What Is an Inventory Planner?
An inventory planner is software that forecasts how much of each SKU you will sell over a coming period, then converts that forecast into a recommended order quantity once supplier lead time and existing stock are accounted for. It sits between your sales data and your purchasing decision. Instead of a buyer eyeballing last year’s numbers in a spreadsheet, the planner runs the same calculation on every SKU, every day, and flags the ones that need attention.
For a brand doing $3M to $30M in revenue on Shopify Plus or a comparable subscription platform, this is the point where the buyer’s own memory stops being a reliable forecasting method. A single buyer can hold maybe a few hundred SKUs’ seasonal patterns in their head with any accuracy. Past that, patterns blur, a launch from eighteen months ago gets mistaken for a launch from six months ago, and the spreadsheet model that worked at a much smaller scale starts producing orders that are either too large or too late.
The planner doesn’t remove judgement from the process. It removes the arithmetic. A buyer still decides whether to trust a given recommendation; the planner just makes sure the starting number reflects the actual sales pattern rather than a guess extrapolated from the last invoice.
What Data Does an Inventory Planner Need to Forecast Demand?
A forecast is only as good as what feeds it, and most of the disappointment operators report with inventory planning software traces back to a thin or dirty input set rather than a bad model. Six inputs matter more than any others.
| Input | What it tells the planner | What breaks if it’s missing |
|---|---|---|
| Sales history, by SKU and channel | The baseline demand pattern | Forecast defaults to a flat average, missing peaks and troughs |
| Supplier lead time, including variance | How far ahead to place the order | Recommendation arrives too late to prevent a stockout |
| Current stock on hand and on order | What’s already covered | Planner suggests reordering stock that’s already inbound |
| Promotional calendar | When demand will spike above baseline | Sale-week orders come in sized for a normal week |
| Seasonality calendar | When demand shifts by category, not just by SKU | A new SKU in a seasonal category gets a flat, unseasonal forecast |
| Supplier MOQs and pack sizes | The smallest valid order | Recommendation gets rounded manually, or ordered short |
Take one sentence from that table: without a clean, per-SKU sales history split by channel, everything downstream of it is a guess dressed up as a number, because the model has no baseline to compare current sales against.
Lead time deserves its own note, because it’s the input most brands supply as a single average when it should be a range. A supplier that ships in 14 days on average but has shipped anywhere from 9 to 25 days over the last year presents a very different risk profile from one that ships in 14 days every time. An inventory planner that only takes an average lead time will size safety stock for the average case and leave you short whenever the supplier runs long, which for an unreliable supplier is more often than the average suggests.
Returns data matters for categories with meaningful return rates, because a sale that later reverses as a return should reduce the demand signal, not inflate it. A planner working from gross sales rather than net sales after returns will consistently overforecast in categories where returns run high, and nobody notices because the error looks like normal forecast noise rather than a structural bias.
What Does an Inventory Planner Change About the Buyer’s Day?
Before a planner is in place, a buyer’s week typically starts with pulling a sales report, comparing it against a stock report in a second tab, applying a rule of thumb for how many weeks of cover to hold, and building a purchase order line by line. That process scales badly. A buyer covering 200 SKUs can do this thoroughly. A buyer covering 2,000 SKUs across several suppliers cannot, and starts triaging by gut feel, which means the SKUs that get proper attention are whichever ones the buyer happened to notice were low.
With a planner running, the buyer’s day starts from a ranked list instead of a blank spreadsheet. SKUs the model is confident about, ones with clean, stable history and no unusual factors, sit at the bottom of the queue with a recommendation the buyer can approve in bulk. SKUs flagged for a promotion, a new launch, a lead-time change or an unusual demand swing sit at the top, because those are the ones where the model’s confidence is lowest and a human decision adds the most value.
A planner’s real shift is this: purchasing moves from a task of building every number from scratch into a task of reviewing exceptions. The buyer’s judgement moves from “what should I order” to “is this recommendation right”, which is a faster question to answer and, done properly, catches more of the situations a spreadsheet model would have missed entirely.
The failure mode operators run into here is trusting the bottom of the queue too readily. A SKU with stable history looks safe precisely because nothing unusual has happened to it recently, and that’s exactly the condition under which a buyer stops checking it. Then a supplier discontinues a component, or a competitor undercuts the price, and the first sign anyone notices is a stockout, because the SKU had quietly stopped being reviewed.
Where Do Inventory Planner Recommendations Go Wrong?
Every inventory planner is a statistical model built on the assumption that the future looks roughly like the recent past, adjusted for known seasonal patterns. That assumption holds for the majority of SKUs most of the time, which is exactly why the exceptions catch people off guard. Three situations account for nearly every recommendation an experienced buyer overrides.
How Promotions Distort an Inventory Planner’s Forecast
A promotion breaks the planner’s core assumption twice: once during the sale, when demand spikes well above baseline, and once afterward, when demand often dips below baseline as customers who would have bought later pulled their purchase forward into the discount window. A planner that isn’t told a promotion is coming will size the order for baseline demand and stock out mid-sale. A planner that isn’t told the promotion has ended will read the post-sale dip as a genuine drop in demand and underorder for weeks afterward.
The fix isn’t complicated, but it requires discipline most teams don’t have: every promotion needs to go into the planner’s calendar before it runs, with an estimate of the expected demand lift, not just logged after the fact for reporting. Brands that only log promotions retrospectively get an accurate history but a forecast that’s still blind going into the next one, because the model has no forward-looking signal to act on.
How Seasonality Trips Up an Inventory Planner
Seasonality sounds like a solved problem, but most planners apply it at too coarse a level. A category-level seasonal curve, say, a lift every November, gets applied evenly across every SKU in that category, when in practice individual SKUs within a category often peak on different weeks or don’t follow the category pattern at all. A gift-focused SKU might peak in the first two weeks of December, while a practical, everyday item in the same category sees a flatter lift spread across the whole quarter. Apply one curve to both and you’ll overorder the practical item and underorder the gift item.
Season length shifting year over year is the second seasonality failure. A planner trained on three years of data where the peak always fell in week 47 will assume week 47 again, and if a shift in the sale calendar, a change in shipping cutoffs, or a competitor’s earlier promotion pulls that year’s peak into week 44, the recommendation arrives after the demand has already happened.
Why an Inventory Planner Has Nothing to Say About a New SKU
A new SKU with no sales history is the case a demand forecast is structurally worst at, because there’s nothing for the model to extrapolate from. Left to its own defaults, most planners either forecast near zero for a brand-new SKU, badly underordering a launch that should sell, or apply a category average that ignores whatever made the launch distinct in the first place.
The workable approach is to give the planner a proxy: name a comparable existing SKU, or an earlier launch in the same line, and tell the model to forecast the new SKU against that history until it accumulates enough of its own sales data to stand on its own, typically several weeks to a full season depending on how fast the category turns. This is a manual step every time, and it’s the step teams skip when they’re moving fast, which is exactly when a launch is most likely to sell through in the first few days and then sit out of stock for the rest of the season.
Is a Store Inventory Planner the Same Thing as a Reorder Point Calculator?
No, and treating them as interchangeable is a common source of confusion. An inventory planner answers “how much will I sell”, producing a demand forecast; a reorder point calculator answers “when do I need to place the next order”, using that forecast plus lead time and safety stock to trigger a purchase order. The mechanics of setting a reorder point, sizing safety stock against lead-time variability and running the purchase order process are covered in detail in our companion article on inventory ordering systems, which this piece deliberately doesn’t repeat.
Who Should Not Buy Inventory Planner Software Yet?
A brand below roughly $3M in revenue, or one running a narrow SKU count where a single buyer can still hold the seasonal pattern of every product in their head, is usually better served by a well-maintained spreadsheet with manually set reorder points than by a dedicated planner. The setup cost of a planner, cleaning sales history, mapping bundles, building a promotional calendar, tends to exceed the time it saves at that scale.
A brand still selling through a channel with unreliable or fragmented sales data, several POS systems that don’t reconcile, a marketplace feed that drops orders, is also a poor candidate until that data problem is fixed first. A planner built on unreliable inputs produces confidently wrong recommendations, which is worse than an admittedly rough spreadsheet, because a spreadsheet’s limitations are visible and a planner’s are not.
Finally, a brand whose catalogue turns over faster than its sales history can accumulate, drop culture, made-to-order, or a business built almost entirely on limited runs, gets little value from a system whose core strength is learning from repeat patterns. That business needs to run new-SKU forecasting, comparing each drop against a named comparable product, as its primary method rather than an occasional exception, and a planner built around stable, repeating SKUs isn’t the right fit for it.
How Do You Know Whether an Inventory Planner’s Forecast Is Working?
Track forecast error at the SKU level, not just in aggregate, because an aggregate number can look healthy while individual SKUs are badly wrong in offsetting directions. A category that’s overforecast on half its SKUs and underforecast on the other half produces a near-perfect aggregate total and a stockroom that’s simultaneously overstocked and out of stock.
Separate the error into two questions: is the forecast biased, consistently high or consistently low, or is it just noisy, right on average but wrong in both directions week to week? A biased forecast usually points to a fixable input problem, a stockout period baked into history, an unflagged promotion, a seasonal curve applied at the wrong level. A noisy forecast with no consistent bias is closer to the natural limit of what’s predictable for that SKU, and the fix is a wider safety stock buffer rather than a better model.
Log every manual override alongside the reason for it. This is the step that separates a planner that improves over time from one that repeats the same mistakes indefinitely. If a buyer overrides a recommendation because a promotion wasn’t in the calendar, that override should feed back into how the next promotion gets logged, not just correct that one purchase order. Without that feedback loop, the same category of error recurs every cycle because the correction lived only in one buyer’s head and left the system unchanged.
What Should You Check Before Turning On Inventory Planner Recommendations?
Before letting a planner’s output drive purchase orders automatically, confirm four settings rather than assuming defaults are right for your business. First, the lookback window: how many weeks or months of history the model uses, and whether that window is long enough to capture a full seasonal cycle for your category. Second, whether stockout days are excluded from the training data or counted as ordinary low-demand days. Third, whether the promotional calendar is a manual entry a person maintains or something the system infers from past sales spikes, because inference after the fact can’t help with a promotion that hasn’t happened yet. Fourth, whether channel-level forecasts are kept separate before being summed, particularly if you sell on Shopify alongside a marketplace or wholesale channel with a different demand pattern.
These four settings need a scheduled review, not a set-once-and-forget configuration. A lookback window that was right at launch drifts wrong as your business grows or your category shifts.
What Happens When Buyers Override the Inventory Planner and Never Log Why?
Silent overrides are where a planner’s value quietly erodes at growing brands, and the failure rarely shows up as a single dramatic event. A buyer sees a recommendation they know is wrong, because they know a promotion is coming that isn’t in the calendar, or they know a supplier just changed lead time, and they change the order quantity by hand. That’s the correct call. The mistake is doing it silently, in the purchase order tool, without recording the reason back in the planner.
The next time that same situation recurs, the model has no memory of the correction. It makes the identical wrong recommendation, and the buyer makes the identical manual fix, quietly, again. Multiply that across a few dozen SKUs and a few promotional cycles a year, and a team can run an inventory planner for two years without its forecast accuracy improving at all, because every correction that would have taught the model something instead lived and died in one buyer’s judgement.
The operator-level fix is procedural, not technical: every override needs a reason code, even a short one, fed back into the system that generated the recommendation. A team that treats this as optional treats their planner as a one-time purchase rather than a system that’s supposed to get better the longer it runs.
Forecast accuracy is fundamentally an operations problem before it’s a software problem: a planner is only as good as the process that feeds it clean data and captures the exceptions it misses, and building that process is what ops automation work is for. Pointerflow’s ops automation service is built around exactly this kind of feedback loop, closing the gap between what a planning tool recommends and what actually needs to happen on the purchasing side.
Sources
No external figures are quoted in this article; it is written from the operational mechanics of how demand-planning software ingests sales, lead-time and calendar data, and where that process commonly fails at growing ecommerce brands.