Harpy Glossary

PBS (Predictive Buying System)

Amazon & D2C glossary · Harpy Media

A PBS (Predictive Buying System) is inventory software that forecasts demand from historical velocity, seasonality, and trend signals, and converts those forecasts into procurement decisions — when to reorder, how much, and at what point the order needs to be placed to arrive in time.

What is PBS?

It exists because manual replenishment has a hard ceiling. Spreadsheets react to what has already happened; a predictive system reads trajectories, recognises seasonal patterns, and calculates reorder points against lead times automatically. For a catalogue of any size, that difference is the gap between inventory that flows and inventory that lurches between panic-orders and stockouts.

What a PBS actually changes

The core improvement is the shift from trailing averages to forward-looking demand. A trailing 30-day average cannot see a trend building or a seasonal spike coming — so it orders late into rising demand and early into falling demand, which is exactly backwards. A predictive model weighted with trend and seasonality produces a demand estimate that anticipates rather than reflects.

The second change is automation of the timing. Reorder points calculated against real lead times, purchase quantities sized to forecast demand plus safety stock, and triggers that fire without a human remembering to check. Every manual step removed is an error mode eliminated — and inventory errors are among the most expensive a seller can make.

What to look for, and how to run it

Three capabilities matter: demand forecasting that accounts for trend and seasonality rather than simple averages; lead-time awareness, including freight and receiving variance; and the mathematics of safety stock — sized to demand variability and service level rather than to a round number. The systems that do this well also reconcile forecasts against actuals continuously, so accuracy improves with use.

The human job becomes oversight rather than calculation: review exception reports, sanity-check the big commitments (a first-quarter order for a peak season is worth a human eye), and keep the inputs honest. Forecast quality depends entirely on data quality — shipping times recorded accurately, stock counts correct, promotional plans visible to the model. A system fed bad data will automate bad decisions confidently.

Target Procurement Quantity (Q) ≈ Forecast Daily Demand × Lead Time + Safety StockSafety stock is sized from demand variability against the service level you choose. The system recalculates as velocity, trend, and lead times change.

In practice

A seller running a five-pound espresso machine with a 45-day combined manufacturing and freight lead time uses a predictive system rather than spreadsheets. The model spots a rising trend in related search volume and combines it with past peak-event data to project a demand spike around 50 days out — and triggers the reorder earlier than a trailing-average report would have. The stock lands days before the event, velocity is uninterrupted, and margin is maximised.

⚠️ Watch out. Running replenishment on static trailing averages. A competing seller orders from a 30-day backward view: when demand starts climbing ahead of a peak, the average still reads normal, the order goes out too late, and the stockout arrives at the worst possible moment. The spreadsheet was accurate about the past and useless about the future — which is the only part the purchase order needed.

💡 Harpy tip. Give the system clean inputs and keep a human on the exceptions. Forecast accuracy depends on accurate lead times, correct stock counts, and promotions declared in advance; the value of the model collapses if any of those are approximate. Then review the large orders and the first-time events personally — automation handles the routine, judgement handles the unusual.

How Harpy Media helps

Replenishment systems are part of the operations stack we build with brands: forecasting wired to real lead times, reorder points automated, and the exception reviews that keep automated ordering trustworthy through peaks.

PBS FAQ

What is a predictive buying system?

Software that forecasts demand from sales history, seasonality, and trends, then converts the forecast into reorder decisions — calculating when and how much to buy so stock covers demand without excessive carrying cost.

How is it different from a spreadsheet?

Spreadsheets typically look backwards at recent averages; predictive systems look forwards, weighting trend and seasonal patterns and calculating reorder points against actual lead times. That difference is what prevents the late orders that cause stockouts.

What data does a PBS need to work well?

Accurate sales history, real lead times including freight and receiving, correct current stock counts, and advance notice of promotions. Forecasts are only as good as the inputs — bad data automates bad decisions efficiently.

Want these numbers watched for you, every week?

Book Free Consultation