Money and reporting
Which lead sources actually make money
Ranking lead sources by revenue is the most common reporting mistake in a trades business. This use case ranks them by gross profit, evidenced line by line out of the accounting system, and publishes the revenue ranking alongside it so the owner can see the difference.
- Who it is for
- Owners spending real money across direct mail, paid search, email and referral who cannot tell which channel is actually profitable.
- Deployed for
- Two separate businesses under one owner in West Michigan: CertaPro of Grand Haven and CertaPro of Kalamazoo, and Freedom Flooring, all named per the owner's 2026-08-28 naming approval. Each carries its own set of books.
- Status
- Live deployment
The goal
Answer the question the owner is actually asking, which is not which source produces the most leads and not even which produces the most revenue. It is which source makes the most money.
Match every paying customer back to the channel that produced them, across every place a lead can live, not just the convenient one.
Publish the method, not just the answer, so the number can be challenged and reproduced.
Cost, against the human alternative
This is analysis plus a reusable reporting skill. Priced through deployment plus skill development hours at the published rate.
Anchor title: Billing and Business Manager reconciling a year of customer-level accounting data against multiple lead sources by hand. Typical fully-loaded cost is $70,000 to $85,000 a year with payroll tax, benefits, paid time off, ramp time, and turnover included. That is a market range for the title, not a quote of anyone actual salary. The goal is not to replace that person. It is to put them in their most productive role. The reconciliation moves to the agent. The person moves to collections and margin review, which is where a business manager earns several times their own cost.
The cost that matters is the misallocated marketing spend, which is the thing a revenue-ranked report causes and a profit-ranked report prevents.
The goal is not to replace humans. It is to put them in their most productive role. Nobody gets cut here. Your billing manager stops doing data entry and starts working collections. Your GM stops chasing paperwork and starts selling. The agent absorbs the repetitive half of a job so the person keeps the half that actually needs a person.
This is a typical fully-loaded annual range for that title, meaning base pay plus payroll tax, benefits, paid time off, ramp time before the person is productive, and the cost of refilling the seat when they leave. It is a market range for the role, not a quote of anyone actual salary.
Security
- Customer-level rows never appear on a published page. Aggregate figures go on a page that is deliberately excluded from search indexing, and the row-level detail lives in a separate document with a deliberately tiny permission list.
- The accounting integration is fenced to an explicit list of companies. Every other set of books is refused by name in the procedure, and the credentials for those other books are simply not present in the container, so the call cannot authenticate even if attempted.
- The fence is honest about what it is. The rule is a behavior the agent follows; the credential absence is the actual control. We document both rather than describing the rule as a technical guarantee.
- Company identifiers are never typed from memory. They are read from the credential store for the company being worked on and checked before any call.
Output
- A ranking of lead sources by gross profit, with the revenue ranking published next to it.
- An aggregate page the owner can read at a glance, plus a restricted spreadsheet holding the customer-level rows behind it.
- An as-of timestamp on every figure, because a number quoted from a live accounting system moves between pulls and a report that hides that is lying by omission.
- Open discrepancies published rather than resolved silently. Where two defensible methods disagree, both numbers appear and the reader is told which is which.
Ongoing cost to maintain
- The extraction and matching are deterministic and run as scheduled scripts at no per-run inference cost. Judgment steps, such as writing the executive summary, are the only paid model usage.
- Reruns are cheap, which is the point. A report you can only afford to run once a year is a report you cannot trust, because you can never check it.
The framework that keeps it safe and keeps it cheap
- Match all lead universes, not the convenient one. The first pass on this engagement matched only one and silently under-reported an entire channel, because most recipients of that channel never appear in a lead sheet at all.
- Never publish revenue and call it profit. A job with no recorded cost looks like a hundred percent margin job. It is not, it is missing data, and reporting it as margin tells the owner his best channel is the one with the worst data.
- Never let an all-time figure sit inside a period report.
- Two separate companies means two separate sets of books. Pull each independently and never sum them without saying you did.
- Every number on a client-facing page traces to an evidence line with a value, a source, and a date, or it does not ship.
The procedures behind this
Each of these is a written document with a plain-English section for a person, a machine-readable section for the agent, and a dated changelog.
- accounting
- quickbooks-online
- qbo-exact-match-attribution
- BI_Build
- intelligence-brief
Where every claim on this page comes from
- Method, the three-universe matching rule, and the three named mistakes are documented in the canonical Multi-Source Lead Attribution and Gross Profit Report document, version 1.1 dated 2026-08-20, status live.
- The published aggregate page and the restricted row-level sheet exist and the split is described in that document.
- The as-of timestamp discipline and the open commercial-count discrepancy, where two methods differ and both are published, are recorded in the same document.
- The accounting scope fence, the named forbidden companies, and the honest note that credential absence is the real control are documented in the Chief of Staff QuickBooks scope document dated 2026-08-28.
- DELIBERATELY OMITTED: the actual revenue, customer counts, and margin percentages from this engagement. They are client financials and no publication permission is on file.
Questions people ask about this
Why rank by gross profit instead of revenue?
Because a source can top the revenue table and still be the worst thing you spend on. Revenue says nothing about what it cost to deliver the work. Ranking by profit routinely reorders the table, and the gap between the two rankings is usually the most useful thing on the page.
What is the most common mistake in this kind of report?
Matching only one lead universe. Most owners have three: the service CRM, the marketing platform, and the lead spreadsheets. Channels like direct mail barely appear in the spreadsheets at all, so matching only spreadsheets deletes an entire channel from the ranking and makes it look worthless.
What if two valid methods give different answers?
Publish both and say which is which. On this engagement a strict method and a looser method disagreed on a subset of commercial customers. Neither was quietly chosen. The reader is entitled to know which number they are looking at and what assumption produced it.
Do customer names end up on a web page?
No. Aggregate totals go on a page that is excluded from search indexing. Customer-level rows live in a separate restricted document with a permission list you can count on one hand.
Is there a version of this in your business?
The discovery call is a working conversation, not a demo. Bring the process that annoys you most.