Academy › Build › Bookings and orders
The buffer of units you never sell
You'll decide how many units to always hold back as a buffer, based on where a sale can slip past your website unnoticed, and know when that number needs to be zero.
By Yodai · Published September 6, 2026 · Updated October 12, 2026
This lesson is about selling something you no longer have less often on your website: the buffer you set aside ahead of time. You pad the stock you show on your website by holding back two or three units that never go up for sale, just in case. The question almost nobody stops to answer is how many. Set aside one, and it might not be enough. Set aside twenty, and those twenty never sell, which is lost money too.
The number doesn't come from a ratio table or a rule that fits every business. In inventory management, the classic buffer, safety stock, is calculated from the error you make forecasting your own sales, as explained in a technical note from the Universitat Politècnica de València. Here it comes from one very specific question: where does a sale slip through that your website never finds out happened?
The buffer solves a different problem than the tie does
You already know what happens when two people buy the last unit at the same time: whoever pays first gets it, and you tell the other person right away. That happens within seconds, and it already has its answer.
The buffer isn't for that case. It's for the gap that opens up over hours, or days, between what your website thinks you have and what's actually on the shelf. The two numbers need to match, and the system believing you have units that aren't really there is, according to another technical note from the same professor at the Universitat Politècnica de València, a problem almost worse than the opposite one. It's not a tie: it's a silent mismatch, and it creeps in through very different places.
Where that mismatch comes from
It almost always comes from one of these places. You also sell at a physical counter, and whoever's ringing up sales there doesn't touch the dashboard until they get a free minute. You sell by phone or WhatsApp to a regular customer and jot the sale down in a notebook instead of on the website. You count inventory by hand once a week, and that day, until you check it, the website still shows last week's number. Or the product breaks, spoils, or gets lost along the way, and nobody subtracts that unit until someone stumbles on the empty spot.
Notice the pattern: in all four cases, the problem isn't sales volume. A product you sell once a week might need a bigger buffer than one you sell twenty times a week, if the first also sells at the counter and the second sells only on your website.
Shrinkage calls for its own kind of care. A study covering about 24,000 SKUs from 11 grocery stores found more mismatches in perishable products, and that counting stock worked better for those than for everything else. And you can't fix it just by adding extra units: an MIT thesis showed that a small loss nobody logs, built up over time, can end up causing serious stockouts, and that holding extra stock isn't an effective way to deal with it. For things that break or expire, the buffer covers the shrinkage until the next count, and it's the count that sets the number straight.
When the buffer is zero
If you sell only through your website, your product doesn't break or expire, and payment deducts the unit instantly like you saw in the first lesson, the buffer you need is zero. There's no mismatch to cover, because there's no other place for a sale to slip through.
Only add a buffer where you've identified the parallel channel. If you can't name what it is, don't add one.
- Easy to understand at a glance, no math required
- Works well when the product moves slowly and you know the exact size of the discrepancy
- Needs a manual review if the sales pace changes
- It comes from the same rule every time you update your stock, without having to come up with a number for each product
- Makes sense when you carry many SKUs with different sales paces and you're not going to review them one by one
- On a product with very little stock, 10% rounds down to zero and protects nothing
Which one do I pick: You don't have to treat every product the same way. In ABC classification, the few products that account for most of your sales are the ones that need more attention and review, while the many that account for little get the simplest control possible, as the MIT logistics course notes teach. Give the first group a fixed number and check it by hand; give the rest a percentage, and for the ones that barely carry any stock, add a fixed minimum on top of the percentage.
The expensive mistake isn't setting too small a buffer: it's setting one for a product that sells only through your website, where payment deducts the unit instantly and there's no counter or notebook to throw it off. There, the buffer doesn't protect anything, it just hides units you could actually sell. Multiply that by a hundred products and that's a hundred sales that never happen, all from a habit copied from a business with a physical store.
The second mistake comes later: the buffer you set on day one and never look at again. You close the store counter, and the three units from the earlier note stay reserved forever, with nobody left who remembers why.
- You know the buffer answers one question: where could a sale slip through that the website doesn't see
- You've gone through your products and know which ones need a buffer of zero
- For the ones that sell through another channel, you have a number or a percentage, plus the note explaining why it's there
- You know that number gets reviewed when the parallel channel changes, it's not set in stone forever
Test yourself
Answer before opening each one. If one gives you trouble, go back to the section that explains it.
1. You sell candles only through your website, with no separate store or WhatsApp channel, and even so, twice this month you've sold a candle you no longer had.
The buffer isn't the solution: with no separate store or WhatsApp channel, there's no parallel channel for a sale to slip through. Check first whether the number your website had matched what was actually on the shelf: if a candle broke and nobody deducted it, a count is what sets the number straight.
2. You have a physical store and twenty clothing items, and for your best seller, the counter sets aside an average of four units between counts.
2. You have a physical store and twenty clothing items, and for your best seller, the counter throws off an average of four units between counts.
Fixed buffer: on the website, you set the quantity four units lower than what you actually have of that item, until the next count. It's your best seller: one of the few that deserve their own number, checked by hand at every count.
Drop the buffer to zero for every product that had that note, since the separate channel that justified it no longer exists. Leaving it in place just hides stock you could actually be selling now.
Sources
- Independent Demand Inventory Management. Technical Note. José Pedro García Sabater, Universitat Politècnica de València (RiuNet repository), 2020.
- Counting: Inventory Data Quality. Technical Note. José Pedro García Sabater, Universitat Politècnica de València (RiuNet repository), 2020.
- Inventory record inaccuracy in grocery retailing: Impact of promotions and product perishability, and targeted effect of audits. Yacine Rekik, Rogelio Oliva, Christoph Glock, and Aris Syntetos, arXiv, 2025 (revised 2026).
- Information inaccuracy in inventory systems. Yun Kyu Kang, doctoral dissertation, Massachusetts Institute of Technology, 2004.
- Supply Chain Fundamentals & Segmentation Analysis. Chris Caplice, Logistics Systems course (ESD.260), MIT OpenCourseWare, 2006.
In Yodai the store comes with the Free plan, with a catalog, a cart and orders. See how to create your online store for free.
NextBefore you say "there's no availability," decide whether a waitlist is worth opening