Removing Work, Not Moving It

The two are recorded in different places, by different people, on different cadences, so a cut that removes work and one that relocates it look identical in the ledger. Both reduce the same line by the same amount. Both are reported as delivered.

The difference surfaces about three quarters later and somewhere else: overtime in an adjacent team, a rising error rate, a control finding, a resignation. Nothing in the accounts links any of it back to the decision that caused it.

What actually happened

Polymath’s founder ran global human capital planning in a CPG business of roughly $2– 3 billion in revenue that had gone public on the TSX, with payroll across approximately twenty-one countries and a workforce P&L above $200 million. He was the finance lead inside HR, working to the chief financial officer, and owned the HR-related internal controls for about eight years.

Cost targets arrived as they always do: a percentage, a date, a plan wanted by Friday. The team sorted candidates into two piles.

The first pile took out cost and nothing else. Duplicate vendor arrangements sat there — the same service bought twice on either side of a regional boundary, roughly $800K of it identified and removed, and not one step of anybody’s Monday changed. Part of the payroll work belonged there too: a twenty-one-country solution carrying about $500K in identified annual savings, and terms on payroll and Workday negotiated at roughly $350K a year. That pile is smaller than any cost programme assumes.

The second pile took out cost and left the work standing. A payroll coordination role across twenty-one countries is not a function you can point at on a chart; it holds a calendar together: country cut-offs, statutory filing dates, the reconciliation between the HR record and the payroll record before each run. Remove the role and all of it still happens, later, by people who already have a full day.

The clearest version was in the controls. A cut made in another function for sound reasons takes a reviewer out of an approval chain, and nobody in that meeting uses the word control. The review still has to occur, so it migrates up to the approver, who is now signing work he initiated. In eight years of owning those controls alongside the auditors, the planning lead never saw one declared when the cost decision was taken. It surfaced in testing, a year later.

The sharpest case ran the other way. The function rebuilt global reward and recognition without incremental budget, and the model carried roughly $5 million in identified organizational value. No line moved in either direction. What changed was where recognition was triggered, who could authorize it and what paid for it. If value can arrive without a budget line, cost can leave one with no work leaving alongside it.

A saving that survives its second audit removed work. Everything else was a transfer with a delay on it. Polymath sorts a client’s target into those two piles while it is still a number and not yet an announcement, because a cut already declared cannot be re-sorted.

Why it happens

Cost is recorded in a structure built to be reconciled: cost centres, named owners, a monthly close, a tie back to the statutory accounts. Someone answers for every line, every period.

Work has no equivalent. It sits in processes that cross cost centres, and nobody closes the books on a process. Where a map exists it was drawn for an implementation or an audit, and describes the design rather than the running state.

So the two are visible in different places, and the cut has to be decided somewhere. It gets decided in the only view where the thing being cut is countable. In the budget a role or a contract is a number with an owner. The steps hanging off it have no field.

From inside that view, two different decisions are indistinguishable. One removes work: the step stops, or stops being needed. The other relocates it: the step still has to happen, somewhere else, by someone with less context.

A budget can be cut in a meeting. A process can only be changed by redesigning it, and redesign does not fit inside the quarter the number was promised in.

That asymmetry settles it whenever a date is attached. The cut is available now; the redesign is not.

The relocated work reappears in a different unit of measure. Not as cost: as hours, as errors, as a cycle that runs longer, as a control finding, as somebody handing in their notice. Different function, later quarter. No expense line records which decision caused it.

Because nothing connects them, the re-emergence is explained locally, and every explanation is plausible. Volume was up. The team is stretched. We made a poor hire. Each earns a remedy — a contractor, an overtime approval, a temporary resource — and each restores the cost without restoring the design, generally at a premium. The ledger is not lying. The line did come down. It came down because the work left the line, and the ledger cannot ask where it went.

The strongest objection

The strongest case against this: we need the number this year, and what you are describing takes three.

Concede it entirely. Sometimes the cash requirement is immediate: a covenant, a demand collapse, a commitment already given to a board. A programme returning benefit in eighteen months is no answer to a problem due in one. There are quarters where you take the transfer knowingly.

But the choice is not between cutting and redesigning. It is between cutting with the work mapped and cutting without. Asking which steps stop and where the rest lands takes days, not quarters. It changes what you write down, which decides whether you cut this same cost again next year.

The better objection: organizations carry slack, and pressure is the only thing that ever finds it. True, and it is the honest case for a blunt cut. Some absorbed work should never have existed, and no analysis was going to identify it. Only the constraint did.

Absorbed and deferred look identical for about a year. The difference is whether the step got faster or only got later, and the only way to know is to look again twelve months on. Almost nobody does.

The method

We run six steps, in this order. The first two are what make the remaining four enforceable.

1. We price the target in work before we price it in dollars. For each candidate cut, we list the process steps attached and mark every one: stops, moves, gets slower, gets automated. We spend two hours at a whiteboard with the people who run it. The trade-off we accept: a fortnight of delay before anything can be announced. Skip it and you learn which category you chose a year later, from somebody else’s number.
2. We book cost removed and work relocated as separate lines. We add one column to your savings register: where the work went, and to whom. Relocating work is sometimes right; we do not let it be pretended about. The trade-off we accept: the headline shrinks in the room where it is presented. Skip it and every cut is carried at gross, and within two cycles the register reconciles to nothing.
3. We get the receiving owner to agree before the cut lands. Relocated work needs a named person who will do it — not the function head, the person — and we get that agreement while the decision is still reversible. The trade-off we accept: some refuse, which turns a clean decision into a negotiation and occasionally kills the saving. We treat a refusal as capacity data, not obstruction. Skip it and the work settles on whoever notices first, reliably the most conscientious person on the team.
4. We run every cut past your control map before approval. A cut touching an approval, a reconciliation or a segregation-of-duties boundary changes a control, whether or not the word came up in the room. Control changes made for cost reasons are discovered rather than declared, so we declare them while the cut is still a proposal. The trade-off we accept: one more review before approval, and a few cuts blocked outright. Skip it and remediation prices the saving for you, at the point in the year when finance has least attention to give it.
5. We fund the redesign cases on a longer clock. Where the work must genuinely continue, we treat redesign as the only cut that survives next year. The twenty-one-country payroll design carried about $500K in identified annual savings, and the recurring part of it was re-keying and reconciliation that no longer had to be performed. The trade-off we accept: nothing lands in-year, and it consumes the same scarce people now running the process. Skip it and you cut this cost again in eighteen months at a higher price.
6. We re-audit last year’s savings before we set this year’s target. We take the ten largest from the prior cycle and ask of each: is the cost line still down, is the work still being done, and by whom. The trade-off we accept: some will have failed in front of the people who booked them, and it costs two weeks in a cycle with none spare. Skip it and this year’s target sits on last year’s fiction.

Three of those steps belong to finance, two to your process owners and one to your control owners, and no cost programme convenes all three, so we convene them. That is why we run this as Process Redesign rather than out of a savings office. A target written as a percentage of a budget can only be answered inside the budget, while the work attached to it runs across Procure-to-Pay, into payroll, into the controls above both. We redesign the steps and the vendor scope, and the line follows without being pushed. Push the line first and the steps redistribute themselves, out of sight.

Evidence and measures

The second pile, written out. Every row reduced a real budget line:

Five measures worth tracking over ninety days. None of them requires a new system:

— Of last year’s ten largest booked savings, how many still show the cost line down today.

— Overtime and out-of-hours system activity in the teams adjacent to each cut, not the one cut.

— How many entries in the savings register name a receiving owner for work that moved rather than stopped.

— Elapsed time from a cut being approved to the first backfill or contractor appearing in the same process.

— Control changes in the last twelve months that an auditor identified rather than the business declaring.

If those move, the work came out. If the register grew while overtime and contractor spend grew with it, what you scheduled was a repurchase.

How Polymath solves this

We start with your target itself, before it becomes a plan. In the first weeks we trace every contract and role inside the number to the process steps it actually carries: what a vendor performs that a platform you already own performs too, what a step exists to correct rather than to produce, which steps stop if this cut lands and which merely move. We run that as Process Redesign applied to Procure-to-Pay, with technology rationalization inside it rather than beside it, and what we hand over is a redesigned process before a savings figure.

We execute the redesign rather than recommend it. We rebuild the steps and consolidate the platform footprint alongside your procurement, HR and finance teams, and we reopen commercial terms only after the process has changed, so what a vendor prices is the redesigned scope rather than the one that produced the duplication. Then we hand the redesigned steps and the register back to the owners who will run them next cycle.

What compounds is the register underneath it: each step with an owner, a system and the control it sits inside. Once you hold that, next year’s target can be set in work rather than in percentages, your segregation-of-duties view is available before an auditor asks for it, and the steps that turn out to have no owner become the honest shortlist for automation. Nothing in that register has to be discovered twice, which is why the cuts made against it are the ones still standing when they are re-audited.

What it costs to do nothing

The first cost is that repurchase. Relocated work comes back as a contractor, an overtime approval or a backfill, at a premium, because by then it is urgent and whoever understood it has gone. You buy the same capacity twice and book one saving. The second is control exposure, priced outside the building. A control weakened by a cost decision is found in testing, and remediation runs on the auditor’s calendar. In a public company that consumes finance and executive attention in the quarter you can least spare it.

The third appears in no register. The people who quietly absorb relocated work are the ones who can see across the whole process, which is why they have somewhere else to go. When they leave, the map of who actually does what leaves with them, and the next cost programme is planned blind.

The ledger will tell you the cut worked. Only the process will tell you whether it held, and it answers about three quarters late. So we open the target first, not the ledger.