The approach
Research first. Rhetoric never.
Built on one assumption, which we have tested the hard way: the next mistake is already sitting in the work, and finding it before you do is the job.
01
Find where the money actually leaks
Not where it is assumed to leak. We start from the transaction — what was advertised, what was paid, what was added on afterwards — and work backwards to the decisions that produced the gap.
02
Verify every figure at its source
A decree number, an article, a published fee schedule, a transacted price. Secondary summaries are a starting point, never a citation. Anything we cannot trace is cut, however useful it would have been.
03
Mark secured against available
Every line is classified before it is written up. A rule-backed position and a negotiable one are different arguments requiring different tone, sequencing and fallback. Conflating them is how strong cases get lost.
04
Build the instrument, not the summary
The output is a document that can be sent, a calculator that can be run, a script that can be read aloud. If it still needs to be turned into something before it is useful, it is not finished.
05
Check the arithmetic programmatically
Every calculator and worked example is run against generated cases before release. Plausible-looking numbers are the ones that survive review, which is exactly why they need testing.
06
Publish it in full
Complete sample documents, on the page, before anyone pays. It is a slower way to sell and a much faster way to be believed.
Verification
Four mistakes we made, and found, and are telling you about.
A process that never reports a failure is not a process, it is a slogan. Every one of these read perfectly well on the page. Every one was caught only when somebody actually ran the numbers.
Finding 01
A document dated before the event it analysed
A sample verdict carried a date earlier than the notice it was assessing. Internally consistent, entirely wrong, and invisible to anyone reading for sense rather than sequence.
Finding 02
A calculator counting the same amount twice
An over-cap amount was being added on top of a total that already contained it. The output looked generous, which is precisely the direction an error must never run.
Finding 03
An implausible multiple from stacked savings
Overlapping savings were being summed as if independent, producing a return no reader would have believed. Caught by testing against generated cases, not by proofreading.
Finding 04
An unclosed tag that would have collapsed a layout
A single missing element in published markup. Trivial to fix, embarrassing to ship, and only found because the file was validated rather than glanced at.
The working assumption: the next one is already in there. Everything is checked again before it goes out, and corrections are dated when they are made.
Where this leads
The standards that fall out of the method.
Publishing documents rather than testimonials, guaranteeing a finding rather than an outcome, drafting rather than corresponding — none of it is positioning. It is what the method leaves you able to say honestly.