How To Interview An S&OP Planner When The Software Already Wrote The Plan

Start The Interview With One Exception The Planner Actually Owned

Open with a week, not a competency. "Take me to the last Monday your planning system escalated something you disagreed with." Then stay in that week for twenty minutes. The candidate who owns exceptions gives you a supplier name, a line, a date, a person they called, and a number that moved. The candidate who owns dashboards gives you a process description.

This works because the escalation is the only part of the job the tooling hands back. The clearest published account of that arrangement comes from the vendor side: an SAP implementation partner, writing about the 2026 supply chain stack, describes agent modules that evaluate dozens of alternatives during a live crisis call and escalate to a human only when a proposed revision breaches a strategic boundary ¹. Read that as a vendor describing its own product rather than as a measurement of the industry, and then go and check it against the configuration in front of you, because the boundary is a setting somebody chose. Everything below it happens without the planner. Everything above it is the hire.

So the questions that used to sort candidates no longer sort anything. Asking someone to explain safety stock calculation, or to talk through a forecast error metric, tests the layer the vendor already sells. Ask instead about the exception queue: how many items hit it in a normal week, which ones they closed without escalating further, and which single one they still think they got wrong. That last question does more work than the rest of the loop.

Push on the boundary itself, too. A planner who has been in the seat for a year has an opinion about where the escalation threshold is set wrong, because they have been woken up by noise or blindsided by something that never reached them. That opinion is unfakeable. It comes from having lived inside a specific configuration, and a candidate who has only read about exception-based planning cannot produce one on demand.

What Separates A Real Scenario Planner From A Performed One?

Three tells. A real one frames the scenario before running it, and can say which question the run was supposed to answer. A real one names the input they distrusted and what they did about it. A real one describes disagreeing with a recommendation in front of someone senior, and remembers how it went, including the times they lost.

Framing first. Ask what scenarios they ran during the last serious disruption and, more importantly, what they chose not to run. When simulation is nearly free, the scarce skill is deciding which futures are worth pricing. A planner who lists eight scenarios is describing a tool's capability. A planner who says they ran three because the other five collapsed into the same decision is describing judgment. That habit of framing before generating is the same one that separates capable AI work in any role that now supervises a model.

The distrusted input is the second tell. Every planning engine sits on a demand signal, a lead time table, and a capacity assumption, and at least one of those is stale in every company. Good candidates volunteer which one they never trusted and how they checked it: a call to a plant scheduler, a look at actual receipts against the master data, a text to a supplier's account manager. Watch for people who verified against something outside the system rather than a second report drawn from the same system.

The third tell is disagreement with a record. "Tell me about a time the recommended plan was defensible and you argued against it." You want the operating fact they had that the model did not: a qualification run that was not in the routing, a customer whose commitment was verbal, a line that had been down twice that month. Then ask what they did when the director wanted the number anyway. Candidates who have never been overruled have usually never pushed.

One anti-tell worth naming. Fluency about AI is not evidence of using it well. Someone who talks about agentic planning in the abstract, with no configuration they have argued with and no recommendation they have rejected, is describing a conference talk. The signal is friction with a specific system, not enthusiasm about the category.

Which Backgrounds Actually Produce Exception-Grade S&OP Planners?

The obvious feeders are demand planning, supply planning, materials management and plant scheduling, and they work. The better-than-expected ones are people who have carried a cross-functional commitment: revenue operations analysts who owned a quota-to-supply handshake, category buyers who negotiated allocation during a shortage, and production supervisors who ran the floor when the plan was wrong at 6 a.m.

What those unexpected backgrounds share is arbitration under a real constraint. S&OP fails on alignment far more often than on math, and BCG's 2026 supply chain planning work argues that AI on its own does not deliver the planning outcome, because judgment and organizational alignment remain the human layer ². A person who has already made sales and operations agree on one painful tradeoff has practiced the actual job.

Some credible non-obvious sources: military logistics officers, who plan against adversarial uncertainty and are trained to state assumptions out loud; ERP and APS implementation consultants at firms that deploy SAP IBP, Kinaxis, o9 or Blue Yonder, who have seen twenty configurations instead of one; clinical or lab operations coordinators, whose scheduling problems are constraint-dense and unforgiving. Consultants come with a caveat worth probing: implementation experience is not the same as living with the consequences of a plan for four quarters, so ask what they would have configured differently after seeing it run.

What matters less than it used to: years of Excel modeling depth, and formal APICS-style certification. Both still signal seriousness and neither predicts exception judgment. Weight them below evidence that the candidate has changed a decision that someone else did not want changed. The same reweighting is happening in editorial and review roles where the draft arrives finished.

Where To Source S&OP Planners, And What It Costs To Hire One

Look where planners argue about configurations rather than where they post resumes. The vendor ecosystems are the densest pools: SAP IBP, Kinaxis, o9 and Blue Yonder user communities and their annual conferences, plus the implementation partners around them. ASCM and its local chapters, the Institute of Business Forecasting's S&OP events, and university supply chain programs at Michigan State, Penn State and Arizona State are reliable adjacent sources.

Feeder companies behave predictably. Consumer packaged goods and pharmaceutical manufacturers train planners in high-cadence S&OP with real executive attention, and their alumni transfer well into industrials and medical devices. Contract manufacturers produce planners who are used to constraints they do not control. If you are hiring your first planner into a company that has never run S&OP, prefer someone from a smaller planning team, because a specialist from a fifty-person planning organization may never have owned the whole cycle.

On compensation, be honest with yourself about what is knowable. No published salary series exists for an AI-era exception-and-scenario S&OP title, because the title is not stable enough to have been surveyed. What you can defend is a proxy, said out loud as one: hire against your senior individual-contributor planning band, because that is where postings for senior S&OP and integrated business planning managers already cluster, and because the exception-and-scenario work is a rescoping of that seat rather than an invention of a new one. Then, as of mid-2026, pull the current figure yourself from a named source with a date on it, such as the Bureau of Labor Statistics occupational wage series for logisticians, a Lightcast or Dice compensation report, or aggregated postings in your own metro. Quote that number in the loop rather than one from an article, this one included.

Two things reliably move the band upward: multi-node global planning with real allocation authority, and ownership of the executive S&OP meeting itself rather than the pre-work behind it. If a candidate has run the meeting where the tradeoffs got decided, you are competing for them against companies that know exactly what that is worth.

How Do You Close An S&OP Planner, And Should The Role Be On Site?

Planners take the offer where the decision rights are real. The single strongest close is a sentence you can prove: name the meeting they will own, the tradeoffs they will arbitrate, and who has to accept their recommendation. Money matters and rarely decides it. What kills offers is discovering during onboarding that the planner is a report producer for a director who makes every call.

The other reliable offer-killer is master data. Planners who have worked in a system with rotten lead times and phantom inventory will ask about it, and if the answer is evasive they will leave the process. Say plainly what state the data is in and what budget exists to fix it. Candidates forgive a mess they were told about; they do not forgive one they discovered in month two.

On location, split the work. The scenario, modeling and exception-review parts are genuinely remote-capable and increasingly done that way, which is why distributed planning teams are common. The alignment part is not, or not entirely. Consensus meetings, plant walks, and the informal conversation with the plant manager that prevents a bad commitment all run better in a room. The pattern that holds up: remote or hybrid for the individual planner, with an expectation of regular presence at the site or the executive S&OP cycle, stated as a specific cadence rather than a vague "as needed."

One broader piece of context to put in the offer conversation, because good candidates are already thinking about it. In BCG's 2026 survey of 11,749 workers across 14 markets, 47 percent reported spending more time managing and directing AI than doing the work themselves ³. Planners are early to that shift, and the honest pitch is that the job is now supervision and arbitration. Say so. The people you want find that more attractive than the version where they run the numbers by hand.

Test The Judgment Before The Offer, Not After

Replace the case-study slide with a working session. Give the candidate a messy scenario, an assistant or planning tool that will confidently overreach, and ninety minutes. Then watch what they check, what they refuse to accept, and how they write up a recommendation they know is contestable. An interview captures a description of judgment; work captures judgment.

Build the exercise around something that is wrong on purpose. A lead time in the data that contradicts a note in the supplier email thread. A demand signal inflated by a one-time promotion. A capacity figure that assumes a line that has been down. The question is not whether the candidate finds all three. It is whether they check anything against a source outside the model before committing to a plan, and whether they say clearly which parts of their recommendation they are unsure about.

Score what actually happened in the session rather than the polish of the output. Did they frame the question before running scenarios? Did they demand a source for the claim that mattered most? Did they keep the judgment they should not have delegated, or did they accept a confident recommendation because it was confident? Write those observations down with the moment each one came from, and share the same write-up with the candidate. That kind of evidence trail is what makes a hiring debate about the work instead of about impressions, the same shift described in red-team and adversarial technical hiring.

Finally, run the debrief on evidence and not on comfort. The failure mode in planning hires is preferring the candidate who sounded most certain, which is precisely the trait the job punishes. The planner you want is the one who told you, unprompted, which part of their own plan they would check first.

See how it works >