53.4, printed 53%). Documents the section's four areas and checks, records that it does not query any AI assistant, and that its own score (71) does not equal the average of its areas (65). Extends the do-not-compare rule to the 2026-11-11 change. New open question 5 | Symphony Core Knowledge Management Team |
| 1.1 | 2026-10-09 | Open question 4 closed: the heatmap grid centres on the prospect's own business address, not on the Prospect AI agent's anchor postal code. Verified by reading the heatmap data for two generated prospect reports, one in Illinois whose grid sits 779 miles from Southbury and lists only local competitors, one in Thomaston CT. The Southbury centring seen on SC's own two reports was SC's own geography, not the agent's. This unblocks describing the ranked competitor table to a prospect, which a cold email was already doing; the wrong-business-match caveat is unchanged | Symphony Core Knowledge Management Team |
| 1.0 | 2026-09-30 | Initial version. Every formula tested against SC's 2026-09-23 and 2026-09-29 audits. Overall and Website Performance reproduced exactly; Local SEO Heatmaps does not match the vendor's formula. Records the effect of hiding Listings through report customization | Symphony Core Knowledge Management Team |