The first performance pack from an expensive transformation is not the one anyone wants to present. Delivery is slower, costs are higher and throughput is down. The slide deck promised a better operating model; the spreadsheet is currently making the opposite case.
The executive who approved the change now has one group asking when it will be stopped and another explaining why the numbers should not be taken too literally.
A transformation getting worse before it gets better doesn’t automatically mean it failed.
That sentence is useful. It is also a convenient place to hide a bad plan.
A serious change can temporarily damage the metrics by which the old operating model was judged. The question is whether that damage was anticipated, limited and exchanged for something observable in the new model. Without those conditions, “it gets worse before it gets better” is not an explanation. It is a request for more time.
A bad number can still be on plan
Take a hypothetical ERP migration. During cutover, order-processing throughput falls as people check migrated records, reconcile exceptions and work across old and new systems. Some of the shortcuts that kept the old process moving have been removed, while the team doing the migration has less capacity for ordinary delivery.
For illustration only, suppose the plan permits a 15% throughput decline for four weeks, provided the proportion of orders requiring manual reconciliation declines each week and billing accuracy does not deteriorate. Those figures are invented; the agreement is the point.
The migration still has to show that the new operation is becoming viable. The proportion of orders requiring manual reconciliation should decline as more orders complete without a spreadsheet or the old ERP. Exception categories should be understood and resolved rather than merely counted. More of the daily work should finish inside the new process without somebody rescuing it by hand.
Some measures do not get the same flexibility. Billing accuracy cannot become an experiment, orders cannot disappear or be duplicated, and customer harm cannot keep growing because the programme is “still learning”. A throughput dip may be an agreed cost of migration. Broken financial records are a guardrail breach.
This is the distinction leaders need: what may deteriorate, what must improve while it does, and what cannot be compromised.
When measurement becomes politics
If those decisions were not made before the transformation, they will be made after the numbers turn bad. By then, everyone has a position to defend.
The programme sponsor points to migrated records, completed training and system usage. Finance points to cost, operations to the backlog, and customer teams to complaints. Every number may be accurate. None settles the argument because nobody agreed which evidence would justify continuing.
The programme report may still contain a green box. Finding it is not the same as finding progress.
This is why success criteria agreed under pressure are weak. Executives begin choosing the metrics that support the decisions they already want to make. The transformation team reaches for adoption activity. Its critics select the outcome that makes stopping look obvious. The dashboard stops being a control and becomes ammunition for the meeting.
The cure is not a larger dashboard. It is an agreement made before anyone needs the numbers to win an argument.
Decide before the dip
Before work starts, the plan should say what may deteriorate, by how much and until when. It should also say what must improve during that period, what cannot be compromised, and who decides when the evidence moves outside the agreement.
“Throughput may fall temporarily” is not an answer. Name the maximum decline and the end date. Name the leading evidence expected at each review. Keep billing accuracy, data integrity and customer-harm limits separate from the measures allowed to flex. Then make the decision authority explicit. It may sit with one executive or a committee, but everyone should know who can continue, pause or reset the work.
The decision then becomes clearer, although not comfortable.
If the permitted outcome has deteriorated within its limit and the viability indicators are improving, continue. The business is paying the cost it agreed to pay and receiving evidence in return.
If the decline remains within the guardrails but the leading evidence is flat, intervene. Find the specific constraint, give one person the authority to resolve it and set the next decision date.
If a guardrail breaks, the deadline expires or the new operation shows no credible improvement, trigger the agreed decision process. A contained miss may need intervention; serious customer or financial harm may justify a pause, reset or termination. Rollback is one option, not a reflex, because reversing a live change creates risk of its own. Time and money already spent explain why the decision hurts. They do not justify another month.
The same logic applies to engineering restructuring, vendor replacement and AI rollout. Each has a period where the old measure may worsen while responsibility, capability or operational independence moves. Each also needs a point beyond which deterioration is no longer the price of progress.
Do not panic at the first bad number. Do not ask for faith in the original plan either. Know which decline you approved, what improvement it was supposed to buy, and who has the authority to say the evidence is no longer good enough.
When this matters
How can leaders tell whether worse performance during a transformation is temporary disruption or evidence of failure?
Execution, not consulting.
If you are looking for another strategy deck, Safyron is probably not the right fit. If you need someone who will still be there when the difficult work starts, read how Safyron operates.