The forecast is only the beginning of the decision. Someone also needs to understand what that change means for inventory, capacity, procurement and customer service. In many organizations, these questions still move through several planners and meetings before the full impact becomes clear.

This is where the end-to-end planner comes in. An end-to-end planner understands how decisions connect across demand, inventory, supply, production and procurement, and can follow an issue beyond the boundaries of one functional plan.

From planning a number to planning its consequences

Planning roles have traditionally been divided for good reason. Demand planners understand customers and forecasting, supply planners understand constraints, while production and procurement planners work closer to execution. End-to-end planning does not require removing that expertise.

The change is in how decisions are managed. BCG describes end-to-end planning as increasingly decision-led, starting with the important business decisions and organizing people, processes, data and technology around them.

Consider our demand increase. Instead of stopping after changing the forecast, the planner needs to understand whether inventory can absorb it, whether capacity is available, whether materials can arrive in time and where shortages could emerge. Specialists may still become involved, but the planner maintains the line of sight across the decision.

The problem is not specialization. The problem starts when the handovers between specialists become the planning process.

What happens to the traditional IBP cycle?

This development also affects Integrated Business Planning. Traditional IBP creates alignment through a recurring sequence of demand, supply, reconciliation and management reviews. Modern planning technology makes it increasingly possible to identify exceptions, evaluate consequences and update plans without waiting for the next meeting.

Deloitte describes a move from cyclical planning toward a more continuous process, with demand and supply monitoring becoming simultaneous activities and operational reviews increasingly focused on exceptions. Importantly, it also cautions against simply removing planning forums: strategic, governance and human decisions still require structured discussion.

A simplified view of that shift looks like this:

Traditional IBPEmerging modelDemand review once per cycleDemand continuously monitoredSupply review follows demandDemand and supply evaluated togetherReconciliation meeting resolves gapsScenarios and trade-offs evaluated as issues emergeFinancial reconciliation later in the cycleFinancial implications considered alongside operational decisionsMBR resolves major open decisionsMBR focuses on major exceptions, policy and strategic trade-offsPlanner prepares a functional planPlanner manages cross-functional decisions and exceptionsCalendar triggers planning activityMaterial changes can trigger planning activity

This does not mean the monthly IBP cycle disappears. It means fewer operational decisions need to wait for it. The formal cycle can increasingly provide governance and alignment while more planning happens continuously underneath it.

That shift also makes the end-to-end planner more relevant. If demand and supply are evaluated together, someone needs enough understanding of both to interpret the trade-off and determine when another specialist or decision-maker needs to become involved.

How do we get there?

Creating an end-to-end planner starts with the process, not the job title. Plans may currently sit in different systems, assumptions may not flow between functions, or decision ownership may simply be unclear. Kearney describes connected planning around a unified decision model that links commercial, supply chain and financial decisions while making trade-offs explicit.

Technology then needs to make broader responsibility manageable. Systems can perform calculations, monitor changes and bring exceptions to the planner's attention, while planners concentrate on interpreting what has changed and deciding what to do. This is closely connected to the shift we discussed in AI Should Move Planning From Analysis to Action.

Scyplan is being developed in this direction, progressively connecting Demand Planning with Inventory, Production and Procurement Planning. The goal is to give planners enough connected information and decision support to manage issues further across the supply chain before another handover is required.

A practical transition can start at a relatively high level:

• Start with decisions. Identify recurring decisions that currently move across several planning functions, such as responding to a projected shortage or a significant change in demand.

• Connect the planning information. Make the relevant demand, inventory, capacity and supply implications visible together so the planner can understand the consequences of a change.

• Broaden planner responsibility gradually. Give planners ownership of a wider part of the decision while keeping specialists involved where deeper expertise is required.

• Adapt the planning cadence. Allow material exceptions to trigger analysis and decisions when they occur, while using the formal IBP cycle for broader alignment, governance and strategic trade-offs.

The destination is not one planner doing everything. It is a planning organization where important decisions can travel further with fewer handovers, without losing the specialist knowledge needed to execute them well.

So why buy planning software at all? It is a fair question, particularly when an internal solution can be created quickly and adapted to exactly what the team needs. The answer becomes clearer once that first working model starts becoming part of the planning process.

You can actually build quite a lot

Imagine a planner who wants to improve the statistical forecast. Historical sales are loaded into Excel, AI helps write the code, a few forecasting approaches are tested and the output looks promising. A dashboard follows, then some exception flags and perhaps an input field where sales can adjust the forecast.

For a small team or a specific use case, this may work perfectly well. It can also be an excellent way to experiment with a new process before making a larger technology decision.

The interesting part starts when colleagues begin to rely on it.

Then other people start using it

Sales wants to enter assumptions. Operations needs another level of aggregation. Management wants to compare this month's forecast with last month's version, while planners need to understand who changed a number and why. Someone asks whether the workbook can refresh automatically from the ERP instead of requiring another monthly export.

None of these requests sounds unreasonable on its own. Gradually, however, the original forecasting workbook starts acquiring data connections, user workflows, hierarchy logic, version management, permissions and business rules. A tool that started as a clever piece of analysis has become operational software.

That changes the economics of the decision. The comparison is no longer between an expensive software subscription and a free Excel workbook. The organization has chosen to build and maintain part of its own planning application.

Congratulations, you are now the software vendor

Once the business depends on that application, somebody has to own what happens next. Changes need to be tested, broken logic has to be fixed, new users need documentation and data connections need maintenance. As more people and processes depend on it, every change also carries more consequences.

AI can make many of those development tasks faster, but it does not remove them. For a manufacturing company, the question therefore becomes whether maintaining an internal planning application is where planners and analysts should spend their time.

Building a useful prototype can be valuable. Supporting an increasingly business-critical application is a different commitment.

Who upgrades the intelligence?

There is another part of ownership that is easier to overlook: the forecasting logic itself. A model that performs well today can remain embedded in a workbook for years simply because nobody has a reason to revisit it.

The M4 forecasting competition evaluated 61 forecasting methods across 100,000 time series. The results illustrate why forecasting is not simply a question of selecting one algorithm and assuming it will remain the right answer indefinitely. Different approaches perform differently depending on the data and forecasting situation.

For a planning organization, model selection should therefore remain something that is tested. Forecast performance can change as demand patterns evolve, more data becomes available or new methods mature.

A specialist planning-software vendor has a commercial reason to keep evaluating methods, improving the product and maintaining the underlying technology. A self-developed solution can do the same, but the responsibility sits with the company that built it. The apparent saving from building internally should therefore be weighed against the capabilities and improvements that still need to be developed later.

Know what you want to own

Vibe coding makes Excel a more powerful planning environment than it was before, which is a good development for planners. It makes prototypes faster, lowers the barrier to experimentation and can solve genuinely useful local problems.

The decision changes when several people depend on the solution every planning cycle, its output drives purchasing or production decisions, and expectations continue to grow. At Scyplan, we would rather see planning teams spend their effort on those decisions than on maintaining the technology underneath them.

The difference becomes clearer as the planning process grows. More SKUs are added, several people contribute to the forecast, different versions start circulating, and the business wants to understand how forecasts have changed from one planning cycle to the next.

At that point, the question is no longer whether Excel can calculate a forecast. It usually can. The question is how much work is required to manage everything around it.

Excel gives you flexibility. Planning software gives the process structure

Excel starts with a blank sheet. The planning team decides how data is organized, which formulas are used, where assumptions are stored, and how the planning process should work. That makes it highly adaptable, particularly when requirements are still evolving.

The same flexibility means the team also owns the mechanics behind the process. Someone needs to understand the formulas, refresh the data, maintain relationships between sheets, protect the right cells, and make sure everyone is working from the correct version.

A dedicated planning tool moves more of that structure into the application. Products, customers, locations, time periods, forecasts, assumptions, and planning versions can be managed within a defined planning model. The team still decides how planning should work, but less time is spent maintaining the infrastructure around it.

Data management becomes part of the planning model

In Excel, planners often need to prepare information before they can start planning. ERP extracts may need to be combined with commercial assumptions, product mappings, customer information, and manually maintained files.

As the planning model grows, these relationships become harder to maintain. A change in the product hierarchy or the addition of a new location may affect formulas, lookup tables, pivot tables, and reporting logic throughout the workbook.

Planning software is designed around these relationships. A planner can work across products, customers, locations, and time while using the same underlying planning model. This becomes important because planning rarely happens at one level. Sales may discuss a category forecast while supply ultimately needs demand by SKU and location.

Collaboration becomes easier to control

Excel can work well when one or two planners manage the process. Complexity increases when sales, finance, operations, and supply chain all contribute information.

Teams often solve this with shared workbooks, input templates, or carefully controlled file versions. The process can work, but it creates additional questions. Which version contains the latest forecast? Who changed a number? Has the latest commercial assumption already been incorporated?

A planning system can maintain users, permissions, changes, and planning versions within the same environment. That makes it easier to distinguish between a baseline forecast, planner adjustments, and the final agreed plan. The planning conversation can focus more on why the number changed and less on finding the right file.

Forecasting is only part of the difference

Excel can calculate statistical forecasts, and sophisticated spreadsheet models can go considerably further than simple averages or trend formulas. Moving to planning software therefore should not be justified simply because a business wants a statistical forecast.

The bigger difference is how forecasting connects with the rest of the planning cycle. Dedicated software can generate forecasts across many planning combinations, organize exceptions, retain adjustments, and connect the output to the wider review process.

This becomes particularly valuable when planners manage thousands of product, customer, and location combinations. Reviewing every forecast manually becomes difficult, so the planning process needs to help direct attention toward the areas where judgment is most useful.

Assumptions, scenarios, and overrides leave a history

Planning decisions regularly require judgment. A customer expects additional volume, a promotion may increase demand, or a new product has limited history. In Excel, a planner can quickly adjust a number or create another scenario.

The challenge often appears later. The team needs to remember what was changed, why it changed, which version contained the adjustment, and whether the assumption should still be included during the next planning cycle.

Planning software can provide a more structured history of those changes. Baseline forecasts, overrides, assumptions, and scenarios can remain identifiable instead of becoming another number in a spreadsheet. This makes it easier to understand how the final plan was created.

Performance can be tracked across planning cycles

Forecast performance is difficult to understand from one planning cycle alone. Planning teams need to see how forecasts, overrides, and forecast error develop over time.

In Excel, doing this consistently often means saving snapshots of previous workbooks, maintaining additional reporting tabs, or creating a separate database of historical forecast versions. The process becomes particularly difficult when the team wants to compare what was forecast three months ago with what was forecast one month ago and what eventually happened.

A planning system can retain these forecast versions as part of the planning history. Once actual demand becomes available, teams can compare the original statistical forecast, subsequent planner adjustments, and final agreed forecast with the outcome. Over time, this helps planners identify persistent bias, understand where forecast performance is improving, and evaluate whether manual overrides are adding useful information.

This cycle history is easy to underestimate when evaluating planning software. It becomes increasingly valuable once the organization wants to improve the planning process itself instead of simply producing the next forecast.

When does Excel actually become the constraint?

There is no universal number of SKUs, planners, or business units at which Excel stops working. Two companies of the same size can have very different planning requirements.

A more useful signal is the amount of effort required to keep the planning process functioning. If planners increasingly spend time refreshing files, reconciling versions, maintaining formulas, preserving historical forecasts, or tracing changes, the spreadsheet itself has become part of the planning workload.

Excel can still remain useful for analysis and ad hoc work. The case for dedicated planning software becomes stronger when the organization needs a repeatable process around the forecast: structured data, controlled collaboration, planning history, performance measurement, and a clear record of how decisions evolved over time.

That is ultimately the real difference. The forecast calculation may only be one part of planning. The larger challenge is managing the process around it consistently, cycle after cycle.

This is a limitation of traditional exception management. Forecast accuracy tells planners something important about forecast quality, but it does not tell them where their limited time will create the most value. With thousands of products to manage, planning teams need to consider the financial and operational consequence of an error alongside the error itself.

Forecast accuracy does not measure business impact

The relationship between forecast accuracy and business performance is less direct than it may appear. Research published in the European Journal of Operational Research used data from the M5 forecasting competition to examine how forecast accuracy translates into inventory outcomes. The researchers found that the relationship depends on factors such as demand characteristics, inventory policies, holding costs and lost-sales costs. Under some conditions, the most accurate forecasting method did not produce the best inventory result.

Consider the two SKUs from the opening example. A 50% error on the $0.10 product looks much worse in an accuracy report than a 10% error on the $10,000 product. Yet the second error may expose substantially more revenue, margin, inventory or working capital. At the same time, price cannot determine priority on its own. The $0.10 item could be a critical component that stops a production line when unavailable, while the $10,000 product might have short lead times and flexible supply.

The relevant question is therefore broader: what is the likely business consequence if this planning issue is left unresolved?

Prioritize by consequence, not percentage error

ABC and XYZ segmentation already introduce some of this thinking. ABC segmentation can differentiate products according to volume or value, while XYZ segmentation distinguishes them according to demand predictability. SAP's guidance on ABC/XYZ segmentation describes how these dimensions can be combined to apply different planning approaches to different products.

For a portfolio of 10,000 SKUs, prioritization can go further. Depending on the business, relevant factors may include selling price, contribution margin, inventory exposure, lead time, service criticality, shelf life, obsolescence risk, availability of substitutes, premium freight and the consequences of a production disruption.

Combining those factors changes the purpose of an exception list. Instead of simply showing the 100 SKUs with the lowest forecast accuracy, it can surface the situations where an intervention has the greatest potential business impact. A 40% forecast error may fall down the list because the financial exposure is small and replenishment is easy, while an 8% error moves upward because it affects a high-value, long-lead-time product with limited alternatives.

A cost dictionary provides the common language

This kind of prioritization requires a consistent way to translate operational situations into financial terms. Supply chain may recognize an excess inventory risk or potential shortage, but comparing those situations becomes difficult when planners and finance use different assumptions about their economic impact.

An internal cost dictionary can provide that common language. Supply chain and finance should agree on how relevant consequences are valued, including inventory carrying cost, contribution margin at risk, obsolescence exposure, premium freight, lost sales and other costs that materially affect the company.

The definitions need enough precision to support consistent decisions. Inventory value, for example, does not represent the full cost of carrying inventory. APQC includes factors such as capital, storage, insurance, taxes, handling, administration, shrinkage and obsolescence in inventory carrying cost. Which components are relevant will vary by company, but the assumptions should be explicit and agreed.

A useful cost dictionary does not need to turn every forecast exception into a detailed financial business case. Its purpose is to give planning teams and their systems a consistent basis for comparing one issue with another.

Software can do the first prioritization

Once those definitions exist, a planner should not need to calculate the financial impact of hundreds of exceptions manually. Planning software can apply the agreed logic across the portfolio by combining forecast performance with inputs such as product value, margin, inventory position, lead time, service importance and relevant cost assumptions.

The output can be a prioritized queue rather than a traditional exception list. A system might identify that a relatively small forecast deviation deserves immediate attention because it affects a high-margin product with low inventory coverage and a long replenishment lead time. Another SKU with a much larger percentage error may receive a lower priority because the financial exposure is limited and supply can be adjusted easily.

AI can assist further by bringing multiple signals together, summarizing why an issue has been prioritized and suggesting the next question or action for the planner. The planner should still be able to inspect the assumptions, understand the drivers behind the recommendation and decide whether intervention is appropriate. This creates a more useful division of work: the technology performs the first screening across thousands of items, while the planner spends time on the smaller number of decisions that require judgment.

The financial estimate does not need to be perfect to improve prioritization. A consistently applied estimate can already help distinguish a statistically large forecast error with little consequence from a smaller error with significant business exposure.

The capability is also organizational

Better software does not remove the need for supply chain and finance expertise. Supply chain teams need enough financial understanding to connect inventory, service and operational decisions with working capital, cost and margin. Finance teams need enough understanding of planning mechanics to recognize why a forecast error does not translate directly into profit or loss.

This broader perspective is reflected in ASCM's SCOR performance framework, which considers customer-oriented measures such as reliability and responsiveness alongside cost, profit and asset performance. Planning decisions often involve trade-offs across these dimensions, which is why one KPI rarely provides enough information to determine priority.

A strong planning organization therefore combines forecasting knowledge, operational understanding and financial fluency. Finance helps establish the economic definitions, supply chain provides the operational context, and technology applies that logic consistently across a portfolio that would be impossible to review manually.

From exception management to value-based prioritization

With 10,000 SKUs, giving every forecast problem equal attention is neither realistic nor useful. Forecast accuracy remains an important diagnostic because it shows where forecasts perform poorly, but prioritization should continue one step further and ask what the error means for the business.

This builds on the move from analysis to action. Planning technology can help determine which actions deserve attention first by combining forecast performance with product economics, inventory exposure, service risk and operational constraints.

For planning leaders, the practical question is what currently determines the order of the exception list. When forecast accuracy remains the main filter, valuable planner time can be spent investigating statistically large problems with very small consequences. A clear cost dictionary, capable supply chain and finance teams, and software that applies their logic at scale can direct that attention toward the decisions where intervention has the greatest potential business value.

That work is changing. Better forecasting models, automation, and AI can increasingly handle parts of the analysis that previously consumed much of a planner’s day. The planner remains central to the process, but the value of the role is shifting toward judgment, prioritization, and decision-making.

From producing the plan to managing decisions

Planning has traditionally involved a significant amount of manual work: collecting inputs, reconciling different versions of the forecast, reviewing individual products, maintaining parameters, and investigating exceptions. Those activities remain necessary, but fewer of them need to depend entirely on the planner.

McKinsey has described a similar direction for autonomous planning, with technology handling more routine planning activities while people focus increasingly on exceptions and decisions that require business judgment.

This changes what good planning looks like. A customer promotion, product phase-out, unexpected order, capacity issue, or commercial commitment may require context that the forecasting model does not have. Instead of reviewing every SKU with the same level of attention, the planner needs to know where intervention is justified and what action should follow.

Technology can prepare the decision. The planner still needs to understand its consequences.

What makes a good planner?

A good planner understands the numbers without automatically accepting or changing them. Statistical forecasts provide an important baseline, while planners can add information about events, customers, products, and operational constraints that may not yet exist in the data.

Knowing when that judgment adds value is an important skill.

A 2022 study in the European Journal of Operational Research examined 30 million AI-generated forecasts at SKU-store-day level for a large European retailer. Planners could adjust the forecasts based on their own knowledge. The researchers found that these interventions did not improve forecast accuracy on average, with large upward adjustments particularly likely to be inaccurate.

That does not make human judgment unnecessary. It shows why intervention needs a reason. A planner who knows that a customer has confirmed a promotion is adding information. Changing a forecast simply because the number feels too low is much harder to justify.

Business outcomes also matter. A 2025 Management Science field experiment at a spare-parts retailer found that allowing merchants to adjust inputs to an inventory algorithm increased profitability by 4.92% on average compared with automation alone. The result comes from one specific retail setting, so it should not be treated as a universal benchmark. It does show that human intervention can create value when people have information relevant to the decision.

A good planner therefore needs analytical discipline as well as business understanding. They should be able to explain why an intervention was made, what information supported it, and whether similar decisions have historically improved outcomes.

Which KPIs should planners focus on?

There is no single KPI that defines a good planner. The most useful scorecard combines planning quality with the operational outcomes that planning is meant to support.

Forecast error and bias provide the starting point. Weighted Absolute Percentage Error, or WAPE, measures total absolute forecast error relative to total actual demand. Bias answers a different question: whether the forecast consistently runs too high or too low. Both need to be measured at a level and horizon that corresponds with actual planning decisions.

Forecast Value Added (FVA) helps determine whether interventions improve the forecast. Instead of measuring only the final number, FVA compares different stages in the forecasting process. The Institute of Business Forecasting & Planning defines FVA as a way to determine which forecasting activities and participants improve the forecast and which make it worse. This can show, for example, whether planner overrides add useful information or consistently introduce additional error.

Service metrics, such as fill rate or on-time-in-full delivery, connect planning decisions with customer outcomes. Forecast accuracy has limited value if customers still face frequent shortages. At the same time, service should not be evaluated in isolation.

Inventory metrics, such as inventory turns, days of supply, and excess or obsolete inventory, provide the other side of the equation. Very high service levels can often be achieved by holding more stock, but that consumes working capital and increases inventory exposure.

ASCM’s SCOR framework reflects this broader perspective by combining metrics around reliability and responsiveness with cost and asset-management performance.

These measures should therefore be read together. Improving forecast accuracy while inventory continues to increase may indicate that the planning process is optimizing the wrong level, horizon, or metric. Reducing inventory while service deteriorates creates a different problem. Good planners understand these relationships rather than optimizing one KPI in isolation.

Better technology should raise the value of planner judgment

As planning technology improves, the planner does not disappear from the process. The work moves toward the situations where human judgment matters most: interpreting unusual events, challenging assumptions, evaluating trade-offs, and coordinating action across the business.

That also changes what planners should expect from their software. Instead of searching through dashboards and investigating every exception themselves, systems can increasingly identify what changed, prioritize where attention is needed, and provide the context required to make a decision. We explored this further in our article on moving planning from analysis to action.

The goal should not be a planner who makes more adjustments or touches more forecasts. A strong planner knows which situations deserve attention, can explain the decision that follows, and measures whether those interventions actually improved the result.

For a mid-market planning team, the following signs are worth watching.

A quick readiness check

• Preparing the plan takes too much manual work. Data is repeatedly exported from the ERP, copied into templates, reformatted, reconciled, and consolidated before planners can begin reviewing the actual forecast.

• You are managing multiple versions of the truth. Sales has one file, Supply Chain another, and Finance may have its own assumptions. Considerable effort goes into establishing which version is current before a decision can be made. ICAEW specifically highlights version control as an important control as organizations become increasingly dependent on spreadsheets.

• More people need to contribute to the plan. A workbook managed by one planner is relatively easy to control. Complexity increases when several functions need to enter assumptions, review changes, or work at different product, customer, and location levels.

• It is becoming difficult to understand what changed and why. Forecast overrides, assumption changes, and corrections happen throughout the planning cycle, but reconstructing who changed a number and why requires emails, meeting notes, or comparisons between files.

• Planners spend too much time maintaining the process. Refreshing data, checking formulas, rebuilding reports, and preparing meeting packs take time away from investigating exceptions and making planning decisions.

• The spreadsheet is becoming increasingly difficult to maintain. More formulas, tabs, macros, data connections, and dependencies make changes harder and increase reliance on the people who originally built the workbook. ICAEW recommends considering whether users can continue to understand and maintain a spreadsheet as requirements evolve.

One question to answer before buying software

Recognizing several of these symptoms is a good reason to start evaluating planning software, but the organization also needs a planning process that is clear enough to put into a system. Before making that move, the team should be able to explain who owns the forecast, which data feeds it, at what level decisions are made, how frequently the plan is reviewed, who can override it, and which metrics are used to judge performance.

That prerequisite matters because software can structure and automate an existing process, but it cannot resolve unclear ownership or an undefined way of working. McKinsey's research on supply-chain planning technology similarly emphasizes the importance of defining the future planning process and its requirements before implementing new technology.

If several of the signs above feel familiar and your planning process is already reasonably well defined, that is the point at which it becomes worth asking whether Excel should remain the primary planning environment. Excel can still play an important role in analysis and one-off calculations, while dedicated planning software takes over the recurring processes that have become difficult to manage through spreadsheets alone.

If you are unsure whether your organization has reached that point, we at Scyplan are happy to help assess your current planning process and determine whether moving beyond Excel makes sense yet.

These different perspectives are exactly why the S&OP or IBP demand review is valuable. Commercial teams and planners can bring information to the forecast that a statistical model does not yet know, such as a confirmed customer order, a promotion, a distribution change or an expected market event. At the same time, the review creates a risk that expectations, targets and operational constraints gradually become mixed into the demand forecast itself.

Recent research suggests that this distinction deserves more attention. A 2025 study in the International Journal of Forecasting analyzed around 147,000 forecasts across six company datasets and examined what happened when people adjusted system-generated forecasts. Judgmental adjustments improved both accuracy and bias for only slightly more than half of the SKUs, while upward adjustments were particularly likely to worsen forecast performance.

The finding does not suggest that human judgment should disappear from demand planning. It raises a more useful question for the demand review: when someone changes the statistical baseline, are they adding information about expected demand or expressing what they would like the number to be?

The baseline gives the discussion an independent starting point

Research into judgmental forecasting has shown that organizational roles and incentives can influence forecasting behavior. That makes the statistical baseline valuable for a reason that goes beyond the quality of the algorithm. It gives everyone in the demand review the same independent starting point before commercial ambition, capacity constraints and management expectations enter the discussion.

That baseline should remain visible even after an adjustment has been made. If Sales expects higher demand because a customer has confirmed additional volume, the review should capture both the original forecast and the adjustment, together with the reason behind it. Replacing the baseline with a new consensus number makes it much harder to determine afterward whether the intervention actually added value.

Watch for signals that the forecast is becoming a negotiation

There is no problem with challenging a statistical forecast. The concern starts when the reasons for changing it become disconnected from new information about demand. A few situations deserve particular attention during the demand review:

• The forecast repeatedly moves toward the budget or sales target. A commercial objective may be achievable, but the target itself is not evidence that customers will buy more.

• Capacity constraints lead to a lower demand forecast. Limited production may affect what the company can supply, but it does not necessarily change underlying customer demand.

• Overrides are explained with confidence rather than evidence. Statements such as “Sales believes the market will be stronger” are difficult to evaluate unless they are connected to a specific customer, promotion, channel or market event.

• The same direction of override appears repeatedly. Consistently increasing or decreasing the baseline is worth investigating, particularly when the adjustments do not improve forecast performance over time.

These are not reasons to reject human input. They are reasons to make the source of that input more explicit.

An override should earn its place

The recent International Journal of Forecasting research is particularly relevant here because it uses Forecast Value Added to examine whether judgmental interventions actually improve the forecast. Rather than evaluating only the final forecast, planning teams can compare the statistical baseline with the adjusted forecast and determine whether the intervention reduced or increased error and bias.

This creates a useful feedback loop for the demand review. A team may discover that adjustments around confirmed promotions consistently improve the forecast, while broad commercial uplifts tend to make it worse. Another business may find that local market knowledge adds significant value for certain products but contributes little for mature, stable categories. Human judgment becomes more useful when the organization learns where it has historically added information that the model was missing.

Keep the target visible without turning it into the forecast

Commercial targets still belong in S&OP and IBP. In fact, the difference between expected demand and commercial ambition can create one of the most useful discussions in the process. If the statistical and market-informed demand forecast says 100 while the business wants to sell 110, that ten-unit gap tells management something important.

Changing the demand forecast to 110 because the target is 110 removes that information. Keeping the two numbers separate allows the conversation to move toward the actions required to close the gap, such as additional promotions, pricing changes, distribution expansion or commercial activity.

The statistical forecast therefore does not need to win every argument in the demand review. Its value is that it provides a neutral reference point before organizational incentives enter the process. Human judgment can then focus on adding information the model could not reasonably know, while the planning team retains enough transparency to determine afterward whether that judgment actually improved the forecast.

That is a normal part of demand planning. Statistical forecasts are built from the information available to the model, while planners often have context that has not yet appeared in the data. The question is therefore not whether planners should override forecasts, but whether those adjustments consistently make the forecast better.

Planner judgment is most valuable when it adds new information

There are many situations where an override makes sense. A planner may know that a large customer has confirmed a promotion, a product will soon be delisted, or an unusual order should not be treated as recurring demand. If that information is not available to the forecasting model, the planner is adding something genuinely new to the forecast.

The distinction becomes less clear when the adjustment is based primarily on judgment that a number seems too high or too low. Commercial optimism, pressure to meet a target, or discomfort with a statistical forecast can all influence the final number without necessarily adding better information.

Research suggests this distinction matters. A study covering more than 60,000 forecasts across four supply chain companies found that judgmental adjustments improved forecast accuracy on average in three of the four companies. However, smaller adjustments frequently reduced accuracy, while larger adjustments were more likely to improve it. Upward adjustments were also less successful than downward adjustments on average. Read the original study in the International Journal of Forecasting.

A more recent analysis covering approximately 147,000 forecasts found a similar pattern. Judgmental adjustments improved forecast performance for only slightly more than half of the SKUs studied, with considerable variation between datasets. In other words, human input can add value, but the act of changing the forecast is not evidence that value was added.

A simple question can improve forecast reviews

One practical way to make overrides more disciplined is to ask what information the adjustment introduces that was not already available to the forecasting model.

A confirmed customer promotion provides new information. So does a planned product discontinuation, a lost customer, a known one-off order, or another business event that has not yet appeared in the historical demand pattern. An adjustment based mainly on the belief that demand “should be higher” is much harder to evaluate.

This does not mean every override needs a lengthy approval process. It means the reason behind the change should be visible enough that the organization can later determine whether that type of intervention actually improved the forecast.

That becomes especially useful when the same behavior appears repeatedly. If planners consistently increase forecasts for a certain category and those changes create upward bias, the process may need to change. If overrides around customer promotions consistently improve accuracy, that suggests the planners have access to information that should perhaps become a more structured input to the forecasting process.

Measure whether the planning process adds value

The most useful way to settle the question is to measure the effect of the adjustment. Planning teams can compare the statistical forecast with the final forecast using the same horizon, level of aggregation and accuracy metric. This is the principle behind Forecast Value Added, or FVA: measuring whether each step in the forecasting process improves the forecast that entered that step.

The objective is not to reduce planner involvement. It is to understand where planner judgment performs well and where additional intervention creates noise. Over time, that allows the planning team to spend more attention on situations where human context genuinely matters while relying more confidently on the statistical forecast where manual changes have historically added little value.

It can also reveal opportunities to improve the forecasting model itself. If planners repeatedly make the same type of correction, the underlying issue may be a missing input rather than a need for permanent manual intervention.

Good demand planning therefore depends on both statistical forecasting and human judgment. The strongest process is one that understands where each source of information performs well, measures the result, and learns from the difference.

For Scyplan, that principle is important: planner adjustments should remain transparent and measurable so that teams can understand whether an intervention improved the forecast and why.

Planning, Simplified.

Modern planning software should do more of the analysis in the background and bring the most relevant issues to the surface. Instead of asking planners to search for problems, the system should help identify what deserves attention, explain why it matters, and provide the context needed to decide what to do next.

More dashboards are not the answer

Most planning platforms already contain more than enough information. Planners can access forecast accuracy, bias, service levels, inventory, historical sales, promotions, customer forecasts, and many other metrics across thousands of products and customers, but access to more information does not necessarily make decisions easier.

A company with 5,000 SKUs should not require a planner to manually review 5,000 forecasts every week. The real objective is to identify the relatively small number of products, customers, or locations where something meaningful has changed, which means planning software should become better at prioritizing information rather than simply displaying more of it.

Exception management needs to become more intelligent

Exception-based planning is not new, but traditional alerts can still create considerable noise. If a system generates hundreds of alerts every morning, the planner is effectively given another list to investigate, and fixed thresholds do not always distinguish between something that is statistically unusual and something that is commercially important.

A 20% forecast change on a low-volume product may have little impact, while a 5% change on a critical high-volume SKU could create a serious inventory or service risk. By considering factors such as volume, margin, inventory exposure, service levels, and historical behavior, AI can help prioritize exceptions based on business impact rather than deviation alone.

The analysis can happen before the planner starts investigating

A more intelligent system could evaluate the latest demand, forecast, inventory, and performance data before the planner begins reviewing it. Instead of opening several reports to search for potential problems, the planner could start with a short list of issues that already include the most relevant context.

The system might identify a sudden increase in demand, show which customer or region is driving it, compare the pattern with previous periods, and highlight the potential impact on inventory. The planner can then focus on determining whether the change is temporary, commercially expected, or something that requires an adjustment, while much of the repetitive investigation has already been completed.

From identifying problems to supporting decisions

The next step is moving from explanation toward action. If the system detects a potential stockout, showing a red warning symbol should only be the starting point because the planner also needs to understand what is causing the risk, when it is likely to occur, and what could change the outcome.

The system might identify an unusual customer order pattern, highlight an inventory imbalance, or show how different forecast adjustments would affect the projected position. The planner should remain in control because supply chain decisions often depend on commercial context and information that is not yet visible in the data, so the role of AI is to prepare the decision rather than replace it.

Human judgment becomes more valuable

As software becomes better at processing data and identifying patterns, planners can spend more time on the areas where human judgment creates the most value. They may know about an upcoming promotion, a customer discussion, a supply constraint, or an unusual event that a model cannot yet see, and that context can fundamentally change the right planning decision.

This is particularly valuable for mid-market companies, where planning teams are often relatively small. The objective should not be to give those teams more data to analyze, but to help them focus on fewer, more relevant decisions and spend more time on the situations where their experience actually matters.

Planning software should help answer “what next?”

Planning systems have traditionally been good at answering what happened, what the forecast is, and where performance deviated from plan. Modern systems need to become better at answering the questions that follow: what deserves attention, why does it matter, and what should the planner consider doing next?

For us at Scyplan, that is an important part of what modern planning technology should deliver. More of the analytical work should happen in the background so planners can spend more of their time applying judgment and making better decisions.

Planning, Simplified.

That creates an odd contradiction: the technology becomes more intelligent, while much of its complexity is still passed on to the user. I believe modern planning software should move in the opposite direction. As the technology becomes more sophisticated, the experience for the planner should become simpler.

Complexity is not the same as capability

Enterprise software is often designed to support a broad range of industries, planning processes, hierarchies, and business rules. That flexibility can be valuable, but it also tends to create systems with extensive configuration options, multiple layers of navigation, and complex workflows.

The problem is not that the software is capable of doing a lot. The problem is when the planner needs to understand too much of that complexity to get value from it.

A forecasting engine may evaluate different statistical models, account for seasonality, detect anomalies, recognize changing demand patterns, and continuously compare forecast performance. These are valuable capabilities, but the planner should not have to manage every underlying parameter to benefit from them. The software should absorb as much of that complexity as possible and surface the result in a way that supports the decision that needs to be made.

Planners should understand planning, not software mechanics

Demand planning is already a complex role. Planners need to understand customers, products, promotions, market changes, historical patterns, and commercial input, while coordinating with sales, finance, operations, and supply. The software should make that work easier rather than introducing another layer of expertise that needs to be mastered.

In many organizations, experienced users become highly skilled at navigating the system itself. They know which reports to open, which sequence of steps is required, where certain information is hidden, and which workaround is needed when the standard process does not quite fit. Over time, this becomes accepted as normal, even though it often creates long training periods and dependency on a relatively small group of experienced users.

A planner who understands the business and the planning process should be able to become productive relatively quickly. The software should reflect how planners think about their work rather than requiring planners to understand how the software is structured.

Sophistication belongs behind the interface

Making planning software easier to use does not mean making the underlying technology less advanced. In fact, the opposite should be possible.

The backend can contain sophisticated forecasting algorithms, machine learning, optimization, anomaly detection, and large-scale data processing while the user experience remains simple. What matters to the planner is not necessarily which model produced a forecast, but whether the system can help answer practical questions such as which products require attention, why forecast accuracy has deteriorated, where bias is increasing, or what changed materially enough to require action.

There can be significant technical sophistication behind those answers, but the planner should not have to reconstruct that analysis manually. Good software should make advanced capabilities accessible without forcing the user to interact with all of the complexity behind them.

AI can make advanced planning more accessible

This is where AI becomes particularly interesting. Traditional planning systems often require users to understand where information is located before they can find an answer. A planner needs to know which report to open, which filters to apply, which level of the hierarchy to select, and which metrics to compare.

AI can make that interaction much more direct. A planner should increasingly be able to ask why forecast accuracy declined last month, which products contributed most to the change, whether a particular customer is creating unusual demand, or what the inventory impact would be if a forecast were adjusted.

The question itself may be simple, while answering it requires the system to work across multiple datasets, time periods, hierarchies, and planning metrics. That is where AI can add real value: not simply by adding another feature, but by making sophisticated capabilities easier to access.

The same principle applies to exception management. Instead of asking planners to continuously search through reports and dashboards to determine where attention is needed, the system can analyze the planning environment in the background and surface the issues that matter most. This allows planners to spend less time finding problems and more time understanding context, challenging assumptions, and deciding what to do next.

Simplicity still requires transparency

A simpler user experience should not mean creating a black box. Planners still need to understand why an exception has been highlighted, what information contributed to a forecast, and what assumptions sit behind a recommendation.

This is especially important because supply chain decisions cannot be made from historical data alone. A system may identify an unusual pattern, while the planner knows that a promotion is about to start, a major customer has changed its strategy, or a market event has occurred that is not yet reflected in the data.

The technology brings analytical scale and consistency, while the planner brings context and judgment. Good planning software should make it easier to combine the two.

For us at Scyplan, that is an important design principle. Forecasting, machine learning, and AI can continue to become more sophisticated in the background, while the everyday experience for the planner becomes simpler and more intuitive.

The goal is not to reduce what planning software can do. It is to make advanced planning easier to use.

Planning, Simplified.

The challenge is that the same philosophy is often applied to mid-market companies, even though their reality is very different.

Too often, a mid-market business ends up buying what is essentially a scaled-down version of an enterprise planning platform. The system may be highly capable, but it often comes with the same underlying complexity: long implementation timelines, heavy configuration requirements, extensive training, significant consulting support, and functionality that many planners will never use.

For a large enterprise, that may be acceptable. For a mid-market company, it can quickly become a burden.

More capability does not always create more value

Enterprise software is typically designed to accommodate as many use cases as possible. From a product perspective, that makes sense. Different customers need different workflows, different planning methodologies, different hierarchies, and different levels of customization.

The downside is that this flexibility often becomes visible to the end user.

Planners may be confronted with large menus, complex settings, transaction IDs, configuration screens, and processes that require extensive training before they become second nature. Internally, we sometimes joke about “Chernobyl interfaces”: systems that look more like a control room than a modern business application.

That may sound harsh, but the underlying point is important. A planning system can be incredibly powerful while still being unnecessarily difficult to use.

In many organizations, planners ultimately use only a fraction of the available functionality. The rest remains in the system because it might be needed by someone, somewhere, at some point. Yet that unused complexity still has a cost. It makes implementation more difficult, increases training requirements, slows adoption, and creates dependency on experts who understand how the system is configured.

For mid-market companies, the question should therefore not be how much functionality a system has. The better question is how much value planners can actually access from it.

Implementation should not become a multi-year transformation

The same issue applies to implementation.

Large enterprises sometimes have the resources to support long transformation programs involving internal project teams, consulting firms, dedicated IT resources, process redesign, integration work, change management, and months of testing. These projects can be justified when the organization is large enough and the expected benefits are significant enough.

Most mid-market companies operate differently. Their planning teams are smaller, their IT resources are more limited, and people often have broader responsibilities. They cannot always afford to dedicate a large team to a planning transformation for a year or more before the software begins to deliver meaningful value.

This makes time to value much more important.

A mid-market planning solution should be designed with that constraint in mind. Integrations should be simpler, configuration should be more intuitive, and the system should be able to evolve as the company grows without every change becoming another implementation project.

The goal should not be to simplify the business until it fits the software. The software should be flexible enough to support the business without requiring unnecessary complexity around it.

Sophistication belongs in the backend

None of this means that mid-market companies need less sophisticated planning capabilities.

In fact, the opposite is increasingly true.

Advances in statistical forecasting, machine learning, optimization, and AI make it possible to perform analysis that would have required significant specialist resources only a few years ago. Mid-market companies should be able to benefit from those capabilities just as much as larger enterprises.

The important distinction is where that sophistication sits.

A planning system can have PhD-level sophistication in the backend without requiring PhD-level knowledge from the person using it. The mathematical models, algorithms, anomaly detection, forecasting methods, and optimization logic can all be advanced while the user experience remains simple and intuitive.

The planner should not need to understand every method being used behind the scenes. What matters is whether the system can help answer practical questions such as: What changed? Which products require attention? Where is forecast bias increasing? Why did performance deteriorate? What should I look at next?

The technical capability already exists to answer many of these questions. The challenge is making that capability accessible to the planner.

AI can change how planners interact with software

This is where AI becomes particularly interesting.

Traditional planning systems have largely been built around the idea that the system provides information and the planner performs the analysis. A planner opens a dashboard, identifies an exception, moves into another screen, investigates historical data, compares forecasts, looks for possible drivers, and eventually decides whether action is required.

That workflow puts a large analytical burden on the planner.

AI creates an opportunity to reverse part of that relationship. Instead of asking planners to continuously search the system for issues, the system can increasingly perform that analysis in the background and bring the most relevant exceptions to the planner.

It can identify unusual behavior, prioritize issues by business impact, investigate likely causes, summarize changes, and suggest where further attention is needed. The planner can then focus on the part where human judgment adds the most value: understanding context, challenging assumptions, collaborating with the business, and making decisions.

This does not mean replacing planners. It means changing how their time is spent.

A planner should not have to spend the first half of the day finding out what went wrong before being able to think about what to do about it.

From analysis-driven to action-driven planning

For me, this is one of the most important changes planning software needs to make.

Many systems are still designed around providing more information. More dashboards, more reports, more metrics, and more ways to slice the data. While those capabilities are useful, they do not necessarily make the planning process better.

The real objective should be to help planners move from information to action more quickly.

That means identifying what matters, explaining why it matters, and making it easier to decide what happens next. It also means recognizing that the planner does not need to investigate every SKU, every forecast, or every exception. The system should help prioritize attention where it will have the greatest impact.

This is especially relevant in the mid-market, where planning teams are often lean. A smaller team cannot spend hours every week navigating through thousands of data points to identify the handful of issues that genuinely require intervention.

Planning software should do more of that work for them.

The mid-market needs a different design philosophy

Mid-market companies should not have to choose between spreadsheets on one side and highly complex enterprise platforms on the other.

There is room for a different generation of planning software: software that is sophisticated enough to handle complex supply chain problems but designed around the realities of smaller planning organizations. That means faster implementation, lower complexity, more intuitive workflows, easier access to advanced capabilities, and a greater focus on action.

The future of planning software is not about making planning less sophisticated. It is about making sophisticated planning easier to use.

That is the philosophy behind Scyplan.

Planning, Simplified.