Sales and Operations Planning: The Process

2026-09-20 · Tommaso Maria Ricci

Most companies that say they run a sales and operations planning process are running a monthly reporting meeting with a planning name on the calendar invite. The difference is simple and brutal: in a reporting meeting, people explain what happened. In a sales and operations planning process, people decide what happens next, with numbers attached and an owner's name on each decision. If nobody leaves the room having committed to something they could be measured on, you do not have S&OP. You have a slideshow.

I have sat in a lot of those rooms. The pattern repeats across industries and company sizes. Sales presents a forecast that is really a quota. Operations presents a capacity picture that is really a complaint. Finance presents a number that matches neither. Everyone agrees the numbers should be reconciled, nobody owns the reconciliation, and the meeting ends on time. Four weeks later the same three versions of reality show up again, aged by a month.

The cost of that is not theoretical. It shows up as inventory in the wrong place, expedited freight nobody budgeted, overtime approved on Thursday for an order promised on Monday, and a service level that the sales team quietly stops promising. None of those line items say "planning failure" in the general ledger. They say freight, labor, obsolescence, discount.

This guide is about the operating mechanism, not the software. It covers what the sales and operations planning process actually is, the five steps and who owns each one, the cadence and horizon that make it work, the failure modes that kill it, the metrics that prove it is working, and a ninety day sequence to stand it up in a company that has never run one.

What sales and operations planning is, and what it is not

S&OP is a monthly decision process that reconciles demand, supply, and money into one plan that the whole company executes. Three words in that sentence carry the weight.

Decision. The output is not a forecast. The output is a set of decisions: what we will build, what we will buy, what we will hold, what we will not promise, and where we will deliberately fall short.

Reconciles. The demand plan, the supply plan, and the financial plan must add up to the same story. If the demand plan says 12% growth and the budget says 6%, one of them is wrong and the meeting exists to name which.

Monthly. Not quarterly, because the business moves faster than that. Not weekly, because the strategic questions do not change that fast and the process collapses into firefighting. Weekly execution meetings are a different mechanism with a different purpose.

What S&OP is not, in practice:

It is not a forecast accuracy exercise. Accuracy matters as an input, not as the goal.
It is not a demand planning tool rollout. Software supports the process and never creates it.
It is not an operations meeting. If finance and sales leadership are not in the room, no decision made there is binding.
It is not a review of last month. The horizon is forward, typically eighteen to twenty four months rolling.

The last point is the one that gets violated most often. When the monthly meeting spends forty minutes on variance analysis of the previous period and ten minutes on the next twelve months, the company has built a rear view mirror and mounted it where the windshield should be.

The five steps of the sales and operations planning process, with owners

The classic structure has five steps that run as a monthly cycle. Each one needs a named owner, an input, an output, and a deadline in the calendar. Without owners, the cycle silently degrades into the one meeting at the end, which is exactly the failure mode described above.

Step 1: product and portfolio review

Owner: product management. What is launching, what is being phased out, what is being repriced, what is changing in packaging or specification. This step exists because new and dying products are where forecasts break. A launch with no volume history and a phase out with a last time buy both require judgment, not statistics.

Output: a dated list of portfolio changes with volume assumptions and cannibalization estimates, agreed with sales.

Step 2: demand review

Owner: commercial leadership, supported by demand planning. Start from a statistical baseline generated from history, then apply the judgment layer: promotions, tenders, customer specific news, competitive moves, price changes. The critical discipline is that every manual override to the statistical baseline must have a reason and a name attached.

Output: an unconstrained demand plan by product family, by month, for the full horizon, in units and in revenue. Unconstrained means it answers "what would customers buy if we could supply everything", not "what can we deliver".

Step 3: supply review

Owner: operations and supply chain. Take the unconstrained demand plan and test it against real capacity: production hours, key equipment, labor, supplier lead times, warehouse space, transport. Where it does not fit, produce options rather than a verdict. Adding a shift costs this much and delivers this. Air freight closes this gap at this cost. Pushing this customer out two weeks frees this capacity.

Output: a constrained supply plan, plus a written list of gaps with costed options for each.

Step 4: financial reconciliation

Owner: finance. Translate both plans into money and compare against the budget and the latest forecast. This is where the process earns its credibility, because it is the step that turns a debate about units into a debate about margin and cash. A volume plan that hits the number with a mix that destroys margin is not a plan that passed.

Output: a financial view of the constrained plan, the gap to budget, and the value of each option from step three.

Step 5: executive S&OP meeting

Owner: the general manager or CEO. Ninety minutes, maximum. The agenda is the gaps and the decisions, not the review. Everything that was already agreed in steps one through four is not re presented. The executive meeting exists to resolve what the functions could not resolve between themselves and to commit the company to one plan.

Output: a signed plan with decisions, owners, dates, and the explicit list of what the company has chosen not to do.

That last item is the most underrated output of the whole process. A plan that does not say no to anything is a wish list, and wish lists are executed by whoever shouts loudest, which in every company I have worked with is the same three people.

The rule that separates real S&OP from theater: one set of numbers

There is one test for whether a sales and operations planning process is real. Ask three people from three functions for the volume plan for a given product family three months out. If you get three different numbers, the process is theater, no matter how polished the deck is.

One set of numbers means:

The demand plan that sales signs is the same one operations schedules against.
The revenue in the financial forecast is derived from that same volume plan at the agreed price, not built separately from the top down.
Changes to the plan between cycles follow a rule, not a phone call.

That third point is where most companies leak. The monthly plan is agreed, and then over the following four weeks it is quietly amended by a hundred individual conversations, none of them recorded. By the time the next cycle starts, nobody can reconstruct why the plan and the reality diverged, so the conversation defaults to blame.

The fix is unglamorous: a change log. Every material change to the agreed plan between cycles gets one line, with date, size, reason, and who approved it. Companies resist this because it feels bureaucratic. Three months in, that log becomes the single most useful document in the process, because it shows the pattern. In most cases the pattern is that one channel or one customer generates the majority of the disruption, and that is a commercial problem with a commercial solution, not a planning problem.

Cadence, horizon, and time fences

The process runs monthly, but the plan does not treat all months the same. Three zones, with different rules.

| Zone | Typical window | Rule |

|---|---|---|

| Frozen | Weeks 0 to 2 to 4 | No changes except emergencies approved by a named executive |

| Slushy | Weeks 4 to 12 | Changes allowed within capacity, mix changes easier than volume changes |

| Liquid | Beyond 12 weeks | Fully open, this is where planning actually happens |

The frozen window length is not a philosophical choice. It is derived from your longest binding lead time: the slowest critical component, the supplier with the longest commitment, the equipment with the longest changeover. Setting a two week frozen zone when your key raw material has a ten week lead time is not agility. It is a decision to hold inventory or to miss dates, made by accident instead of on purpose.

Horizon should extend far enough to cover the longest decision you actually make. If capacity investments take eighteen months from approval to output, an eighteen month horizon is the minimum, otherwise the process can never surface a capacity decision early enough to act on it. Most mid sized companies land on eighteen to twenty four months rolling, reviewed monthly.

What each function has to bring, in writing

The cycle collapses when people arrive with opinions instead of inputs. These are the inputs each step requires, and each one should exist as a file before the meeting, not as a comment during it.

Commercial. Statistical baseline plus overrides with reasons. Pipeline for large deals with probability. Promotional calendar with expected uplift. Customer specific risks and opportunities above a materiality threshold.

Operations. Available capacity by constraint, not average capacity. Planned maintenance and shutdowns. Current and projected inventory by family. Supplier risk list with lead times.

Procurement. Material availability and price movement on the top spend categories, with the commitments already made. This is the natural link to the upstream work described in the guide on the strategic sourcing process, because sourcing decisions made a quarter earlier set the constraints this process has to plan around.

Finance. Latest view of revenue, margin and cash against budget. Cost assumptions. Working capital targets.

Product. Launch and phase out calendar with dates and volume assumptions.

A useful discipline: any input that arrives after the deadline is excluded from that cycle. Once, when I helped a distribution business put this in place, that rule alone changed behavior faster than any amount of process training. The first month, two functions missed the cutoff and had to watch the plan get agreed without their numbers. From the second month on, nobody missed it.

Seven failure modes, and what each one actually signals

One: the meeting is a review, not a decision forum. Signal: the pack is fifty slides and the discussion is about last month. Fix: cap the executive pack at the gaps, the options, and the decisions required. Move history to a pre read that nobody presents.

Two: sales submits quota, not forecast. Signal: the demand plan equals the target, month after month, and then misses. Fix: separate the two explicitly. The forecast is what we expect to happen. The quota is what we are asking people to achieve. Both are legitimate, they are just different numbers, and planning against the second one guarantees excess inventory.

Three: operations submits capacity as a complaint. Signal: "we cannot do it" without options or costs. Fix: require every constraint to come with at least two costed options. Constraints without options move the decision to the executive meeting, where it is made with less information.

Four: finance runs a parallel forecast. Signal: the company has two revenue numbers and everyone knows which one is "the real one". Fix: derive the financial forecast from the agreed volume plan. If finance disagrees with the volumes, that disagreement belongs in the demand review, not in a separate spreadsheet.

Five: no decision rights. Signal: decisions are deferred because nobody is sure who can make them. Fix: write the decision rights down. Who can approve overtime, expedited freight, a customer allocation, a price exception, a capacity investment, and up to what value.

Six: the plan does not survive contact with week one. Signal: within days of the meeting, the schedule is being changed by informal requests. Fix: the change log plus a named gatekeeper. Not to block changes, but to make their volume visible.

Seven: too much detail. Signal: the demand review is arguing about individual stock keeping units. Fix: plan at product family level, where the numbers are stable enough to be meaningful, and leave item level detail to the weekly execution process. S&OP answers how much capacity and cash we need. It does not answer what to build on Tuesday.

Failure mode two is the most expensive one in absolute terms, and the most common. It converts optimism directly into working capital, because the supply plan faithfully builds to a demand plan that was never a forecast in the first place. The link between that and cash tied up in stock is direct, and I have unpacked the inventory side of it in the guide on AI for inventory management.

The metrics that prove the process works

Six metrics, in causal order. The first three describe the quality of the plan. The last three describe what the plan produced.

| Metric | Definition | Starting target |

|---|---|---|

| Forecast accuracy | 1 minus weighted absolute error, at family level, at lag 1 and lag 3 | Set from your own baseline, then improve |

| Forecast bias | Signed average error over rolling 6 months | Within plus or minus 5% |

| Plan attainment | Actual output against the committed supply plan | Above 95% |

| Schedule adherence | Production or dispatch executed as scheduled | Above 90% |

| Inventory days of supply | Inventory value divided by average daily cost of sales | Down at constant service |

| On time in full | Orders delivered complete on the promised date | Up, measured against the original promise |

If you want definitions that will survive an argument with a consultant or a new operations director, standardize on the ones in the SCOR Digital Standard maintained by ASCM rather than inventing your own. The value is not the sophistication of the model, it is that nobody gets to redefine plan attainment halfway through a bad quarter.

Two practical notes. First, measure accuracy at more than one lag. Accuracy at lag 1, one month ahead, tells you almost nothing useful, because at that horizon most of the plan is already committed. Accuracy at lag 3 is where planning decisions actually live.

Second, bias matters more than accuracy for most companies starting out. A process that is consistently 12% over is easier to fix than one that is randomly wrong by 12%, and it is doing more damage, because every month it adds inventory that nobody decided to buy. Bias is a behavioral measurement disguised as a statistical one: persistent positive bias usually means the demand plan is a quota, and persistent negative bias in one region usually means somebody is sandbagging to protect a bonus.

On time in full deserves one qualifier. Measure it against the date first promised to the customer, not against the date after it was revised. Most reporting quietly uses the revised date, which produces a number that looks respectable while customers experience something else entirely.

Maturity: four honest stages

Companies rarely go from nothing to integrated business planning. They pass through stages, and it helps to know which one you are in, because trying to install stage four behavior in a stage one company is how these programs die.

Stage one, reactive. No formal cycle. Planning happens in the weekly operations meeting and by email. Forecast is whatever sales said last. Inventory is the shock absorber for everything.

Stage two, functional. Each function plans its own world competently. Demand planning exists. Operations schedules properly. But the plans are reconciled only when they collide, and the collision is usually visible to customers first.

Stage three, integrated. The five step cycle runs monthly with owners and deadlines. One set of numbers. Decisions are made and logged. This is where most well run mid sized companies should aim, and it is achievable within two to three quarters.

Stage four, integrated business planning. The cycle drives the financial plan and the strategic plan, scenarios are run routinely rather than for crises, and the annual budget becomes a rolling output rather than an event. Powerful, and genuinely hard, because it requires the finance calendar to bend.

Trying to jump from stage one to stage three in a single quarter is possible only if the executive sponsor treats the calendar as non negotiable for the first six cycles. If the meeting slips twice, the process is dead and reinstalling it is harder the second time, because everyone now has evidence that it was optional.

S&OP and integrated business planning: the difference that matters

The vocabulary debate is mostly noise, but one distinction is worth holding. S&OP reconciles the volume plan across demand and supply and expresses it in money. Integrated business planning starts from the financial and strategic objectives and uses the same cycle to test whether the operating plan will deliver them, including the projects, the pricing actions, and the investments.

The practical implication for a company standing this up now: build the S&OP cycle first, with clean inputs and real decisions, and do not promise the board integrated business planning in year one. The failure I see most often in ambitious programs is scope, not capability. Six good cycles of basic S&OP beats one quarter of sophisticated design that nobody executes.

Tools: when a spreadsheet is enough, and when it stops being enough

A spreadsheet based S&OP process is not a sign of immaturity. Plenty of companies with a few hundred million in revenue run a perfectly sound cycle in a workbook, because the constraint is the decision discipline, not the calculation.

The spreadsheet stops being enough when one of these becomes true:

The demand plan takes more than two working days to assemble, so it is always stale by the time it is discussed.
More than one person needs to edit the plan at the same time, and versions start diverging.
You need scenarios. Comparing three supply options across a twelve month horizon is where workbooks break.
Product families exceed roughly fifty, or the number of planning combinations makes manual overrides untraceable.
You cannot reconstruct who changed what, which makes the change log impossible.

When you do move to a planning system, the sequence matters. Install the process first and the tool second. A planning system deployed onto an undefined process automates the confusion and adds a licence fee, which is the same lesson that shows up in every operational software decision, including the one covered in the guide on AI supply chain optimization.

What AI genuinely adds to planning in 2026, and what it does not

Every planning vendor now leads with artificial intelligence. Some of it is real and some of it is a rebranded regression that has been in the product for a decade. Three distinctions worth making.

Where it works. Statistical baselines built from more signals than a human planner can weigh, including weather, promotions, price, and competitor activity. Anomaly detection on the plan, flagging the families where the override pattern looks unusual. Automatic classification of demand variability so that the process focuses judgment on the items that need it. Scenario generation, where the value is speed rather than intelligence.

Where it helps only after the basics. Forecast improvement requires clean history: consistent product hierarchies, sales cleaned of stock out periods, promotions flagged. Most companies that want better forecasts have a data problem, not a model problem, and the honest sequence is to fix the hierarchy and the history first. I have gone deeper on that sequence in the guide on AI for demand forecasting.

Where it does not help. No model resolves the conflict between a sales leader's quota and an operations leader's capacity. That conflict is the entire point of the process, and it is resolved by a person with authority in a room, on a date, in front of the numbers.

The adoption data supports the sequencing argument. In the 2026 MHI and Deloitte Annual Industry Report, based on responses from more than 500 supply chain leaders collected in December 2025, 41% of respondents said their company is currently using artificial intelligence, up from 30% the year before. Growing fast, but still a minority, and the companies getting value are overwhelmingly those that already had a functioning planning cycle to plug it into.

The expectations are running ahead of that. Deloitte's 2026 Retail Industry Global Outlook, based on a survey of 330 global retail executives conducted between October and November 2025, found that 30% currently use artificial intelligence for supply chain visibility with 41% expecting to within a year, and 59% anticipate positive returns from AI driven supply chain initiatives within twelve months. Twelve months is an aggressive horizon for a technology that depends on data most companies have not cleaned yet.

A ninety day sequence to stand up the process

Days 1 to 30: define and baseline

Week 1. Measure the baseline honestly. Forecast accuracy and bias at family level for the last six months, plan attainment, inventory days of supply, on time in full against the original promise. If any of these cannot be produced from existing systems, that is itself the first finding, and it usually explains a lot.

Week 2. Define the planning hierarchy. Product families at the level where volumes are stable, typically twenty to eighty families for a mid sized business. This single decision determines whether the process will be arguable or useful.

Week 3. Write the calendar and the owners. Five steps, five dates, five names, repeating monthly, published for six months ahead. Write the decision rights document, one page, approved by the general manager.

Week 4. Define the input templates. One per function, no more than a page each. Agree the materiality thresholds that decide what gets discussed and what does not.

Days 31 to 60: run two cycles badly on purpose

Run the full five step cycle twice. It will be rough, inputs will arrive late, and the first executive meeting will drift into a review. That is expected and it is not a reason to redesign anything yet.

Two disciplines make these cycles worth running. Hold every date, even when the inputs are incomplete, because moving a date once teaches everyone that the calendar is negotiable. And log every decision, with owner and date, including the decisions to defer.

At the end of the second cycle, run a thirty minute retrospective on the process, not on the numbers. What arrived late, what was argued twice, what nobody used.

Days 61 to 90: tighten and prove

Third cycle with corrected templates and a disciplined executive agenda: gaps, options, decisions, and nothing else. Introduce the change log between cycles. Start reporting the six metrics as a standing one page scorecard.

At day ninety, compare against the week one baseline. The realistic result after three cycles is not a dramatic accuracy improvement, because three data points do not move a statistic. What should have changed is visible: decisions are being made on a date instead of drifting, the number of informal plan changes is measurable for the first time, and bias has a name and an owner.

If none of that has changed, the cause is almost always sponsorship. When the executive who chairs the final meeting sends a delegate twice, the organization correctly concludes that the process does not matter, and no amount of template quality recovers that.

If you are standing this up now and want a second opinion on whether the design fits your business before you burn six cycles finding out, describe how your demand and supply decisions are made today and I will tell you where I would look first. In most companies the binding constraint is not the planning method, it is a decision right that was never assigned to anyone.

What this looks like when it works

A sports distribution business I worked with had the classic version of the problem. Demand was seasonal and promotion driven, the sales team forecast in the language of targets, and operations absorbed the resulting variability with inventory and expedited shipping. The commercial side had also started using AI for marketing, which lifted sales by about 30%, and that success made the planning problem worse rather than better, because the demand signal became more volatile and the planning process was not built to absorb it.

The fix was not a planning system. It was a monthly cycle with owners, a demand plan separated from the quota, a frozen window derived from actual supplier lead times, and a change log that made visible how many mid month promises were being made and by whom. The change log was the part nobody wanted and the part that produced the biggest shift, because it turned an interpersonal argument into a countable pattern.

The general lesson from that project applies broadly. Companies reach for a tool when what is missing is a decision rhythm. The rhythm is cheap to install and hard to sustain. The tool is expensive to install and does nothing for the sustaining part.

Eighteen questions to ask your own team

Run these in a single session with the leadership group. The pattern of hesitation tells you more than the answers.

What is our forecast accuracy at family level, at lag 3, over the last six months?
What is our bias, and in which direction?
Is our demand plan a forecast or a target, honestly?
Who owns the demand number when it is wrong?
How long is our frozen window, and what lead time is it derived from?
How many changes were made to last month's agreed plan after it was agreed?
Who approved those changes?
Do sales, operations and finance work from the same volume number?
If not, which number does the factory or the warehouse actually use?
What decisions can operations make without escalation, and up to what value?
What is our plan attainment, and do we measure it?
Do we measure on time in full against the original promise or the revised one?
How many product families do we plan at, and are they stable?
What is the longest lead time decision we make, and does our horizon cover it?
Who chairs the executive planning meeting, and how many times have they missed it?
What did we explicitly decide not to do last cycle?
What does a stock out actually cost us, in a number?
If demand came in 20% above plan next quarter, what would break first?

Question sixteen is the one that produces the longest silence. A planning process that never says no is not allocating anything, and allocation is the whole job. Question eighteen is the one that produces the most useful answer, because the first thing that breaks is almost never the thing people expect, and knowing it changes where you invest.

Where the process connects to the rest of the operating system

S&OP does not live alone. Upstream, the sourcing strategy determines which suppliers can flex and at what cost, which sets the shape of every supply option the process can consider. Downstream, the purchase to pay chain executes what the plan commits to, and its control points determine whether the plan translates into actual material on actual dates. That execution layer is covered in the guide on the procure to pay process and its controls.

The practical implication is a sequencing one. If purchase orders are issued late and retroactively, the supply plan is fiction regardless of how good the demand plan is. If sourcing has single sourced a critical component with a twelve week lead time, the frozen window is twelve weeks whether the planning calendar says so or not. Planning quality is bounded by the execution discipline around it, which is why a company that fixes only the planning meeting often sees the gains evaporate by the second quarter.

If your planning cycle and your execution processes are currently owned by people who do not meet, that gap is worth a structured conversation before the next budget cycle rather than after it.

FAQ

What are the sales and operations planning process steps and owners?

Five steps, each with a named owner and a fixed date in the monthly calendar. Product and portfolio review, owned by product management, covering launches, phase outs and repricing. Demand review, owned by commercial leadership, producing an unconstrained demand plan with every override justified. Supply review, owned by operations, testing that plan against real capacity and producing costed options for each gap. Financial reconciliation, owned by finance, converting both plans into margin and cash and comparing against budget. Executive S&OP, chaired by the general manager, resolving the remaining gaps and committing the company to one plan with decisions, owners and dates.

How often should S&OP run, and how far ahead should it look?

Monthly, on a horizon of eighteen to twenty four months rolling. Monthly is fast enough to react to real business change and slow enough to keep the conversation strategic. The horizon should cover your longest binding decision: if a capacity investment takes eighteen months from approval to output, a twelve month horizon means those decisions can never be surfaced in time. Weekly execution meetings are a separate mechanism with a different agenda, and merging the two turns the planning cycle into firefighting.

What is the difference between S&OP and integrated business planning?

S&OP reconciles the demand plan, the supply plan and the financial view into one operating plan, primarily around volume and its financial translation. Integrated business planning starts from the financial and strategic objectives and uses the same monthly cycle to test whether the operating plan, the project portfolio and the commercial actions will deliver them, usually replacing the annual budget with a rolling forecast. The distinction matters less than the sequence: build a working S&OP cycle first, because integrated business planning built on an unreliable cycle simply distributes the unreliability to more people.

Do we need planning software to run S&OP?

No. Many companies with a few hundred million in revenue run a sound cycle in a spreadsheet, because the binding constraint is decision discipline, not calculation. Move to a planning system when the demand plan takes more than two days to assemble, when multiple people need to edit simultaneously, when scenario comparison becomes routine, or when overrides can no longer be traced to a person and a reason. Install the process first and the tool second: a planning system deployed onto an undefined process automates the confusion and adds a licence fee.

Which metrics show that the planning process is working?

Six, in causal order: forecast accuracy at family level measured at lag 3, forecast bias over a rolling six months, plan attainment against the committed supply plan, schedule adherence, inventory days of supply, and on time in full measured against the date first promised to the customer. In the first year, bias matters more than accuracy, because persistent bias is behavioral and fixable, and every month of positive bias adds inventory nobody decided to buy. A scorecard of six numbers on one page beats a planning dashboard nobody opens.

How long does it take to implement sales and operations planning?

Ninety days to run three full cycles, with the first two deliberately imperfect. Thirty days to define the product family hierarchy, the calendar, the owners, the decision rights and the input templates. Thirty days to run two cycles while holding every date, even with incomplete inputs. Thirty days to tighten the executive agenda, introduce the change log and report the metrics. Maturity, meaning a cycle people trust and plan their own work around, typically takes six to nine cycles. The variable that decides the outcome is whether the executive sponsor chairs the final meeting personally every single month.

What is the most common reason S&OP fails?

The demand plan is a quota rather than a forecast. When the number submitted equals the target every month, the supply plan faithfully builds to an expectation that was never realistic, and the gap shows up as excess inventory, expedited freight and discounting at quarter end. The second most common reason is the absence of written decision rights, which pushes every operational choice to the executive meeting and makes the cycle a bottleneck rather than an accelerator. Both are governance failures, not planning failures, which is why buying software rarely fixes either.