Demand forecasting software that runs on your actual Philippine sales data.

We replace. We build. We optimize.

Every line of code we write is yours at cutover. No license. No annual increase. No lock-in.

What is demand forecasting software and why does it matter for Philippine businesses?

Demand forecasting software predicts how much of each SKU you will sell in a future period, so you order the right amount at the right time — not too much, not too little.

In a Philippine distribution context, the stakes are concrete. A 3-day stockout on a fast-moving consumer good during peak season can cost you a key account — retailers in the Philippines have short patience for empty shelves when a competitor is available. An overstock of 60 days on a slow-moving SKU in a 40°C warehouse means a write-off when the batch expires. Most Philippine companies manage this tension with experience and memory: the sales manager knows which months are heavy and tells procurement to order more. That works when you have 50 SKUs and one warehouse. It fails when you have 400 SKUs across three regional depots and the sales manager who held that knowledge has left.

How much does demand forecasting software cost in the Philippines?

Global subscription tools run ₱1.5M–₱3.5M per year for mid-market SKU volumes. An Orkids build is a one-time fee — typically 15–25% of a three-year subscription, trained on your data, owned by you, no annual license.

At 400 SKUs across three warehouses, a Slimstock subscription runs approximately $25K–$40K per year — roughly ₱1.5M–₱2.4M annually at current exchange rates. Over three years, that is ₱4.5M–₱7.2M with no equity in the system at the end. A one-time Orkids build, owned outright, produces a forecasting engine trained specifically on your sell-through history, your supplier lead times, and your Philippine seasonal patterns. Optional operations support — monitoring, model retraining as your business changes, and new SKU onboarding — is ₱50K–250K per month with no lock-in. Your exact scope and price are confirmed on the first call.

How does the forecast handle Philippine seasonal demand peaks like Holy Week and Christmas?

Philippine seasonal peaks are explicitly modeled — Holy Week, Christmas, back-to-school, and your business-specific peaks — not derived from global averages that have no Philippine market data.

Global demand forecasting tools are trained on anonymized datasets dominated by North American and European retail patterns. Christmas in those markets peaks on December 25. In the Philippines, Christmas purchasing starts in September, accelerates through October and November, and the single-day peak is December 24. Holy Week is a consumption event in the Philippines, not a shopping event — convenience stores restock differently than supermarkets during that week. Back-to-school in the Philippines is July, not August. These are not edge cases; they are the majority of your annual volume variance. The Orkids forecasting engine trains on your historical data, which already reflects these patterns, and then structures them explicitly so the model can extrapolate forward without distortion.

What data does the system need to start forecasting?

Two years of SKU-level transaction data and your current supplier lead times. We extract and clean the data from your existing system — you do not rekey anything.

The minimum useful dataset is 24 months of daily or weekly sales per SKU per location. Beyond that, additional inputs improve accuracy: promotional event flags (dates and scope of past promotions), price change history, supplier lead time by vendor, warehouse-level stockout dates, and any externally sourced demand data you already track. We connect directly to your existing ERP, distribution system, or accounting database to pull the data. We then clean it: remove stockout-distorted weeks, flag promotion spikes, and identify data entry errors before training the model. The cleaning step typically takes two to three days and is included in the implementation. You do not see the raw accuracy figure — you see the clean accuracy figure after the distortions are removed.

How does the system handle new SKUs with no sales history?

New SKUs are forecast using a combination of product family analogs and the launch curve of historically similar SKUs — no minimum history required to generate a first-week purchase order.

New product forecasting is a real problem for Philippine distributors who launch 20–40 new SKUs per quarter. Without history, most companies apply a flat opening order based on the product manager's intuition and a similar product's first-month sales — then adjust after 60 days when it is too late to avoid the first stockout or the first overstock. The Orkids system uses product family groupings you define during setup. When a new SKU in the 'personal care — moisturizer — mass market' family launches, the model starts with the launch curve of the last five SKUs in that family and adjusts for distribution point count, price tier, and promotional support. The first purchase order is data-informed on day one.

How does the system clean promotion spikes out of the baseline?

Promotion dates and scope are logged as events. When the model trains on historical data, promotion-period sales are replaced with the pre-promotion baseline rate for those SKUs.

Promotion cleansing is what separates a genuine forecasting system from a trend-extrapolation tool. When you ran a buy-one-get-one-free promotion on a product in March, your customers stocked up at home. The week before the promotion, sales were below normal. The promotion week, sales spiked to three to four times normal. The two weeks after, sales were below normal as customers worked through their at-home stock. If you train a forecast model on that raw data, it sees a March spike and predicts a March spike next year. With promotion cleansing enabled, the model strips the spike and the pre/post dip, replaces them with what the underlying trend line suggests baseline demand was, and preserves only the genuine demand signal. The promotion event is stored separately so you can rerun its impact analysis when planning the next one.

How does the system handle slow-moving and dead stock SKUs?

Slow-moving SKUs are forecasted at their actual demand rate, with safety stock set to match their real service-level contribution — not the same policy applied to your A-movers.

ABC-XYZ segmentation is built into the forecasting engine. A-movers with high velocity and stable demand get tight safety stock and frequent replenishment. C-movers with low velocity and erratic demand get longer review cycles and minimum order quantities aligned to their actual consumption rate. D-movers — SKUs with fewer than three transactions in the past 90 days — are flagged for either discontinuation or a minimum-stock hold decision. The system does not automatically discontinue SKUs; that is a commercial decision. What it does is surface the cost of the decision: 'SKU X has ₱180,000 of inventory on hand, has sold 12 units in the past 90 days, and the nearest expiry is 45 days away.' That information, delivered automatically each Monday morning, is what the category manager needs to act before the write-off occurs.

Does the system integrate with existing ERP and warehouse management systems?

Yes. The forecasting engine reads from your existing ERP or WMS and writes purchase order recommendations back into it — you do not run two separate systems.

Philippine FMCG distributors typically run one of a handful of systems: HashMicro, SAP Business One, Netsuite, Microsoft Dynamics, or a local ERP built in the early 2000s. The Orkids integration layer is built during implementation for your specific system. The forecast engine reads sales transactions, current stock levels, and open purchase orders on a daily schedule. It writes a purchase order recommendation file — in your ERP's native import format — back to the system each planning cycle. Your procurement team reviews the recommendations in the same screen they already use for purchase orders. They approve, adjust, or reject each line. The system learns from their adjustments: if a buyer consistently orders 10% more than the recommendation on a particular SKU, the model notes that and adjusts its future recommendations upward for that SKU.

How does the system account for variable supplier lead times?

Lead time variability is modeled per vendor, not assumed constant. Safety stock is calculated using actual lead time distributions from your purchase order history, not a fixed number of days.

Philippine distributors importing from suppliers in China, Europe, and the US face real lead time variability: a shipment quoted at 21 days arrives in 14 days one quarter and 34 days the next because of port congestion, customs holds, or vessel schedule changes. If your safety stock is calculated on the assumption of a fixed 21-day lead time, you will have stockouts in the quarters when it arrives late. The Orkids system pulls your historical purchase order data — ordered date, expected arrival, actual arrival — and builds a lead time distribution per vendor per product category. Safety stock is then calculated to cover the 90th or 95th percentile of that distribution, depending on the service level you specify. A vendor with consistent 21-day lead times gets a tight safety stock. A vendor with lead times that range from 14 to 40 days gets a buffer that reflects the real risk.

Can the system handle multiple warehouses and regional depots?

Yes. Each warehouse or depot has its own demand signal and safety stock calculation, with a consolidated view for the planning team across all locations.

Philippine distributors with regional depots face a specific problem: the demand pattern in Metro Manila is different from Cebu, which is different from Davao. A single national forecast does not tell you how much to stock at each location — it tells you how much to buy in total, which is a different calculation. The Orkids system maintains a separate demand model per location and generates location-specific purchase order recommendations. The central planning team sees a consolidated dashboard showing coverage days by SKU across all depots. When a SKU is overstocked in one depot and understocked in another, the system flags the imbalance and proposes an inter-depot transfer before a purchase order is placed. Inter-depot transfers are cheaper than expedited supplier orders and faster than waiting for the next replenishment cycle.

How does the system support financial planning and budget vs. actual analysis?

The demand forecast feeds directly into the procurement budget — projected purchase orders generate a rolling 90-day cash requirement forecast that the CFO can compare to actual spend each week.

Demand forecasting and financial planning are the same calculation viewed from two directions. The demand forecast says: we expect to sell this much of each SKU. Multiplied by cost of goods, that becomes: we expect to spend this much on procurement this quarter. When the CFO asks why the procurement budget is over by 12% this month, the answer should not be 'we had to expedite several shipments.' It should be a traceable chain from the demand signal that triggered the expedite, back through the lead time failure that made the expedite necessary, back to the safety stock policy that did not buffer enough for that vendor's variability. The Orkids system makes that chain visible. Budget overruns are diagnosed before they become chronic, because the root signal — an understocked SKU on a high-variability vendor — is surfaced two weeks before the expedite freight invoice arrives.

Can the system handle FEFO inventory management for perishable goods?

Yes. For perishable SKUs, the system forecasts demand in a way that ensures older stock is depleted before newer stock — first-expiry, first-out — and flags upcoming expiry risk automatically.

Philippine distributors handling food, pharmaceutical, and personal care products face a FEFO requirement that pure FIFO systems do not fully address. FIFO says issue the oldest stock first. FEFO says issue the stock that expires soonest first — and those are different when a product is received across multiple batches with different production dates and shelf lives. The Orkids demand forecast integrates with the warehouse management module to track batch-level expiry dates. When a slow-moving SKU has a batch expiring in 60 days, the system flags it in the weekly planning report and reduces the purchase order recommendation for that SKU — there is no point buying more of something you cannot sell before the current stock expires. If the risk of expiry is severe, the system generates a markdown recommendation so the product can be cleared through a promotional channel before it becomes a write-off.

How does the system handle demand signals from multiple channels — store, ecommerce, and B2B?

Each channel has its own demand model. The forecast aggregates channel signals into a single total demand view while keeping channel-level detail for purchasing and allocation decisions.

Philippine businesses running simultaneous retail store, ecommerce, and B2B distribution channels face a demand aggregation problem: the same SKU has different velocity patterns on Lazada versus in a sari-sari store versus in a hospital supply chain. A forecasting system that pools all channels into a single demand signal loses the information needed to make channel-specific stocking decisions. The Orkids system maintains separate models per channel and per location. The total demand view shows aggregate purchase requirements for the planning team. The channel view shows how much of the total purchase is allocated to each channel, which SKUs are channel-exclusive, and where a stockout in one channel is being masked by surplus in another. Allocation decisions are made with full channel visibility — not post-hoc when a key B2B account calls to say their order was not fulfilled.

What is the accuracy improvement over Excel-based forecasting?

Most Philippine companies that switch from Excel-based forecasting see forecast error drop by 20–40% in the first 90 days, primarily because the model captures demand patterns the spreadsheet cannot hold.

The accuracy claim is not a marketing number — it is the result of a specific structural advantage. An Excel model trained on monthly aggregates cannot see weekly demand patterns like the Friday-before-payday spike that many FMCG SKUs exhibit in Philippine mass-market retail. It cannot weight recent data more heavily than older data when a trend is changing. It cannot apply different seasonality coefficients per channel. And it cannot automatically re-estimate when a new data point arrives — someone has to update the spreadsheet. The Orkids forecasting engine retrain runs every time new transaction data is available, typically nightly. The model re-estimates seasonality, trend, and volatility coefficients based on the full available history, weighted toward recent observations. The accuracy metric — mean absolute percentage error per SKU — is visible in the planning dashboard. You can see which SKUs the model forecasts well and which still require planner override, and track whether overrides improve or worsen accuracy.

How long does implementation take?

Eight to twelve weeks from contract to a live forecasting cycle covering your full active SKU catalog — the first machine-generated purchase order recommendation replaces your manual planning process.

Implementation has three phases. Weeks one and two: data extraction from your existing systems, cleaning and validation, and configuration of your product hierarchy, warehouse structure, and supplier database. We identify and resolve data quality issues — duplicate SKU codes, missing lead times, truncated sales history — before they can distort the forecast. Weeks three through six: model training on your cleaned historical data, initial forecast generation, and a parallel-run period where the system generates recommendations alongside your existing planning process. Your planners review both and flag discrepancies. We use those flags to improve the model and the cleansing logic. Weeks seven through twelve: the model goes live as the primary planning tool. Your planners review and override recommendations rather than building them from scratch. We monitor accuracy metrics weekly and adjust the model as new data arrives and business conditions change.

How does the system adapt when demand patterns change permanently?

The model detects structural breaks in demand patterns and flags them for planner review — a new competitor, a channel shift, or a regulatory change is not assumed to be noise.

Demand forecasting systems fail when business conditions change faster than the model can adapt. A new competitor opens a store near your key accounts. A large B2B customer changes their ordering frequency. A BIR regulation changes the structure of your distribution contracts. Each of these events creates a visible break in the demand series — a period where the forecast diverges from actuals by more than the historical variance would predict. The Orkids system monitors forecast versus actual at the SKU level on a rolling basis. When the divergence exceeds a configured threshold for three or more consecutive periods, it flags the SKU as 'demand pattern change — planner review required.' The planner investigates the root cause and, if it is a permanent shift, resets the model baseline for that SKU using only the post-change data. The model does not blindly extend the pre-change pattern into a market that no longer exists.

Can the forecast recommend when to run markdowns to clear excess stock?

Yes. For SKUs with excess stock relative to the forecast demand rate, the system calculates the markdown price that clears the surplus before expiry while minimizing the margin loss.

Markdown optimization is the demand planning question that most systems do not address. The calculation is: given the current stock level, the forecast demand rate at full price, the days until expiry (for perishables) or the next season (for seasonal goods), and the price elasticity estimated from your historical promotion data — what is the lowest markdown percentage that clears the surplus in time? Running the markdown too deep leaves margin on the table. Running it too shallow means the stock does not clear and you end up writing it off anyway. The Orkids system does not make the markdown decision — that is a commercial judgment involving your customer relationships and your pricing strategy. What it does is give the category manager the clearance calculation, the expected sell-through at each markdown depth, and the expected residual stock if the markdown runs for the recommended period. The manager decides; the system supplies the arithmetic.

Is this system built for Philippine business or adapted from a global product?

Built specifically for Philippine operating conditions — BIR compliance triggers, Philippine seasonal patterns, supplier geography, and peso-denominated costing — not adapted from a global product.

The distinction shows up in two ways. First, the data model. A global demand forecasting product models a supply chain where the regulatory environment is stable and the primary uncertainty is demand variability. In the Philippines, regulatory changes — BIR form updates, BOC import policy adjustments, BFAD labeling requirements for pharma distributors — affect procurement timing and supplier relationships in ways the global model does not account for. Second, the integration layer. The Philippine ERP landscape is fragmented: many mid-market distributors run systems built in the early 2000s with proprietary data formats and no modern API. We have built connectors for the most common local systems encountered in Philippine FMCG and pharma distribution. That connector library does not exist in a global product's standard implementation package.