Target Canada: empty shelves, full warehouses and data nobody asked to get right

February 2013. Target Canada's head office in Mississauga, on the outskirts of Toronto. The first stores were a month away from opening, and the subsidiary's president, Tony Fisher, was meeting with his leadership team every day to find out whether they were ready. That afternoon there were about a dozen people in the room. The news, according to the reconstruction that Canadian Business magazine published three years later, was rarely good: it was hard to get merchandise out of the distribution centres and onto the shelves, the checkout system was glitchy, and the technology that governed inventory was new and no one seemed to fully understand how it worked [1].

Two of those present expressed reservations about opening on the planned date. Neither one asked, out loud, to postpone. A former employee summed it up for the magazine: "Nobody wanted to be the one to say, 'This is a disaster'" [1]. By the end of the meeting, the decision was made: the stores would open on time. The same former employee later judged it without hedging to the magazine: "That was the biggest mistake we could have made" [1].

Less than two years later, on January 15, 2015, Target announced it was leaving Canada. It would close its 133 stores, where some 17,600 people worked [2].

This is the story of how one of the most admired chains in North America had full warehouses and empty shelves at the same time. It is not a story of bad luck or of a bad product. It is, among other things, the story of a piece of data: the data that described every item, every case and every shipment, and that was entered against the clock without anyone asking that it be right. It is for whoever runs commerce, operations or technology in retail, and for whoever signs off on a launch date.

One clarification before we start: Hábil did not take part in this case. We tell it from what Target reported publicly and what the serious press reported.

The clock started ticking in 2011

On January 13, 2011, Target announced that it would pay 1,825 million Canadian dollars for the leases of up to 220 stores of Zellers, a Canadian discount chain. It planned to open between 100 and 150 stores in 2013 and 2014 [3]. Later, the goal became more ambitious: 124 stores in a single year, 2013 [4].

Buying the leases raised the pressure, financial and operational, to open fast. Canadian Business describes it as a commitment to opening stores as quickly as possible [1]. One former employee put it in words that sum up everything that followed: "From the very beginning, there was a clock that was ticking. And that clock was absurd" [1].

Timelines were compressed across the board. Building a distribution centre from scratch can take a few years; Target was going to build three in less than two [1]. And it had to decide on the piece that connects everything: the system that makes it possible to order merchandise from suppliers, process it in the warehouses and get it onto the shelves.

A new brain, with no memories

In the United States, Target ran on its own systems, fine-tuned over years. Adapting them to Canada required work: another currency, French characters, another country. There was no time. The team chose a commercial enterprise resource planning system (ERP), from SAP, which Canadian Business describes as "the gold standard" in retail and as the brain of the business: it stores the data for every product and hands it to the other systems: the one that forecasts demand and replenishes stock, and the one that manages the warehouses [1].

Other Canadian chains had struggled to implement that kind of system because of a well-known problem: converting the data from the old system to the new one, a process in which it is easy to make mistakes. Target believed it was safe from that. According to the magazine, its reasoning went like this: "There was no data to convert, only new information to input" [1].

That sentence is the heart of the story. Target had no old data to clean up. It had to create all of it, from scratch, for roughly 75,000 products, and each product could have dozens of fields: manufacturer, model, barcode, dimensions, weight, how many units fit into a shipping case [1]. That information is called master data: the record for each item that every other system consults to decide what to order, where to store it and how much fits on a shelf.

The data entry fell, to a large extent, on young merchandising assistants, recently hired, with a few weeks of training and under pressure from the calendar [1]. The data came from suppliers, and in the industry information from vendors is notoriously unreliable; the assistants did not always have the experience to challenge it [1]. "There was never any talk about accuracy," a former employee recalled. "You had these people we hired, straight out of school, pressured to do this insane amount of data entry, and nobody told them it had to be right" [1].

And the system, at that point, had no safety net: it did not warn anyone when an entered value was invalid [1].

The 30%

In 2012, when orders for the launch began, strange things started to happen. Products from overseas did not fit into containers as expected. Tariff codes were missing. Merchandise arriving at a distribution centre could not be processed to be sent to a store. Other items did not fit on the shelves [1].

A team investigated and found the cause. Dimensions were in inches rather than centimetres, or in the wrong order: width by height by length instead of length by width by height. Sometimes the currency was wrong. There were vague descriptions, empty fields, typos. "You name it, it was wrong," said a former employee. "It was a disaster" [1].

That team estimated that the information in the system was accurate about 30% of the time. In the United States, the same measure was between 98% and 99% [1]. The figure should be read for what it is: an internal estimate, cited by the magazine, not a published audit. Even so, it describes the size of the problem.

The response was drastic. In the fall of 2012, Target halted its entire merchandising division for a week to check by hand every piece of data for every product with its suppliers: one buyer might have 1,500 products and between 50 and 80 fields to verify for each. They called it "data week." It stretched into two. There was also test data, entered when the system was set up, that was still there and confusing the system [1]. The week served to remove the worst errors. It was not enough to attack the cause: data kept coming in without an automatic validation to stop it.

March 5, 2013: the doors open

On March 4, 2013, Target announced the opening of three pilot stores in Ontario, in Guelph, Milton and Fergus. They would open the next day, at 8 in the morning. The purpose, according to the announcement, was to "prepare systems, train team members and determine operational readiness" before moving on to the rest of the 124 stores planned for that year [4]. Shortly after, according to Canadian Business, another 21 would follow [1].

There were lines from early on. But complaints about empty shelves soon reached social media. "Target in Guelph, please stock up and fill the shelves," wrote one shopper quoted by the magazine. "How can I or anyone purchase if there is nothing left for me to buy?" [1]. The company told the press that demand had overwhelmed it. According to the magazine's reconstruction, the reality was different: data quality problems were still holding back the supply chain, and there was no time to attack the cause before opening the next wave [1].

The automatic replenishment system, which calculates when a shelf needs to be refilled, needed the exact dimensions of every product and every shelf. Many were still wrong. From the pilot stores on, Fisher ordered it switched off and replenishment done by hand: employees walking the aisles, checking shelf by shelf. It was not switched back on until later that year [1].

Meanwhile, in the warehouses

This is where the story takes its hardest-to-believe turn. While customers were finding empty shelves, Target's distribution centres were bursting [1].

There were two reasons. The first: Target ordered far more than it could sell. Its forecasting system needed years of sales history to be useful, and it had none; purchasing was based on optimistic projections made in the United States [1]. "Once you signed up to do 124 Zellers locations, it felt like there was a point where it's like we have to assume sales will be good," said a former employee. "It's very backwards" [1].

The second reason is the one that matters for this article. The warehouse system and the central system did not understand each other well, and the data was caught in between. The magazine explains it with an illustrative example: someone at head office orders 1,000 toothbrushes and records that they will arrive in 10 boxes of 100. The supplier sends them in 4 boxes of 250. For the warehouse system, that shipment did not exist. It could not be processed. It was set aside in what was called the "problem area" [1].

Hang-ups like this happen at any warehouse, the magazine concedes. At Target Canada they happened with alarming frequency [1]. By the fall of 2013, the three distribution centres, some four million square feet in all, were overflowing. Tractor-trailers sat waiting their turn to unload. The company rushed to rent additional warehouses, and the decision about which merchandise to send to each one was improvised, so it was hard to find later. "It was like a massive black hole," said a former employee [1].

Think about what that sentence means for an executive. The merchandise existed. It was paid for. It was in the country, a few kilometres from the stores that needed it. And nobody could say for certain where it was.

What head office saw

There is a detail that makes the story even more unsettling. In Mississauga, the merchandising system often said that products were in stock. Then the calls came from the stores asking why they did not have them. "We almost didn't see what the customer was seeing," a former employee recalled. "We'd look on paper and think we're OK. Then we'd go to the store, and it's like, 'Oh my god'" [1].

In February 2014, Target published its results for the previous year. It had 124 stores open in Canada, and the Canadian segment had lost 941 million dollars before interest and taxes on sales of 1.3 billion. In the last quarter, its gross margin was 4.4%, which the company attributed to its efforts "to clear excess inventory" [5].

That same year, a small group of employees discovered something that helped explain why head office saw stock that the stores did not have. Within the replenishment system there was a feature that alerted the distribution centres when a store ran out of a product. Some head-office analysts, young and evaluated on their in-stock percentage, were turning it off on purpose for their products: with the alert off, the system did not report the shortage and their numbers looked good. Two people involved in the discovery allow that human error may also have played a part, because the system was new to everyone [1]. When the new leadership ordered the feature fully activated, the true stock levels appeared for the first time: "pitifully low," in the magazine's words. Afterwards, a tool was built that reported when that alert was switched on or off, and whether there was a legitimate reason [1].

In 2014, the automatic validation that prevented invalid data from being entered, such as a barcode one digit short, finally arrived as well. A former employee who was involved in it put it this way: "This happened very late in the game" [1].

January 15, 2015

By the 2014 holiday season, according to former employees, the operation was finally working better: there was a year of sales history, the data was cleaner and products were arriving on time more often [1]. It was not enough.

On January 15, 2015, Target announced that it would stop operating in Canada. That same morning, its subsidiary filed for creditor protection (under the CCAA, the Canadian law that governs it) before the Ontario Superior Court of Justice [2][6]: a court-supervised restructuring, not a bankruptcy. Target's CEO, Brian Cornell, explained the decision: "we were unable to find a realistic scenario that would get Target Canada to profitability until at least 2021" [2].

That day, the company estimated losses from discontinued operations of about 5.4 billion dollars before taxes in the last quarter of 2014 [2]. In its annual report, it recorded 5.1 billion dollars in pretax charges for the deconsolidation and exit; the loss from discontinued operations for all of fiscal 2014, net of taxes, was 4,085 million dollars [6]. It offered 70 million Canadian dollars for a trust to guarantee almost all of its employees in Canada at least 16 weeks of pay [2].

On April 1, Target Canada announced that the last of its 133 stores would close to the public on April 12; the three distribution centres and the head office had already closed [7]. Its subsequent quarterly report confirms it: as of May 2, 2015, the end of its first fiscal quarter, all stores were closed [8].

From the February 2013 meeting to the last closing, a little more than two years went by.

What this case teaches

It would be unfair to reduce Target Canada's fall to a single mistake. Canadian Business's own reconstruction points to several causes that added up: an overly ambitious opening schedule, leases that forced it to open fast, locations that were not always good, optimistic forecasts, prices that did not win customers over, a team with little experience facing the biggest crisis in the company's history [1]. The consulting firm that led the system implementation told the magazine that the implementation was successful and that an independent review concluded the firm had no connection to the problems described [1]. It is not the system that this article judges.

But among all those causes there is one that recurs in every scene: nobody could say where a case was, or whether a piece of data was true. And that is a lesson any retail operation can apply. Five ideas.

1. If nobody can say where an order or a case is, the problem is not the warehouse

This is the central lesson. Target's warehouses overflowed because of a combination of causes: too much was ordered, the forecasts had no history and, again and again, the shipment that arrived did not match the one the system expected [1]. The toothbrushes existed; for the system, they did not [1]. When the merchandise is physically there and the system does not recognize it, hiring more forklifts solves nothing. The problem is the data, and the end-to-end trail of that data: from the moment the product record is entered until the case reaches the shelf.

2. "Starting from scratch" does not mean "starting clean"

Target believed that, with no old data to convert, it would have no data problems [1]. It was the opposite: 75,000 new records, entered in a hurry by inexperienced people, without a validation to flag the error [1]. New data is not correct data. Correct data is data that someone validated, against a source, with a rule.

3. A green dashboard can be lying

Head office saw stock that the stores did not have. And part of that gap came from a switch turned off so the numbers would look good [1]. The lesson is not to distrust people; it is that an indicator that can be switched off without leaving a trace ends up switched off. What fixed the problem was not a reprimand: it was a tool that kept a record of who switched the alert on or off, and why [1].

4. Systems can say different things about the same case

In the magazine's example, the central system expected 10 boxes and the warehouse received 4 [1]. Each was right within its own records, and the case stayed in the problem area. It is the same phenomenon that shows up today in an omnichannel operation when e-commerce, inventory and the parcel carrier have different versions of an order. What gets lost is not the merchandise: it is the conversation between the systems.

5. The launch date is also data

In February 2013, two executives expressed reservations and neither asked to postpone [1]. The signals were in the room: merchandise was not leaving the distribution centres, the checkout system was glitchy and nobody fully understood the inventory system [1]. What was missing was shared evidence, impossible to ignore, that would turn a hunch into a decision. A clear trail does not guarantee that someone will say "let's wait"; it does make it harder to stay silent.

Five questions for your next committee meeting

1. If I ask today for the path of an order or a case, how long does it take someone to tell me where it is, with evidence? Time three real cases.
2. What percentage of our product records is complete and verified? If nobody has measured it, the number is not the one you think.
3. What happens when two systems disagree about the same order? Does someone find out in the moment, or is it discovered at closing?
4. Which indicators can be switched off or adjusted without leaving a trace? Who would know?
5. Does the next launch have a data condition for opening, in addition to a date?

If two or more answers are "I don't know," the risk is not in the warehouse.

Where to start this week

Ask your operations team for the list of everything currently set aside as a "problem" in your warehouses or in your systems: receipts that do not reconcile, orders on hold, shipments without confirmation. Pick ten cases. For each one, write down why it stopped: was it the case, or was it the data that described it? If most turn out to be data that did not match between two systems, you already know where to start, and it is not the warehouse.

Frequently asked questions

Did Target Canada fail only because of its data?

No. Canadian Business describes several combined causes: a very tight schedule, the purchase of leases that forced it to open fast, optimistic forecasts, uneven locations and a team inexperienced in the face of the crisis [1]. Data quality appears in almost every operational scene, but it was not the only cause.

Was the software to blame?

The magazine does not frame it that way, and the consulting firm that led the implementation stated that its work was successful and that an independent review found no connection between the firm and the problems [1]. What the report describes is badly entered data, without automatic validation, in systems that were new to the team.

What is master data, and why does it matter so much?

It is the record for each item: code, dimensions, weight, units per case, supplier [1]. The purchasing, forecasting, warehouse and store systems consult it to decide what to order, where to store it and how much fits on a shelf. If the record is wrong, each system makes the wrong decision with complete confidence.

How can shelves be empty while the warehouses are full?

At Target Canada two things coincided: too much was ordered because of optimistic forecasts, and the merchandise that arrived did not always match what the warehouse system expected, so it sat on hold without being sent to the stores [1]. There was inventory; there was no reliable way to move it or to know where it was.

Could this happen to a smaller Mexican operation?

Size changes the scale, not the mechanism. An operation that depends on several systems to move an order, from e-commerce to inventory, to the parcel carrier and to the invoice, can have orders and cases that exist and that no system, on its own, can explain.

What Hábil does: the trail of each order, across systems

Hábil works in the space between the systems of a retail operation: e-commerce, inventory, fulfillment, shipping and invoicing. It does not enter or clean up your master catalogue; what it does is show at which step a piece of data that does not reconcile stopped an order. The proposal is described in Order traceability for retail: type in an order number and see, on a single sheet, everything that happened to it, what was sent to each connected system and what each one answered, what is missing and who it depends on for it to move forward. When a system did not answer, or answered something that does not reconcile, it shows up flagged, instead of sitting in a "problem area" nobody looks at. It connects to the systems you already have, without replacing them, and it also monitors continuously: when an order gets stuck or a step fails silently, it alerts whoever is responsible. The scope, the systems that are connected and the controls are reviewed against your case.

If you want to dig deeper into why the trail gets lost between systems and what leadership should demand, read "The order is not lost; the trail is". And if the problems go beyond a single order, the page Retail omnichannel software describes how we approach them.

Request your first conversation, at no cost. It can start with a single order: pick one that cost you a week of emails and we will review together where the trail gets lost in your operation. Write to us on WhatsApp; to protect your customers, do not share order numbers or personal data through that channel.

Quotations appear in their original English. Those from former employees are verbatim from the* Canadian Business *report, which protected their anonymity. Dollar figures are U.S. dollars unless marked "Canadian."