When the Target Becomes the Goal: 5 True Stories of Metrics Gone Wrong
Set a target and people will hit it. That is the point of a target. The trouble starts when the number that stands for the result becomes more important than the result itself.
Over the years, a rat bounty, a bank’s sales goals, a car’s emissions test, a hospital wait-time target and a school test score have all gone the same way. People found the cheapest route to the number, and the number stopped telling the truth.
Here are five well-documented cases, each backed by a public source, and what they mean for any team that runs on dashboards, bonuses or KPIs.
Hanoi’s rat-tail bounty that bred more rats

In 1902, rats were spilling out of the new sewers of French colonial Hanoi, and plague was a real fear. City-hired rat hunters were not making a dent, so the authorities opened the hunt to everyone with a bounty of one cent per rat. To avoid piles of dead rats, they asked for just the tail as proof.
It looked like a success. According to historian Michael Vann, whose archive research is the source of the story, 7,985 rats were killed in the last week of April, and the daily total hit 20,112 on 21 June. Then people began spotting healthy rats with no tails. Hunters were cutting the tails off live rats and letting them go, so they could breed and produce more tails. Health inspectors also found rat-breeding operations on the outskirts of the city, and there were reports of rats being smuggled in. The bounty was scrapped.
Who is the poor guy counting all these rat tails?
Michael Vann, historian, quoted by Atlas Obscura (2017)
Takeaway: A bounty paid for a proxy (tails) invites people to produce the proxy, not the result (fewer rats). Before you launch a bonus or reward, ask how someone could earn it without delivering what you actually want.
Wells Fargo’s sales targets and secret accounts

On 8 September 2016, the US Consumer Financial Protection Bureau (CFPB) fined Wells Fargo $100 million, the largest penalty it had imposed up to that point. The bank had set up compensation incentives and sales targets to encourage “cross selling”, getting existing customers to take more products.
According to the CFPB, employees hit those targets by opening accounts customers had not asked for. By the bank’s own analysis, employees opened about 1.5 million deposit accounts and applied for about 565,000 credit card accounts that may not have been authorised. They moved customers’ money into new accounts without consent, sometimes triggering overdraft or insufficient-funds fees, and created phony email addresses to enrol people in online banking.
The CFPB did not condemn cross-selling itself, which it called a common and accepted business practice when it grows out of good service. Its complaint was that these incentive programs pushed staff to sign customers up without their consent, and that the bank “failed to monitor” them with adequate care. Wells Fargo was ordered to refund affected customers, and paid separate penalties to the Office of the Comptroller of the Currency and to the City and County of Los Angeles.
Today’s action should serve notice to the entire industry that financial incentive programs, if not monitored carefully, carry serious risks that can have serious legal consequences.
Richard Cordray, CFPB Director, 8 September 2016
Takeaway: A target needs a check that the activity behind it is real. Watch downstream signals such as complaints, reversals and refunds, not just the count you pay people to raise.
Volkswagen’s engines that passed only the test

Sometimes the target is not a bonus but a test. In September 2015, the US Environmental Protection Agency (EPA) issued a Notice of Violation to Volkswagen. Certain diesel cars had software built to detect when the vehicle was undergoing emissions testing and to turn full emissions controls on only during the test.
On the road, those cars emitted nitrogen oxides at up to 40 times the standard, according to the EPA. The agency’s later civil case covered approximately 590,000 diesel vehicles from model years 2009 to 2016. In January 2017, Volkswagen agreed to plead guilty to three criminal felony counts and pay a $2.8 billion criminal penalty, plus $1.5 billion in separate civil resolutions.
The result was a product that looked perfect to the measurement and behaved differently in real life. The measurement itself had become the thing being optimised.
Takeaway: If a system behaves well only when it knows it is being measured, you are not in control of it. Test products the way customers really use them, and add checks nobody can predict.
Phoenix VA’s wait times that looked too good

In 2014, the US Department of Veterans Affairs Office of Inspector General (OIG) reviewed the Phoenix VA Health Care System after allegations reported to its hotline. The system was working to a 14-day target for scheduling primary care appointments, and its director had a “Wildly Important Goal” to improve access.
The OIG sampled 226 new-patient primary care appointments. VA data reported by the Phoenix system showed those veterans waited 24 days on average, and 43 percent waited more than 14 days. The OIG’s review of the records found an average of 115 days, with about 84 percent waiting more than 14 days.
In interviews with 79 scheduling employees, 30 said they used the wrong “desired date” and produced a false zero-day wait, and 11 said they “fixed” appointments with waits over 14 days by rebooking them with a later desired date. The OIG also identified more than 3,500 additional veterans waiting for appointments who were not on the official electronic wait list, many of them on unofficial lists. It opened investigations at 93 sites after further allegations and confirmed wait-time manipulation was prevalent. The report documented serious lapses in care, while stating it could not conclusively say that delayed care caused the deaths it reviewed.
Takeaway: If staff can edit the date, status or category a metric reads from, the metric will drift. Lock those fields, keep an audit trail, and compare reported figures against a sample of raw records.
Atlanta’s test scores that were too good to be true

In July 2011, special investigators appointed by Georgia’s governor released a report on Atlanta Public Schools. They found cheating on the 2009 state standardised test in 44 of the 56 schools they examined, and identified 178 educators as involved, including 38 principals. Of the 178, 82 confessed. The report traced misconduct back to 2001.
The investigators said the cause was “primarily” the pressure to meet targets in a data-driven environment. They also found that clear warnings had been raised as early as December 2005 and ignored, and that a culture of fear stopped teachers from speaking up. Thousands of children were harmed, the report said, because inflated scores denied them remedial help.
I think the overall conclusion was that testing, and results, and targets being reached became more important than actual learning on the part of children.
Nathan Deal, Governor of Georgia, 5 July 2011 (via Georgia Public Broadcasting)
Takeaway: When a single score decides careers or bonuses, verify it independently. Treat early warnings as seriously as late ones, and make it safe for people to report problems.
What this means for your business
None of these organisations set out to fail. Each picked a number to represent something real, then rewarded the number. The same pattern shows up in everyday companies: sales quotas, on-time delivery rates, ticket-closure counts, stock accuracy, days-to-pay.
A few habits make metrics much harder to game:
- Reward outcomes, not proxies. Ask what the number stands for, and how someone could raise it without delivering that.
- Pair every target with a counter-metric. Sales volume with cancellations, closed tickets with reopened tickets, speed with quality.
- Audit the data behind the dashboard. Sample raw records regularly and compare them with the reported figure.
- Lock the fields a metric depends on. Dates, statuses and categories should not be freely editable, and every change should leave a trail.
- Test the way things really run. Add unannounced checks and review exceptions, not just averages.
- Make it safe to raise a flag. The warnings in these stories existed. Nobody was rewarded for acting on them.
At SOD Infotech, we help businesses build reporting that traces back to the original transactions, so the numbers on a dashboard can be checked, not just trusted.
Sources
- Hanoi — Atlas Obscura, Shay Maunz, "The Great Hanoi Rat Massacre of 1902 Did Not Go as Planned" (6 Jun 2017)
- Hanoi — Michael G. Vann, "Confessions of a Rogue Historian: Why I Wrote a Graphic History of Colonial Hanoi," H-France
- Wells Fargo — CFPB press release, "CFPB Fines Wells Fargo $100 Million for Widespread Illegal Practice of Secretly Opening Unauthorized Accounts" (8 Sep 2016)
- Volkswagen — US EPA, "Learn About Volkswagen Violations"
- Phoenix VA — VA OIG, Interim Report, Report #14-02603-178 (28 May 2014)
- Phoenix VA — VA OIG, Review of Alleged Patient Deaths, Patient Wait Times, and Scheduling Practices at the Phoenix VA Health Care System, Report #14-02603-267 (26 Aug 2014)
- Atlanta — Georgia Public Broadcasting, Maura Walz, "Widespread Cheating Found In Atlanta Schools" (5 Jul 2011)
- Atlanta — NPR, Eyder Peralta, "Governor’s Report Details Widespread, Organized Cheating In Atlanta Schools" (6 Jul 2011)
