Documenting a thirty-year process before retirement

A view of the San Francisco Bay from the client’s offices on the Embarcadero, on a foggy morning during the engagement. The Ferry Building tower is in the middle distance; a sailboat is barely discernible further out. Photograph: house archive.

This entry describes a five-day kaizen at an employee-owned services operator in the western United States, run on the fixed-asset accounting process the firm uses to track its capital base. The process — creating, transferring, and disposing of fixed assets in the firm’s ERP, with monthly reporting back to corporate finance — had been run for three decades by a single accountant, almost entirely from her head. She was about to retire; her replacement had been hired but had not yet started. The kaizen’s job was to get the process out of her head, into the firm, and into shape for the new accountant in five days. The client’s name and the names of individuals have been altered at the client’s request.

The team’s Monday-morning framing for the engagement was a metaphor. Lean, in this account, is the discipline of walking through fog.

In a fog, you do not know you are getting wet, but as you keep walking you get wet little by little. If your mind has ideas of progress, you may say, “Oh, this pace is terrible!” But actually it is not. When you get wet in a fog it is very difficult to dry yourself.

The fixed-asset function, on the Monday morning the kaizen began, was a fog the firm had been walking in for thirty years.

The firm runs a large and unusually heterogeneous fixed-asset register. Trucks, transfer-station equipment, organics-processing plant, and many tens of thousands of wheeled refuse carts — each cart is a fixed asset on the books, each one has to be created when it is bought, transferred when it moves between operating subsidiaries, and disposed of when it is taken out of service. The combined work of running that register through the firm’s ERP — the routine creations, transfers, and dispositions, plus the monthly reconciliation back to corporate finance — had been performed for over thirty years by a single accountant. Call her Marian. The work occupied something close to fifty-four hours a week of her time, with the close compressed into a hectic end-of-month and the capitalisation work fitted in around it.

By the time the firm engaged the practice, Marian had announced her retirement. Her successor had been hired but had not yet started. The process was almost entirely undocumented; nothing on paper described how the work was done; the firm’s confidence in its capital reporting depended, in practice, on her continuing to do it. The four objectives the kaizen team set on the charter Monday morning were straightforward in statement and difficult in execution: secure the financial-deliverable quality of the work, secure the non-financial information quality, automate as much of the manual data entry as could reasonably be laid out in five days, and produce a documentation set good enough that the new accountant could be trained from it.

The Monday interviews mattered as much as the charter did. Marian, asked how confident she was that her workload could be inherited by a younger accountant, said plainly that it could not. The new person is going to die on the job, in her words. The fifty-four-hour weeks she had run for thirty years were, in her experienced judgement, undoable by anyone the firm could plausibly hire. Asked whether she had children, she said she had a daughter. Asked whether her daughter could do the work, she said her daughter was quite bright and could very probably do it. The thin end of the wedge had been inserted: the job was doable, by another competent person, but not in its current configuration. The work, not the worker, was the constraint.

The other Monday interview was with the firm’s controller. The team’s lead asked him whether, when he made the offer to the new accountant, he intended to describe the job as the fifty-four-hour-a-week role it actually was, or whether he would describe it as a forty-hour role and allow the new hire to discover the truth at the close of the first month. The question was deliberately ungenerous. The controller, asked to choose between dishonesty and a redesign, chose the redesign — and in the same conversation committed his finance organisation to making the changes the kaizen would propose.

The Tuesday-morning observation was that Marian was not, for most of her week, doing accounting. She was doing research. A line item came up the chain from one of the operating subsidiaries — a bollard installed in a new transfer-station bathroom, say — and Marian had to determine what a bollard was, what asset class it belonged to, whether the price the subsidiary had paid for it looked reasonable, and whether the entry the subsidiary had submitted contained any of the small mistakes that, accumulated across thousands of items, would distort the firm’s capital reporting. She had been the firm’s central encoder for fixed-asset transactions for thirty years on the strength of her accounting expertise, but the binding constraint on her throughput was not accounting expertise. It was operational expertise — the slow accumulation of knowledge about what each of the things on the books actually was.

The pivot was to ask which way around the asymmetry of expertise pointed. The accounting knowledge required to encode a fixed-asset transaction correctly is bounded — five or six principles a non-accountant can learn in an afternoon. The operational knowledge required to recognise the long tail of equipment, materials, and structures the firm was buying every month was vast and lived in the operating subsidiaries, with the people who had ordered the equipment in the first place. Marian had been sitting at the wrong end of the asymmetry: the central accountant doing the operational research, when the operational research was easier to push down than the accounting was to push up. The team’s working framing for the rest of the week: it is easier to teach a thousand operators five accounting principles than to teach one accountant a thousand operational items.

From Wednesday the team walked the redesign back. Five principal moves came out of the event.

Push encoding upstream. The central move. The kaizen team relocated the encoding work from Marian to the originators across the firm — site accountants, project managers, procurement clerks — who already had the operational expertise to recognise what they had bought. They were equipped with a short accounting toolkit and a one-page reference card; the standards-keeping function stayed with Marian (and her successor), but the encoding itself moved to the source. The flip is a textbook reduction in excess processing, sized to the asymmetry of expertise rather than to the convenience of the central function.

Tighten the boundary of the work. Three moves under one heading. The capitalisation threshold was raised from $500 to $1,500, removing the long tail of low-value items that had been generating administrative work without affecting the reported capital base materially. Precision was reduced where the noise floor allowed: for used-truck disposals, prior-year average sale prices uplifted by a fixed percentage replaced individual price discovery, on the operational observation that the variance from the average was smaller than the rounding the larger figures were already absorbing. And several monthly reports the function had been producing for years were deleted, because no one was reading them — the cleanest underutilised-talent and NVA processing reduction available to any back-office Lean engagement.

Level the work across the month. Marian’s work had been concentrated at month-end, with the close compressed into a hectic week and the capitalisation fitted in afterwards. The redesign split the work temporally — closing in the first half of the month, capitalisation in the second — using the same hours but spreading them over a more predictable cadence. The peaks came down. The lead time on closing came down with the predictability.

Lay out automation for the largest data-entry block. The wheeled refuse carts — the firm calls them by the brand name Toter — accounted for ten of the twenty hours of weekly data-entry work, on the strength of volume alone. The team laid out a JDE-compatible automation scheme that would handle eight of those ten hours through a structured upload from the procurement records. The work to build the scheme was scheduled for the sixty days after the event; the architecture was settled during the kaizen.

Capture the audit logic as an error registry. Marian’s internal audit — the rules by which she examined the monthly FAT report for errors before submitting it — would not reduce to a checklist; her checks were a habit built across thirty years of seeing the same kinds of mistakes. The kaizen replaced the attempt to extract her rules with a registry of error types, prioritised by occurrence, detectability, and severity. Her successor would not arrive with Marian’s pattern recognition, but would inherit the structure within which the pattern recognition could be rebuilt.

By the close of Friday the team had drawn a function in which the documentation existed, the encoding work had been reassigned to the people closest to the operational facts, the boundary of the work had been tightened, the cadence had been levelled, the largest data-entry block had a scheduled path to two-thirds automation, and the audit logic was a registry rather than a habit. The state, by closing day:

Marian’s total weekly workload: ~54 hours → ~28 hours.
Encoding responsibility: Centralised to one accountant → distributed to operating-subsidiary originators, with a five-principle accounting toolkit.
Capitalisation threshold: $500 → $1,500.
Pricing precision (low-noise transactions): Individual price discovery → prior-year average × fixed uplift.
Monthly reports without readers: Eliminated.
Monthly cadence: End-of-month peak → close in first half of month, capitalisation in second.
FAT process documentation: None → SIPOC diagrams covering the four principal processes.
Audit methodology: Tacit, in one accountant’s head → published error registry, prioritised by occurrence, detectability, and severity.
Data-entry hours per week: 20 → projected 12 once the cart-asset automation lands.
60-day tasks list: Eight items, with named owners and due dates.

The twenty-eight hour figure was not an accidental landing. The practice designs knowledge-work roles for about seventy per cent utilisation — twenty-eight hours of book load on a forty-hour scheduled week — on the operational observation that load above that margin makes the work brittle under the variability knowledge work always carries. The role’s previous occupant had run it at a hundred and thirty-five per cent of a forty-hour week for thirty years, on the strength of irreplaceable expertise. The new role would be inherited at thirty per cent below the standard week, with the headroom built in for the variance the close brings every month and the training a new hire needs in her first year.

The single most consequential move was the first one. Marian had been the firm’s central encoder because she had the accounting expertise, and the firm had organised the work around that expertise without ever asking whether it was the binding one. The work she had been asked to do was, considered carefully, inductive: to reason from sparse line-item descriptions back to what each item actually was — bollard, $1,200, project XYZ — before she could begin to classify it. The same work, performed by the originator who had ordered the item in the first place, was deductive: a known thing, classified against a small set of accounting principles. In the kind of classification work the FAT process required, induction is slower and more error-prone than deduction; the kaizen moved the work from the slower side of the asymmetry to the faster one. The Lean canon’s eight-wastes vocabulary names the same fact as excess processing and underutilised talent, but the epistemological frame may be the cleaner statement of why the redesign worked: the firm had been asking the wrong direction of inference of its accountant for thirty years. Three months after the event Marian retired on schedule. Her successor closed the books on time, in twenty-eight hours a week.

· · ·