Most outsourced CFO firms run a monthly cleanup. Six Sigma factories run DMAIC. The difference shows up exactly when it matters — in your portfolio company’s diligence room, when the QoE team asks why the close cycle is 18 days and the categorization rules have changed three times this year.
What DMAIC actually is
DMAIC — Define, Measure, Analyze, Improve, Control — is the structured problem-solving cycle Bill Smith introduced at Motorola in the 1980s and Jack Welch scaled across GE in the 1990s. The discipline pushes a process from the industry-default of 3-sigma performance (66,800 defects per million opportunities) toward 6-sigma (3.4 DPMO). Welch put it plainly: “Six Sigma is the most important initiative GE has ever undertaken.” GE reported over $12 billion in savings in the first five years.
DMAIC is not a cleanup. A cleanup is event-driven — something breaks, you fix it, you move on. DMAIC is cycle-driven. It runs against the same process every period, asking: what’s the defect rate this cycle? Is it lower than last cycle? The compounding gains come from the cycle, not from any single fix.
The five phases:
- Define — What’s the process? What’s a “defect”? What does the customer (in our case, the QoE buyer) require?
- Measure — Establish the current baseline. Cycle time. Defect rate. DPMO.
- Analyze — Root cause. Pareto the defects. Statistical tests if the data supports them.
- Improve — Implement the fix. Pilot, measure, expand.
- Control — Lock the gain. Control charts. SOPs. Defect rate monitored continuously.
Applied to the monthly financial close
Most portfolio company closes are run intuitively. The controller closes when the books “feel done.” That’s the 3-sigma default. Here’s what DMAIC looks like on the same problem:
Define. This is the phase most close projects skip — and the one that determines whether everything downstream is honest. A complete, accurate close = books reconciled, no uncategorized transactions, no post-close reclasses. Target cycle time: 4 business days. Defect = any transaction that gets reclassified after close, any uncategorized transaction lingering past day 4, any reconciling difference flagged in lender or QoE review. Without a written definition of “defect,” every subsequent measurement is meaningless.
Measure. A real baseline from a recent trade-services engagement: 18-day close, 1,400 uncategorized transactions per month, 12 post-close reclasses, two AR reconciling differences flagged in lender diligence the prior quarter. DPMO baseline ≈ 28,000. Sigma level ≈ 3.4. About what you’d expect from a hand-managed close.
Analyze. Pareto the defects. In that engagement, three sources accounted for 81% of the volume: (a) Ramp transactions auto-categorizing to “COGS — Supplies” instead of specific GLs, (b) vendor master inconsistency producing duplicate vendor names, (c) missing job-cost tags on field expenses. Root cause was structural, not personnel.
Improve. A rules-based categorizer encoded the top 60 vendor → GL mappings. The vendor master was deduplicated and locked. A job-cost linker tied every Ramp/QBO line to a project at the time of capture, not at close. Each improvement was piloted on one division for two cycles before expanding.
Control. A 16-check monthly data-quality scan runs on workday 2. A control chart tracks close cycle time and uncategorized count. Any month that breaches the upper control limit triggers a same-cycle root-cause review, not a “we’ll look at it next month.”
Result, three cycles later: close cycle 4 days, uncategorized transactions <20/month, post-close reclasses 0. DPMO ≈ 3,000. Sigma level ≈ 4.2 — and still improving each cycle. The lender’s next quarterly review produced zero data tickets. The CFO got an extra week of his life back every month, and the operating partner stopped getting Friday-evening text messages about what the trailing-twelve looked like.
What this means for diligence
A PE/QoE buyer should ask three questions of any portco’s finance function:
- What’s your defect rate on the close, in DPMO? If the controller can’t answer, the close is being run intuitively. That’s the 66,800 DPMO baseline. Diligence will surface it.
- What’s your control chart for cycle time? No chart = no control. No control = the gain is not durable. The next QoE review will find the same issues again.
- What’s your DMAIC cycle for the close, by phase? If the answer is “we just close the books,” you’re buying a 3-sigma operation and paying a 6-sigma multiple.
The gap between “books that pass diligence in one round” and “books that take three rounds and a $50K legal-and-accounting fee to defend” is exactly the gap between intuitive monthly close and DMAIC monthly close. It’s not a personnel issue. It’s a methodology issue.
The discipline we apply
Every Kaizen CFO engagement runs the same DMAIC cycle on the close — and on AR, AP, and revenue completeness alongside it. Cycle time goes down. Defect rate goes down. The diligence room gets easier each quarter, not harder.