A specialty trade-services operator. About $8 million in revenue, two operating divisions, 15 years owner-run. A minority growth-equity partner had just come in. The new arrangement included quarterly diligence reviews, a 3-year liquidity horizon, and the expectation that the books could carry weight under outside scrutiny. The “good enough” close that had served the owner for a decade was about to become the most visible bottleneck in the business.
This is what the teardown looked like.
The baseline
The close was finishing on workday 18 of the following month — by the time the prior month’s P&L was final, the next month was almost over. Approximately 1,400 transactions per month posted to “Ask My Accountant” or a generic COGS bucket. The vendor master had 340+ duplicates: the same vendor present two, three, sometimes four times under spelling variations. About $80K of unapplied customer payments and undeposited funds sat on the balance sheet, growing each cycle. Roughly 10% of pay periods had at least one employee posted to the wrong department. Inventory adjusted once per year, which meant the intra-year P&L was structurally misstated by four to seven points of gross margin.
The lender’s quarterly review produced six to eleven data tickets per cycle. Three rounds of back-and-forth was the norm.
That’s not a personnel problem. That’s a methodology problem.
What got measured
The first month of the engagement was instrumentation, not fixing. A 16-check monthly data-quality scan went into place to provide a baseline measurement instrument — the same instrument that would track every subsequent improvement. The first-run output: roughly 28,000 defects per million opportunities, sigma level ~3.4. That’s the industry default for a hand-managed close. The QoE buyer’s bar is 6-sigma (3.4 DPMO).
A Pareto analysis surfaced three structural sources accounting for 81% of the defect volume: (a) Ramp transactions defaulting to generic COGS, (b) vendor master fragmentation, (c) missing job-cost tags on field expenses. The remaining 19% was concentrated in payroll department errors and the annual-only inventory cadence. None of it required new headcount to fix. All of it required the workflow itself to change.
The five sprints
Each improvement was a one-month sprint. Each one was piloted on a single division for two cycles before being expanded. None of them moved a single line in the books without leaving an audit trail.
- Vendor master deduplication. 340 vendors collapsed to 184. The master was locked, with a workflow gate on new vendor creation.
- Categorizer rules engine. The top 60 vendor → GL mappings were encoded. About 85% of transactions began auto-categorizing at capture, before any human touched them.
- Job-cost linker. Every Ramp and QBO line was tied to a project at the moment of capture, not at close. The job-cost view became real-time for the first time in the company’s history.
- Payroll reclass engine. A two-stage journal-entry automation pulled from a Payroll_Splits sheet, eliminating the manual re-coding cycle that produced the 10% department-error rate.
- Inventory JE engine. A monthly reverse/post pattern by division replaced the once-yearly adjustment. The intra-year P&L started reflecting reality immediately.
Underneath each sprint was the same DMAIC discipline: define what good looks like, measure the current state, analyze root cause, improve in a controlled pilot, lock the gain in a control chart.
What changed
Three close cycles in:
- Close cycle: 18 days → 4 days (−78%).
- Uncategorized transactions: 1,400/month → <20/month (−99%).
- AR unapplied + undeposited funds: $80K → $4K.
- Payroll department errors: ~10% → 0%.
- Inventory: monthly true-up. GM% materially accurate every month.
- Defect rate: 28,000 DPMO → ~3,000 DPMO. Sigma 3.4 → 4.2, and still improving.
- Lender review: zero data tickets in the next quarterly cycle.
The CFO got a week of his life back every month. The operating partner stopped getting Friday-evening text messages about what the trailing-twelve looked like.
What it meant when the company sold
Three quarters after the teardown, the operating company began a sell-side process. The QoE ran in one round, not three. No proposed EBITDA adjustments held under scrutiny. Working capital normalization was defensible to the dollar. The diligence experience was clean.
Our internal estimate: somewhere between $1.5M and $2.5M of enterprise value was preserved on a $4–5M-EBITDA business — entirely on the back of a clean diligence room and the ability to defend every line. That preserved value showed up the day of the LOI, not the day of the engagement. But it accrued every month, one DMAIC cycle at a time.
What to ask before the next close
A PE buyer or owner-operator can run this audit on any portfolio company in 30 minutes:
- What’s our close cycle, in business days?
- What’s our defect rate, in DPMO?
- Where are the top three structural sources of those defects?
- What’s our control chart on cycle time?
If any of those questions don’t have answers, the close is being run intuitively. There’s a 6-sigma operation hiding inside it. The teardown is how you find it.