“The computer only wants to buy one,” said Bill, looking at a printout.
“But we don't have any,” countered Nick.
Bill consulted the printout. “Yes, we don't have any,” he concurred.
“So then we'll have one,” Nick pointed out.
“That's how many the computer wants.”
“You can't have one sawhorse in stock,” Nick said.
“They are sold in pairs.”
“But the computer ....”
“Ask the computer how you hold anything up with one sawhorse!”
This was a typical conversation that took place every week since we had put our hardware stores online 10 or 12 years ago. Bill would have the computer generate a suggested order based on sales history, inventory stock levels, desired gross margin return on investment, and other factors. Then, Bill would review the order with the various department managers and the fun would begin.
Often, like the sawhorses, the computer would suggest quantities that just didn't make sense. Other times, factors that affect the sales history couldn't be programmed into the equations creating the suggested order.
“The computer wants to buy 223 railroad ties,” Bill told Ken, the yard manager, one day.
“Please don't buy 223 railroad ties,” Ken begged.
“The computer sees a strong sales pattern developing.”
“We don't need 223 railroad ties.”
“We sold 430 last month,” Bill pointed out.
“I know,” said Ken.
“And 193 the month before that.”
“Stock levels are zero.”
“So the computer wants to buy 223 railroad ties ....”
“Please don't buy 223 railroad ties!” Ken pleaded.
The problem with the ties was that one customer had bought hundreds of them through a special order for a large commercial project. The project was done, and we'd be lucky to sell another 10 ties before the end of the year.
We worked hard to address these problems. We assigned order multiples to items such as the sawhorses, so the suggested order from the system would reflect the multiples in which the items were bought and sold. Special order items were tagged to be ignored by the suggested order.
But there were always issues.
“We need pallets!” Bill screamed once, waving a suggested order in his hand. “No, truckloads!”
Everyone ignored him.
“Did you hear me? TRUCKLOADS!!!!”
By this point, we knew the system well enough to anticipate the hang-ups. No matter how carefully we refined our inventory tracking methods, we could never fully depend on the computer-generated suggested orders. Every year, with further refinements and new features and functions added to the software, the proposed orders became more and more helpful. But the human element, in our experience, could not be totally removed from the ordering process.
“The GMROI and sales history indicate this could be our best inventory item!” Bill said another time, with yet another suggested order printout in his hand. “Get rid of the lumber, the computer wants us to CORNER THE MARKET ON ICE MELT!!!”
He was partially correct. We had sold an amazing amount of ice melt. We had had the worst winter in years, with record cold and frequent snow and ice storms. So the computer had a point.
“I'm calling the owner!” Bill raved. “I need authorization for six—no, SEVEN—-truckloads of ice melt!!!”
We continued to ignore Bill's screeching, working around him as he waved the suggested order and shouted his predictions.
We knew something that the computer didn't—it was July.