The Plan Should Never Go Stale

I've been thinking about a basic contradiction in modern planning.

Companies have spent years making operational data more current. Transactions arrive faster, forecasts refresh more often, and planning engines can recalculate large models in minutes.

But when something important changes, the official plan can still take days to catch up.

Consider a supplier delay that becomes known on Monday. The planning team extracts and reconciles the data on Tuesday. It investigates the consequences on Wednesday, reviews alternatives on Thursday, and gets a decision approved on Friday.

The company had the data on Monday. It did not have an authorized response until Friday.

This is not an unusual rhythm. ASCM still describes demand planning in many organizations as a monthly cycle built around S&OP, where functions come together to agree on one plan. That process matters because planning is how an organization reconciles competing priorities and decides what it will actually do.^[1]

The problem is that reality does not wait for the next cycle.

The delay after the data arrives

Most planning technology is evaluated through forecast accuracy, calculation speed, scenario generation, or the percentage of decisions made without human touch. These are sensible measures. But they do not tell us how long it takes the organization to respond when operating reality changes.

We have started calling that gap planning latency: the elapsed time between a material change in operating reality and a valid, authorized change in the executable plan.

Data latency and planning latency are not the same thing. A supplier delay can be visible in the system while the organization is still determining what demand is affected, whether inventory is usable, which commitments have already been made, and who can change the plan.

Research has been pointing toward this distinction for some time. In a widely cited supply-chain study, Gérard Cachon and Marshall Fisher found that sharing demand and inventory information reduced costs in their model, but shortening lead times and reducing batch sizes produced much larger benefits. The study was not testing Zero-Day Planning, and I do not want to stretch its conclusion. But it does reinforce a practical point: making information available is not the same as converting that information into a better operating response.^[2]

Maintain the state instead of rebuilding it

Our working hypothesis is that the next meaningful advance in planning will not come from making the existing planning cycle faster. It will come from no longer treating that cycle as the primary way the plan is kept current.

A plan is a maintained state of demand, inventory, supply, constraints, commitments, policy, and risk. When accepted evidence changes, deterministic software should recalculate the consequences. Routine changes within approved bounds should not require the organization to revisit the plan. Material exceptions should reach the person with the authority and context to decide.

That changes the operating model.

Instead of repeatedly collecting data, rebuilding the plan, investigating the result, meeting, approving, and publishing, the organization maintains its understanding of reality, calculates the consequences of change, resolves bounded decisions, and surfaces the exceptions that require judgment.

The distinction is important because planning is not just analytical work. A forecast estimates what may happen. A plan records what the organization has decided to do. That decision may reserve inventory, change production, create a purchase commitment, expose one customer to protect another, or spend money. Calculation can show the consequence. Calculation cannot grant the authority to act.

Research on S&OP makes a similar point. In a detailed case study, Rogelio Oliva and Noel Watson found that alignment in executing the plan could matter more than information or process quality alone. A separate study of 725 manufacturers across 34 countries found a positive, moderate-to-large relationship between internal S&OP practices and manufacturing performance.^[3][4]

The planning system, in other words, has to support organizational commitment, not just produce a better answer.

Speed needs a control system

We call the desired state Zero-Day Planning: the organization begins each day with a plan that is already current, decision-ready, and governed.

Zero-day does not mean every signal changes the plan, every decision happens instantly, or people disappear from planning.

In fact, research on rolling-horizon planning has documented the trade-off between responsiveness and plan stability. Frequent revisions can create schedule nervousness, and approaches that perform well on cost do not necessarily perform well on stability. A continuously recalculated plan without materiality rules, decision bounds, and explicit authority could create continuous disruption.^[5]

That is why governance cannot be added after automation. It is what makes responsible speed possible.

Deterministic software should own the calculations, constraints, and committed planning state. AI can help interpret evidence, investigate exceptions, compare feasible alternatives, and explain consequences. People and institutions should govern policy, authority, judgment, and accountability.

The objective is not maximum automation. It is the minimum necessary human intervention required to maintain a correct, authorized, executable plan.

What we are testing

Our research is not finished. The evidence supports the importance of integrated planning, timely response, and plan stability. It does not yet prove that continuously maintaining the planning state will outperform periodic reconstruction in every setting.

That is the hypothesis we now need to test.

We need to learn which parts of planning latency dominate in different operating environments, whether reducing it improves service, inventory, stability, and planning labor, and which decisions can safely proceed within delegated authority.

We are building Vista as one implementation of this theory. The system is designed to maintain an authoritative planning state, use reproducible calculations to determine what changed, apply policy and authority to routine decisions, and bring material exceptions to the right person with the evidence needed to act.

I do not think the next generation of planning software will be defined by how many forecasts it produces or what percentage of decisions are touchless. It will be defined by how quickly an organization can turn a meaningful change in operating reality into a valid, authorized plan.

If the hypothesis holds, the planning cycle will not merely get faster. It will stop being the primary way the plan is kept current.

Every day will start with a plan the organization can act on.