Direct Pack + Pack-to-Light Case Study

How Picking Errors Dropped by 93%

ShipHero measured Pack-to-Light across six rounds of testing at real Direct Pack stations. The study recorded roughly 350 orders on video. At the large station, picking time fell 31% to 36% and estimated daily capacity rose 7% to 15%. This means 10 to 24 extra orders per day from the same station and the same person. The error and search results were the most consistent finding of all, holding at every station size: in about 120 lights-on orders watched, no picker ever went back to the screen. The value compounds as stations get larger and orders get more complex.

A picker moves faster when they don’t have to think about where anything is. That is the entire idea behind Pack-to-Light, and it sounds obvious until someone asks how much. “Faster” is cheap to claim. So a team at ShipHero spent six rounds measuring it on video, across real production stations and controlled lab replicas, with a stopwatch running on some 350 orders

The tests were conducted in very specific settings: Direct Pack stations. This type of pack station puts inventory in bins right around the packing table, allowing one person to pick and pack each order in a single spot. No need to walk to a separate picking area.

The ShipHero Pack-to-Light tool adds LED lights to those bins. When an order opens, every bin it needs lights up. The picker goes to a lit bin, picks the item, and its light goes out, then they move to the next lit bin, until none are left. If an order needs several units of the same item, the light alternates colors until the picker has them all. The picker doesn’t need to read the order screen: they just follow the lights.

How the study worked

Timing whole orders, lights on against lights off, produced nothing usable. That’s because the time a picker spends boxing, wrapping and labeling varies widely from order to order, burying the difference the lights made. So instead of timing whole orders, the team counted steps. Once the walking effect was clear, time came back in, now with packing carved out of the measurement entirely.

Every comparison was the same picker, at the same station, working real orders with the lights on versus off. At every station, each bin held a single product. And for the purposes of this study, one order line corresponds to one location visited.

The study tested six rounds, run at ShipHero’s Dallas warehouse and in the Buenos Aires Lab (*):

• 2 rounds at Company 1 conducted in a large station: 205 sq ft (19 m²) with 160 pick locations, with two different pickers.
  
  
• 1 round at Company 2 conducted in a compact station: roughly 65 sq ft (6 m²) with 65 locations, worked by a veteran who runs it daily.
  
  
• 3 rounds conducted at the lab, in replica stations of 59 to 87 sq ft, built to change one variable at a time.

Regarding density, all the tested stations have one location per square foot. There’s one exception though: because of its particular setting, Company 3 ranges from 15 to 20 locations per square foot, which completely changes the picture. Since this article focuses on standard pack stations, we analysed the impact of higher densities in a separate study: Card Walls + Pack-to-Light Case Study: Picking Time Dropped by up to 67%.

The clock started the moment the order was on screen and the picker was free to act on it, not the moment they began walking. Those are rarely the same instant. An order often opens while the picker is still closing the previous box.

That detail decides the result. The seconds a picker burns reading the screen and memorizing locations are precisely what the lights delete, so a stopwatch that starts any later measures everything except the point.

What the lights actually remove

Without lights, a picker works from memory. They must read the screen, hold three or four locations in their head, walk, forget one, walk back and check. While it is not possible to show this on a stopwatch, all of it shows up on video.

The rounds tested 146 lights-off orders and about 120 lights-on orders, showing that:

• Pickers went back to re-check the screen mid-pick in 6 of every 10 orders without lights. With lights on, that did not happen once in 120 orders.
  
  
• Walking toward the wrong bin happened in about 1 in 4 orders without lights, against 1 in 120 with them.
  
  
• Reaching into the wrong bin happened 1 in about 8 orders without lights, against 1 in 120 with them.
  
  
• Finishing an order with an item missing, then going back for it, happened in roughly 1 in 15 orders without lights, and 1 order in 120 with them.

With lights, missed items dropped to about an eighth of the rate, though not to zero.

Also, with lights on, 92% of orders ran clean from start to finish. Without lights, 31% did. In the other 69%, the picker’s memory failed and got patched on the fly, mid-order, while the clock ran.

And with lights on, order-to-order times varied 2.5 to 3 times less than without them. And lost orders, where a picker spends 30 to 70 seconds hunting for a single location, never happened with them. That means picking became both faster and more predictable, which makes a station's output easier to plan.

What it does to the clock

At Company 1’s large station, on a typical order:

• Picking time, measured from the order appearing on screen and the picker being ready to start, to every item sitting on the pack table, fell from 29 to 33 seconds down to 20 to 21 seconds. That is a 31% to 36% reduction in picking time.
  
  
• Steps walked per order fell from 18 to 24 down to 13 to 14, roughly 30% to 40% fewer.
  
  
• Trips between the pack table and the shelves fell from 1.5 to 2.1 per order down to 1.0 to 1.2, roughly 30% to 40% fewer.

Behind those numbers sits a pattern that is held across all six rounds. Every order gets a base saving of 2 to 2.5 seconds, which is the skipped screen read. On top of that comes a per-line saving that grows with the size of the station. At small stations, roughly 43 to 75 sq ft, a picker who knows the layout cancels that per-line saving almost entirely. At a large one they can’t, because no picker reliably holds 160 bin locations in memory.

Two consequences follow:

1. Lights make large stations behave like small ones, absorbing the cost of bigger setups. With lights on, each additional line on an order costs about the same wherever the picker is standing, roughly 3 to 4 seconds. Without lights, an extra line at the large station costs up to 13.5 seconds.
   
   
2. Lights flatten the experience gap. The two pickers at the large station had visibly different working styles. Without lights, Picker B was the quicker of the two, because she barely checked the screen and worked from memory instead. It cost her 3 completion errors in 26 orders, against 1 for the picker who kept checking. Turn the lights on and the trade-off disappears: both pickers hit the same times, around 20 to 21 seconds per order, with zero screen re-checks and zero errors between them. This reflects that the gain belongs to the station, it does not depend on who is working, which for any operation staffing temps at Peak matters more than the average.

What it does to a day

Folded into each station’s real order mix, using full order cycles from production data:

• Company 1’s large station, averaging 3.8 lines per order across 150 to 165 orders a day, gained 7% to 15%. That is 10 to 24 extra orders every day, from one station.
  
  
• Company 2’s compact station, running mostly single-line orders at 60 to 90 a day, gained 4%, or 2 to 4 extra orders.

At the large station, that works out to roughly 20 to 35 minutes of picker time freed every day. A small station adds only a few minutes on top, since its orders are mostly single-line. That gain can be taken as more orders out the door or as time given back.

Where the lights do less, and why that is useful

At Company 2’s compact station the per-line saving disappears completely. Every location is fewer than three steps away, the picker can see the whole station at a glance, and the screen can be re-checked without walking back, so timewise, working from memory costs almost nothing. It still costs accuracy, which is where the lights earn their place here.The time win is just the base saving, about 2.5 seconds per order, whatever its size. The error reduction did not shrink with the size of the station. Without lights the picker still stopped to re-check the screen about once an order, on average. With lights, not once.

Bin density turned out not to matter across the densities standard stations run, which is around one pick location per square foot. Packing the bins more tightly made no difference to picking times. Walking distance and station familiarity did.

What this adds up to

Six rounds at real stations, roughly 350 orders coded on video, and the findings point the same way:

• Picking time at the large station fell 31% to 36%. Every order, at every station, gets a base saving of 2 to 2.5 seconds from the screen read the picker no longer has to do.
  
  
• Capacity rose 7% to 15% at the large station, which is 10 to 24 more orders a day, and 4% at the compact one.
  
  
• Errors and search behavior improved at every station size, which is the one result that did not depend on how big the station was. Orders that ran clean from start to finish went from 31% to 92%.
  
  
• Output became easier to forecast, with order-to-order times 2.5 to 3 times less variable.
  
  
• Experience stopped deciding the outcome. Both pickers at the large station reached the same times once the lights were on.

The reason this is worth spending money on is not only the seconds. A station that produces more without anyone new being hired and without a bin being moved, that is capacity bought with hardware instead of headcount. Plus, it arrives without the ramp-up time a new hire needs.

The error side compounds that.

A mispick that reaches a customer costs a replacement shipment, a support conversation and the customer’s confidence, all of which sit downstream of a mistake that took two seconds to make. And because the lights close the gap between a veteran and someone new, the operation gets less fragile exactly when it is under the most pressure, which for most warehouses equals to Peak Season.

Packing is the other half of the cycle, now 40% to 50% of it at every station measured. But that issue already has an answer: Tap-to-Pack, ShipHero’s hardware controller that takes about 8 seconds off every order.

Lights on the pick side and a controller on the pack side act on different parts of the same cycle, increasing the overall efficiency.

‍(*) A note on names: the companies whose stations were tested are referred to here as Company 1, Company 2 and Company 3. Their operations were the real-world environment for the measurements, not a research partnership, so we describe their stations without identifying them.