Runner, repeater, stranger is a way to sort products by how much and how often you make them: runners are high-volume items made almost every day, repeaters run on a predictable cycle, and strangers are the sporadic long tail. The classification decides what flows continuously, what runs to a fixed schedule, and what gets planned one job at a time.
The idea is simple, but it changes how a plant plans. Most schedules treat every SKU the same, one big planning queue, first in, first out, expedite whatever the customer is shouting about today. Runner-repeater-stranger says stop doing that. A product you make every shift and a product you make twice a year do not belong in the same scheduling logic. Sort them, then give each group the planning approach it deserves.
What do runner, repeater, and stranger actually mean?
Each label describes a demand pattern, not a part number. The same physical product can drift between categories as demand rises and falls, which is why you re-sort on a schedule rather than carving the labels in stone.
| Class | Demand pattern | Typical share | How you plan it |
|---|---|---|---|
| Runner | High volume, steady, ordered almost every day | A small handful of SKUs | Continuous or near-continuous flow; fixed line, minimal changeover |
| Repeater | Moderate volume, predictable, recurring on a cycle | The middle band | A fixed repeating schedule (every part every interval) on a shared line |
| Stranger | Low volume, sporadic, hard to forecast | The long tail of many SKUs | Made to order, slotted into open capacity as jobs |
Notice the pattern: as volume drops and unpredictability rises, you move from flow, to a fixed cycle, to one-off scheduling. That progression is the whole point. You spend your engineering and standard-work effort on the runners, where it pays back every day, and you keep the strangers from disrupting that rhythm.
Where does the idea come from?
Runner-repeater-stranger grew out of lean production leveling work and is closely tied to the Glenday Sieve, a product-sorting method developed by lean practitioner Ian Glenday. Glenday analyzed volume data across many companies and kept finding the same lopsided distribution: a very small share of products drove a very large share of volume. Rather than the familiar 80/20 split, his data pointed to something sharper, roughly 6% of products accounting for about half of total volume, and around half of the products producing 95% of it. The rest was a long tail of infrequent items.
That empirical shape is what makes the classification worth doing. If your volume were spread evenly across every SKU, there would be nothing to sort. Because it almost never is, sorting lets you run the vital few products in a stable, repeating pattern, the foundation of heijunka and level scheduling, while the erratic tail stays out of the way. If you want the color-banded version of this sort, see the Glenday Sieve method; if you want the repeating-cycle mechanics for repeaters, see every part every interval.
By the numbers. In his production-leveling work, Ian Glenday reported that roughly 6% of products typically accounted for about 50% of volume, with about half of all products delivering 95% of volume, a distribution far steeper than the classic 80/20 rule. The Lean Enterprise Institute documents this runner-first approach to levelled production, and iSixSigma catalogs the Glenday Sieve as a recognized product-classification method. See the Lean Enterprise Institute on levelled production and iSixSigma: Glenday Sieve.
How do you classify your own products?
You classify products by pulling real order history, sorting by volume and frequency, and drawing two cut lines. The mechanics are deliberately simple so the plant can repeat them every quarter.
- Pull 12 months of demand history. Use actual orders or shipments by SKU, not the sales forecast. Twelve months smooths out seasonality; a shorter window can misread a seasonal product as a stranger.
- Compute volume and frequency for each SKU. Total units (or standard hours of work) tells you size; number of distinct order periods tells you how often it comes up. You need both, a product can be high-volume but lumpy, or small but ordered every week.
- Sort descending by volume and add a running total. This is the same cumulative math behind a Pareto chart. The running percentage shows where the vital few end and the tail begins.
- Draw the runner line. The top SKUs that together make up roughly half your volume are runner candidates. Confirm with frequency: a true runner is ordered almost every day, not once in a big batch.
- Draw the stranger line. The long tail of low-volume, sporadic SKUs are strangers. Everything between the two lines is a repeater.
- Sanity-check the edges by hand. New products with short history, promotional spikes, and end-of-life items all lie to the math. Have planning and sales eyeball the borderline SKUs before you lock the list.
- Assign a planning rule to each class. Runners flow; repeaters go on a fixed repeating cycle; strangers are made to order into open capacity. Write the rule down so it survives a planner changing jobs.
- Re-sort on a cadence. Demand moves. Re-run the sort quarterly (or after any big product-mix change) so a product that has become a runner stops being scheduled like a stranger.
How does each class change the way you schedule?
Each class gets a different scheduling engine, and mixing them is what makes plants feel chaotic. Runners want a fixed, repeating rhythm, ideally a dedicated line or cell where changeover is near zero and the sequence barely changes week to week. That stability is what lets you build standard work, tune takt time and pull materials with a simple kanban loop instead of chasing every order.
Repeaters live on a shared line and run to a set interval, Monday this family, Wednesday that one, so the cycle is predictable even though no single product runs continuously. Strangers are the disruptors: infrequent, hard to forecast, often needing setup and materials you do not keep on hand. The trick is to protect the runner and repeater rhythm from them by reserving a slice of capacity for stranger jobs rather than letting a rush order blow up the whole week's sequence.
What should you do with the strangers?
Strangers are where the classification pays a second dividend: they are your rationalization list. A stranger that is strategically important, a spare, a regulated variant, a key-account special, stays, and you plan it as a job. But a stranger that is low-volume, low-margin, and only there out of habit is a candidate to cut, standardize, or push to a longer lead time. Every stranger you retire removes a setup, a slug of slow inventory, and a disruption to the runner rhythm. The point is not to kill the tail wholesale; it is to make a deliberate keep-or-cut decision on each one instead of carrying it by default. Run that review once a year and the tail stops quietly growing.
Where does runner-repeater-stranger go wrong?
The classification fails in a few predictable ways, and all of them come from treating the labels as permanent or letting the data lie to you.
- Never re-sorting. A list built two years ago describes a plant that no longer exists. Products graduate and demote between classes; if you do not re-run the sort, you end up flowing a dying product and job-shopping a new hit.
- Confusing volume with frequency. A product ordered in one giant annual batch is high-volume but is not a runner, it does not want a flow line, it wants one well-planned campaign. Always check both dimensions.
- Letting strangers set the tempo. If every rush order reshuffles the runner sequence, you have thrown away the entire benefit. Protect the rhythm; give strangers their own reserved slot.
- Dirty demand data. Free-text SKUs, one-time promotions, and inter-plant transfers logged as customer orders all distort the sort. Clean the data first, the classification is only as honest as the order history behind it.
- Sorting once and never acting. The list is not the deliverable. The deliverable is a changed schedule: a dedicated runner line, a fixed repeater cycle, a reserved stranger slot. A tidy spreadsheet nobody plans from is wasted effort.
That last point is where accurate data matters most. Runner-repeater-stranger depends on trustworthy demand and run history, and in plants still reconstructing production from paper at shift end, the numbers are usually too soft to sort by. When run data is captured in real time at the line, the sort reflects what the plant actually made, which is the foundation the whole method sits on. That is the shift CLS made by replacing paper production logging with real-time capture.
How it fits the rest of lean
Runner-repeater-stranger is a planning lens, not a standalone program. It feeds directly into lean manufacturing by telling you where to invest in flow, and it is the natural first step before heijunka and level scheduling, you cannot level a schedule you have not sorted. Do the sort, act on it, and re-sort on a cadence, and the plant stops treating a twice-a-year product like a daily one. That single change quietly removes a surprising amount of the churn that makes scheduling feel impossible. For the broader toolkit this sits inside, see the seven basic quality tools several of which, Pareto and stratification especially, are exactly how you build and defend the sort.