Numbers

The numbers I can share publicly.

These are the numbers I can talk about publicly. Some are rounded. Some are anonymized. When the source is private, I say so.

What this page does

It shows enough to make the claims inspectable. It does not publish private records.

AI avatar summary

The caveats are part of the point.

Some numbers are public. Some are rounded. Some are anonymized. If the source is private, I say that instead of pretending otherwise.

Here's a summary of this page by my AI Avatar

170+
agency clients served
rounded public aggregate
165+
ad accounts modeled
internal ad-data warehouse
30+
production systems
rounded public count
3.77%
forecasting MAPE
anonymized B2B case
Evidence ledger

Each claim says what kind of source it has.

I am not dumping private records onto the site. I am showing enough of the public version that you can check the claim.

The agencies have real delivery volume

170+ clients
Internal

Source: Public-safe aggregate across the agency portfolio.

Why it matters: The software was built against work the team actually had to do.

See the companies

Paid-media data is modeled daily

165+ ad accounts
Internal

Source: Internal ad-data warehouse across paving, garage door, and Heaviside-managed accounts.

Why it matters: Account decisions get better when the data is not trapped in stale dashboards.

Inspect the ML work

The software is not a demo wall

30+ systems
Public

Source: Public site inventory plus internal systems that still do work.

Why it matters: These are not screenshots. They run sales, delivery, reporting, content, data, and operations work.

Inspect the software

Forecasting work beat a real baseline

3.77% MAPE
Anonymized

Source: Anonymized B2B forecasting case study with walk-forward validation.

Why it matters: The ML work is measured, not just described. The customer stays private, but the method and result are inspectable.

Open the case study

Sales research moved out of slide decks

1-2 hours/report
Estimate

Source: Conservative estimate from replaced manual data pulls and deck assembly.

Why it matters: A buyer can get useful market signal without waiting on a sales deck. The team spends more time on fit and follow-up.

Inspect HG Market Report

Recurring SEO delivery moved into a command center

3-4 hours/client/month
Estimate

Source: Conservative estimate from replaced recurring reporting, GBP, audit, and content-prep work.

Why it matters: The useful gain is fewer handoffs, fewer stale reports, and more work ready for review.

Inspect SEO Commander

Revenue per paid-team FTE improved materially

Nearly 2x since 2023
Internal

Source: Internal P&L and payroll model. The raw table stays private.

Why it matters: This is one of the cleanest signs that the business got simpler after an overstaffed period.

Read how I work

Agent work has owners and stop rules

80+ scheduled jobs
Internal

Source: Internal Hermes and open-agents fleet.

Why it matters: AI output is cheap. The work is deciding who owns each job and when it gets paused, which is why every one has an owner and a stop rule.

Inspect Agent Commander
Source labels

The labels mean what they say.

Public

Visible on this site or linked publicly.

Internal

Backed by private records: P&L, payroll, CRM, account data, dashboards, or operating logs.

Anonymized

Real work with the customer, account, or dataset name removed.

Estimate

A conservative count from replaced manual steps.

What stays private

Some proof does not belong on a public page.

I will not publish raw P&L exports, payroll tables, CRM records, client dashboards, Slack screenshots, or account-level data I do not own.

Public readers need enough detail to understand the pattern. Private review can go deeper.