Nobody told them it had to be right

In 2013 Target opened in Canada. Two years later it left: 133 stores closed, about 17,600 people out of work, and a pre-tax loss of US$5.4 billion in a single quarter, most of it writing off the Canadian business.

There were many reasons. Prices, the range on the shelves, and the speed of the whole thing all played a part. Target’s own chief executive said they had taken on too much, too fast. But one reason sits underneath the others, and it is the one worth learning from, because it is so ordinary.

What happened

Before the first store opened, the details of roughly 75,000 products had to go into SAP, the system that would run the supply chain. Each product had dozens of fields: dimensions, weight, how many fit in a case, the supplier’s code. That system fed everything else: forecasting, replenishment, the distribution centres.

The work fell largely to junior staff on a fixed timetable, typing in what suppliers sent them. According to the Canadian Business account, built from interviews with former employees, there was no talk of accuracy, and the system had no checks to catch a mistake as it was typed. One former employee put it plainly: “nobody told them it had to be right.”

When a team was sent to investigate, they found dimensions in inches instead of centimetres, measurements in the wrong order, the wrong currency, missing fields, and typos. They estimated the information was accurate about 30% of the time. In Target’s US business the figure was 98 to 99%.

The consequences were physical. Shipments arrived in case sizes the warehouse system didn’t know existed. The automatic replenishment relied on exact product and shelf dimensions, so it was switched off at the first test stores, and staff walked the aisles checking shelves by hand. Shoppers complained publicly about empty shelves. The tool that would have stopped bad data at the point of entry went in during 2014, by which time the stores were open and the reputation was set.

What was actually wrong

Nothing here was hidden. Every one of those rules was known before a single record was typed: dimensions are in centimetres in Canada, prices are in Canadian dollars, a barcode has a fixed number of digits, a case holds a whole number of items. The system was being filled with data that broke rules everybody already agreed on.

The problem was not that the data was bad. Data entered quickly by inexperienced people under pressure is always bad at first. The problem was that nobody measured how bad, while it was still cheap to fix.

What a check would have shown

Testing records against their own rules is not a research project. A count of products whose dimensions only make sense in inches, or whose currency is not the one the business trades in, is a query that runs in minutes. Run in the first week of data entry, the 30% figure would have been a warning with months still on the clock. Found by an investigation team, it was a post-mortem with the stores already open.

That is the whole of the argument for checking before you act. The check is small. The moment it is done is everything.

Before your own version

If you are about to load, move or rebuild a large set of records, these are the questions worth answering before the deadline, not after it:

• What are the rules? Units, currencies, formats, required fields. Write them down. Most of them already exist in the system’s own settings.
• Who is checking the records against them, and how often? If the answer is “the people typing them in”, nobody is.
• What does “good enough to go live” mean as a number? Target’s US business ran at 98%. Canada was at 30, and nobody knew until a team went looking.

Sources: Joe Castaldo, “The Last Days of Target”, Canadian Business, January 2016; Target Corporation announcement of the Canadian exit, 15 January 2015, as reported by the Associated Press.