Fernbank's developer is $33,420 cheaper as a contractor — until three answers change, and then the tool won't price it at all

The moment

Fernbank Studio, the one-person brand-and-web shop in Athens, Georgia, closed the credit union rebrand — the job priced out at $18,500, with 180 of the owner's own hours and 40 hours of development subcontracted out, at $3,800, to a developer named Jordan. It is the second week of October. The credit union has referred two more clients, both wanting the same kind of build: Fernbank designs it, Jordan builds it.

That is more development work landing on one person's plate than has ever come through Fernbank before, and the owner is doing the obvious next-step math: if this keeps up, should Jordan be brought on as staff instead of booked project by project? The instinct is to treat that as a scheduling question — can Jordan clear the time — when it is actually a classification question first, and the two do not get asked in the order most owners ask them.

What they did instead

The usual move is a contractor agreement pulled from a template site, signed once, and never looked at again while the actual working relationship quietly changes shape underneath it. The word on the agreement is still "independent contractor." Whether Jordan still functions like one six months later is a different question, and the paperwork does not track it.

Nobody sits down and asks the classification question again until something forces it — a state audit, a disgruntled worker, an accountant flagging it in passing. By then the arrangement being examined is not the one that was signed, and the exposure it created runs backward through every month it was true, not forward from the day someone noticed. Owners underprice this risk for the same reason they underprice most invisible costs: nothing about a Tuesday spent working this way feels different from a Tuesday spent working lawfully, right up until it is examined.

The questions a CFO asks first

1. Who actually decides when and how the work gets done — not what the contract says, what happens in practice?
2. Whose tools, equipment and software is the work done with?
3. Is this person paid for their time, or for a finished deliverable?
4. Is the work the thing your business sells, or something adjacent to it that you buy in?
5. Does the engagement have an end date, or does it just continue?
6. Do they have other clients right now, this week, not in theory?

Money is deliberately last, and in the first pass it is not asked at all. Classification is a facts question. A tool — or an owner — that opens with a price has already treated it like a shopping decision, and that is the mistake this ordering exists to prevent.

The arithmetic, shown

Same studio, same developer, two versions of the arrangement, run through the same eight questions. Georgia has not been verified as one of the states with its own stricter test, so both runs sit under the federal common-law factors — which the IRS is explicit cannot be scored, only weighed.

Scenario one — the arrangement as it runs today.

All seven factors point the same way. The tool's headline for this run: A clean contractor pattern on every factor — though the federal test still yields no formal determination, because no federal test does. With a real choice established, the cost comparison renders: Fernbank has paid Jordan roughly $27,000 a year across projects like this one. Brought on staff at a $52,000 base, 8.5% employer payroll tax and $4,000 of benefits, the same work would cost $60,420 a year loaded. The gap is $33,420, the saving from keeping this project-based rather than converting it to payroll, on the facts as they stand today.

Scenario two — the tighter version being considered.

To guarantee Jordan's time through the referral wave, the owner is weighing three changes: pay by the hour instead of by the deliverable, ask Jordan to pause other client work for the stretch, and run daily check-ins on what gets built and when. Development would also become something Fernbank actively sells as part of a standard package, not an occasional subcontract.

Five of seven factors now point toward employment. The tool's headline flips: This tool will not put a price on the alternative. No $27,000, no $60,420, no saving shown — because a saving shown beside a fact pattern that leans employee prices a decision that is not lawfully available to make. The tool's own next step for this run is the same one it gives every time the facts lean this way: settle it on Form SS-8, or with an employment lawyer licensed in Georgia, before changing anything.

What changed

Fernbank kept the arrangement with Jordan as it was: paid per project, on Jordan's own schedule and equipment, with the referral work booked the same way the credit union job was. The tighter version — hourly pay, paused other clients, daily check-ins — did not happen, not because the owner decided against it on principle, but because seeing it lean employee on five of seven factors was enough to table it until there was time to ask a lawyer whether that specific shape of arrangement could be run lawfully as a contractor relationship at all.

The one concrete change: the owner started keeping a short written note each quarter of how the arrangement with Jordan actually runs — separate clients, separate tools, project-by-project pricing — so that if this is ever questioned, the evidence is dated contemporaneously rather than reconstructed from memory after the fact.

What the tool will not do

It will not rule on scenario one, either, even though every factor lined up the same way. A clean pattern under the federal test is still not a determination — only Form SS-8 or an employment attorney can make one. The tool says which way the facts point and stops there, on purpose, because going further would be inventing a scoring method the IRS itself says does not exist.

It will not verify Georgia's own state test. Several states layer a stricter rule — an ABC test like California's — on top of the federal factors, and one of those can turn a clean federal pattern into an employee anyway. This tool has only verified California's ABC test against the statute directly; Georgia is reported as unverified, not as clear.

It will not size the exposure if the classification turns out to be wrong. Misclassification reaches backward through every month the arrangement ran, in back taxes, interest and penalties, and pricing that needs the whole engagement history — this tool works from the facts of how the work happens today, not from a ledger of how it happened last year.

It will not let a saving be shown next to a fact pattern that leans employee, under any circumstance, including scenario two above. That suppression is not a bug to route around by adjusting an answer until the number reappears — it is the reason this ordering exists at all.