The hire is not the finish line
The first 90 days after you hire a controller decide whether you bought a closer or you bought a $140K report writer. The difference is the packet, the close calendar, and who still owns the forecast.
Most founders treat the start date as the win. Then they leave the controller to get in the books with no definition of done. Day 90 arrives. The close is still late. The CEO still has the forecast in a private spreadsheet. You did not install finance. You added a person to a mess.
I sit in this gap as a fractional COO and CFO, often the person who helped justify the hire and then has to make it take. The job of the first 90 days is not for the controller to learn your tribal knowledge. It is for tribal knowledge to become a packet, a calendar, and a named split of forecast ownership. If that does not happen, the hire will look expensive and you will be right.
A controller is not a fractional COO. The controller seat is close, controls, and the numbers. Do not dump operating cadence on a new controller and then wonder why the close slipped. The role split is in what a fractional COO does.
Days 1-30: the packet has to exist
By day 30 the company should produce one finance packet, the same way every month, without the founder rebuilding it. The packet is not every report the accounting system can print. It is the owner’s version.
Page 1 is P&L actual versus budget versus prior year, with variance in dollars and percent, down to the 15 to 25 lines that explain the month.
Page 2 is cash. Beginning cash, receipts, disbursements, ending cash, versus the 13-week forecast for that week and versus the monthly cash budget. If you do not have a 13-week view yet, that is a day-30 gap, not a year-two project.
Page 3 is the balance sheet and three working-capital numbers: DSO, DPO, and inventory or WIP days if you have them. Plus the cash floor.
Page 4 is the close notes. Material variances with cause, not adjectives. Gross margin is 34.1% versus 36.5% budget, 2.4 points short, because overtime on two jobs ran $36K. Risk is those two jobs still have three weeks of work.
Page 5 is open items. Unreconciled accounts, unmatched rec items, unpaid tax estimates, and any control failure. If page 5 is empty because people are hiding items, that is a control failure of its own.
The controller owns the assembly. The CEO owns two hours in week one to say which lines they actually use. If the CEO will not sit for those two hours, you will get a 40-page packet nobody reads. That is on the CEO, not the controller.
Do not wait for a perfect ERP. By day 30, map access: who can pay, who can book, who can only read. If the new controller inherited admin rights on the bank and the GL, split that in week one.
Days 31-60: the close calendar has to hold
A close calendar is a dated list of who delivers what, in what order, with a day-number target. It is not a hope that we close by the 15th. Write it backwards from the packet date. If owners get the packet on working day 10, bank rec cannot start on day 9.
| Close day | Artifact | Owner | Blocker rule |
|---|---|---|---|
| 1 | AP cutoff, inventory or WIP snapshot, payroll accrual | Controller, with ops on WIP | Missing snapshot slips the close. It does not skip the snapshot. |
| 2-3 | Bank rec, credit card rec, cash tie to GL | Controller | Unmatched items stay on the list. They do not disappear into timing. |
| 4-5 | Revenue, COGS, deferred items, material JEs | Controller, with sales ops on bookings | No private spreadsheet. If it is not in the system, it is not revenue. |
| 6-7 | Flux versus budget, cash recast, packet draft | Controller | Every material line has actual vs budget vs variance, then cause, then risk. |
| 8-10 | Owner review, adjustments, packet out | CEO plus controller | Adjustments after day 8 need a reason that will still make sense in a quarter. |
Hold the calendar in public. Misses get names, not “finance was late.” The day-60 test is two closes on the calendar, not one heroic close. Anyone can make day 10 once. The second month tells you if the calendar is real.
If the close is still past day 15 at day 60, stop adding reports. Fix cutoff and rec. A controller producing dashboards on an unreconciled set of books is decorating. Kill the decoration. Hire-before-automation applies inside finance too. Do not add reporting on a close that still lies. See when to hire before automating.
Days 61-90: who still owns the forecast
This is the fight most companies skip, and it is the expensive one.
The controller should own the mechanical forecast. The model, the actuals feed, the 13-week cash view, and the monthly P&L recast. They are responsible for the math being internally consistent and tied to the books.
The CEO still owns the demand assumptions. Pipeline, close rates, price, hiring plan, and any this-customer-will-land-in-October claim. If the controller owns those, you will get a forecast that is internally pretty and commercially naive. If the CEO owns the model, you will get hope in week two of the cash view.
Put the split in writing. Controller owns actuals, trailing conversion rates, payroll calendar, AP timing, debt service, tax estimates, and the recast after the close. CEO owns bookings forecast, any deal over a named size, hiring adds or freezes, owner draws, and capex. Shared, in a 30-minute weekly meeting: actual versus forecast versus variance on receipts and on revenue, then cause, then risk, then a keep or change on the assumption.
I will make the call. If the CEO keeps a parallel forecast after day 90, you do not have a controller. You have a historian. Kill the parallel file. One model. Two owners of inputs. One packet. If you think you need the shadow file just in case, the case is that you have not accepted the hire.
The day-90 test is a recast that the CEO will run the company on. If the quarterly priorities still come from the founder’s gut while the controller’s model sits unread, the 90 days failed even if the close is on time.
A worked example: $140K hire, two forecasts, one survivor
An $11M field-services company. Forty people. Bookkeeper plus a fractional CFO. Close on day 17. The founder kept an Excel forecast that had beaten the last two years on revenue and missed cash twice badly enough to tap the line. They hired a controller at $140K and kept the bookkeeper part-time for AP.
Days 1-30 we did not touch the ERP. We defined the five-page packet. The founder sat for two hours and cut a 60-report wish list to those five pages. Access was split. The controller lost the ability to release a payment run alone.
First close on the new calendar landed on day 13. Bank rec had 14 unmatched items. We left them on page 5 instead of forcing day 10. That was the correct ugly. Actual EBITDA for that month was $92K versus a $140K budget, a $48K miss. Cause was two jobs where labor ran 18% over bid. Risk was four similar jobs still open. That paragraph did not exist in the old pack.
Days 31-60, second close hit day 10 with 3 unmatched items. AP cutoff started to hold because ops got a day-1 snapshot with a name. The founder tried to add a utilization dashboard. We killed it until rec was clean.
Days 61-90 was the forecast fight. The founder still updated the private file on Sundays. The controller’s recast, tied to the books, showed Q4 cash $220K tighter because the founder booked a $400K job in October that sales had at 30% probability. We ran both files into the same meeting once. The $400K job did not close. The controller’s cash view was the one that would have kept a $175K equipment PO off week 8.
We killed the private file. Wrote the split. CEO owns deals over $75K and hiring. Controller owns the model and the 13-week cash. The equipment PO moved to week 14, after a deposit. Floor held.
What had to go right was the founder accepting one model. By day 90 the $140K started to look cheap against a $175K PO that would have landed in a hole.
Common mistakes
Hiring a controller to fix the books with no packet definition. They will fix something. It may not be what you need to run the company.
Keeping the founder forecast as the real one. Then the hire is a reconciliation clerk for a file they do not control.
Measuring the controller on close speed only. Day 8 with unmatched items and no cause on the variance is a worse close.
Dumping HR, benefits, and systems admin on the new seat. Protect the 90 days.
Starting an ERP project in week two. Implement the packet and the calendar on what you have.
Leaving bank and GL admin on the same login. Split it in week one.
FAQ
Who should own the forecast after a controller is hired? The controller owns the model, the actuals feed, and the 13-week cash recast. The CEO owns demand assumptions, large deals, hiring, and owner draws. Meet weekly on actual versus forecast versus variance. If the CEO still keeps a parallel file after day 90, the hire has not taken. Kill the parallel file.
What should be in the finance packet by day 30? Five pages: P&L actual versus budget versus prior year, cash versus the 13-week view, balance sheet plus DSO and the cash floor, flux notes with cause and risk, and open rec items. The controller assembles it. The CEO spends two hours in week one naming the lines they will actually use. More pages is not more control.
When is the first 90 days a fail even if the person is strong? If the close is still past day 15, if unmatched rec items are hidden instead of listed, or if the company is still running on a founder spreadsheet the controller cannot recast. Those are install failures, not people failures. Fix the packet, the calendar, and the forecast split before you judge the hire.
