The day Helen was raped it was cold. She remembers exactly what she was wearing when her ex-boyfriend knocked at the door, wanting to come in. It was the same thing she always wore while studying at home in winter - a hooded sweatshirt, trackpants, socks.

"Why are you here," she asked. They'd broken up about a week earlier, after a short relationship. He'd been mean and controlling. She had two young children. She decided it was never going to work.

Unfortunately, her ex didn't see it the same way. He'd been texting her incessantly since the break-up, and now he wanted to talk in person, he said. He was quite aggressive about it. Reluctantly, Helen let him in.

Three years on, she still has flashbacks to the assault - of him pushing her down, the pain. Afterwards, while collecting her youngest child from kindergarten, she broke down crying. The teacher advised her to call the police, but all Helen wanted was a shower. Standing sobbing under the water, she scrubbed her thighs so raw the skin came off.

When she phoned the police that night, they booked a medical check straight away. Doctors were able to collect DNA the next day. The bruises on her legs matched where her attacker had forced her against the metal bed. She had other, internal, injuries as well. Detectives examined her bedding, her clothes. Helen gave a long video interview.

Police believed her, Helen says. But her case never got to court.

During questioning, the man's lawyer said they planned to argue Helen had opened the door in skimpy clothing, that she seduced him. They also wanted to use her sexual history against her, the fact that she had two children to two different fathers, one at age 19.

With DNA effectively ruled out because of the recent relationship, the case would largely come down to Helen's word against her attacker's, police told her. By then, Helen was suffering from post-traumatic stress disorder, including anxiety and insomnia.

"I was struggling mentally. Police thought I would be a mess in court," she says. "They thought I would be too weak ... and that we would lose."

Lacking the necessary evidence, police closed the file. Somewhere out there is Helen's assailant, an untried rapist. In official statistics, her case is considered "unresolved".

Helen's story is not isolated.

A Herald investigation found 2400 reported violent sex crimes went "unresolved" in 2016. Each unresolved case means police believe an assault occurred, but an offender was never charged and taken to court for that crime.

Since 1994, official data says almost 14,000 aggravated sexual assaults in total were unresolved. However, the Herald can reveal that, because an unknown swathe of sex crimes were categorised in a way that meant they were effectively removed from statistics, the true number of unresolved cases over the past 25 years is likely to be thousands more.

According to current police data analysed by the Herald, as of 2016 up to 80 per cent of reported aggravated sexual assaults go unresolved. For the crime "male rapes female 16 and over", that number is even higher, at 85 per cent. Rape cases are four times less likely to go to court in comparison with other types of physical assault, where only 24 per cent of offences are unresolved.

Official statistics say the rate of unresolved violent sex offences is higher now than at any time since records began. For most of the 1990s, and early 2000s, the unresolved rate hovered around 55 per cent. But after 2007 the rate began to climb and climb, so that when plotted on a graph the data looks like the steep, jagged rise of a mountain incline. The empty space below the incline represents 13,914 victims left in limbo, their stories unheard, their attackers untried.

On the face of it, one could assume the driver behind the spike in unresolved cases was the overall increase in reporting rates. Data shows reporting has also climbed steadily since last decade. But while more reporting might account for some of the upward swing in unresolved cases, it doesn't tally with such a sharp, rapid change. Instead, the more likely explanation is that the true rate has simply been misreported all along.

During its investigation, the Herald discovered that historic police data on sex crimes is deeply misleading. For at least 20 years since collection began in 1994 police were incorrectly coding some sexual assaults as "no crime". Where cases had limited evidence, or victims were intoxicated, or consent was unclear, instead of recording the incident as a "K6 - crime reported" police would instead label it "K3 - no offence disclosed" or "no crime".

Documents released by police to the Herald after Ombudsman intervention suggest the inappropriate use of the "no crime" code in upwards of 15 per cent of cases. Not only did this distort crime statistics to seem as though fewer sexual assaults were being reported, but it kept the "unresolved" rate artificially low for years, because K3 cases disappear from official counts.

Worse, because the catch-all K3 code was also used for the small minority of fake claims, academics say it led some police officers to conflate the two totals, resulting in the incorrect belief that a huge proportion of women were liars, and not to be trusted.

Police documents show the K3 practice continued despite Victoria University criminologist Jan Jordan highlighting the issue in a 2004 research paper, and the issue arising again in a 2009 study for the Ministry of Women's Affairs. In fact, while its use began to decline after 2007, the inappropriate use of K3 wasn't acknowledged by police until 2013, when an informal review was commissioned into 279 sex cases coded "no crime", in which the reviewer said its overuse should cease immediately.

According to the Herald's findings, the inappropriate use of the K3 code means that in addition to the thousands of people with official "unresolved" cases like Helen's, there could have been at least 2300 more men and women from the 20 years to 2014, who, despite reporting aggravated sexual assaults to police, never had it recognised they were a probable victim of a sex crime. If you include non-aggravated sex assaults as well - which the Herald did not - that number would be thousands higher again.

"That confirms everything we've seen in the community," National Collective of Rape Crisis spokesperson Andrea Black says.

"It's left so many people traumatised. People end up taking sides, saying 'she lied because police aren't investigating, she's crazy, she's chaotic.'"

Police, who have spent the past 10 years improving the way they deal with sexual assault following a Commission of Inquiry into Police Conduct, say they are no longer coding sex crimes they can't resolve otherwise as "no offence". Just 3.8 per cent of adult sexual assault cases designated K3 in 2017, police said.

Jan Jordan, who recently reviewed three months of police files from 2015, say her preliminary results suggest detectives are now truly reluctant to use the code for reported sex crimes. Compared to 2004, files are much more detailed, with more work put in before any decisions about prosecution are made, Jordan said.

Additionally, police have introduced a new recording system, with codes for outcomes such as "not pursued" or "investigation suspended", allowing officers to add nuance when they close a file, and for more detail in official data sets. They have also educated investigators about compliance with reporting standards.

What that all means, is that for the first time, we are probably seeing an accurate picture of the true rate of unresolved sexual assault cases in New Zealand.

It comes amid a sea-change moment for the campaign against sexual violence, where dozens of women have named their high-profile abusers as part of the #MeToo movement, where judges who blame victims will have their rulings overturned, and where the stigma and shame of sexual violence is being replaced with a mood of empowerment and change.

Much of the current debate centres around low reporting rates, and the need for victims to come forward, to follow the right channels, to be brave. However, what the Herald's investigation has found is even when victims do the "right thing" there is a very low chance their attacker will even get to court, let alone be convicted, despite police certainty their victims are telling the truth.

"The public awareness and international awareness is awesome but brings with it people who say I'm going to do something about my assault, I feel strong, society has changed... and then they go into that process and they wait in limbo," Black says. "They speak up and take action and then it's like: bam, stop, block, wait."

For women like Helen, it leaves a deep sense of frustration with the system that is supposed to help them.

"It's flawed," she says. "And that's why so many women don't go through with it."

"The dectective said he believed me"

In New Zealand, the crime of rape carries a maximum penalty of 20 years' imprisonment. In terms of severity, it is second only to murder, and the most serious methamphetamine offices, on the sentencing scale. Unlawful sexual connection has a maximum penalty of 18 years. Between them, the two categories make up the majority of reported cases of aggravated sexual assault.

While some less serious offences are able to be resolved out of court, alternative resolutions do not apply to sex crimes. There are no warnings, no mediation. Rape either goes to court, to be heard in front of a jury, or it doesn't. Aside from the very small proportion of cases now deemed "no crime", almost all of the cases which don't make it are deemed "unresolved".

To find out why so many of New Zealand's reported sexual assaults do not reach court, the Herald spent six months speaking with victims, police, sexual violence advocates, lawyers, police and academics.

It reviewed studies and government reports, including Dame Margaret Bazley's 2007 Inquiry into Police Conduct, and the subsequent monitoring updates. It found the reasons cases aren't put forward for prosecution not only involve police practices, but in large part stem from inadequacies in the wider justice system, and the deeply entrenched views about rape still prevalent within society itself.

Ten years ago, it is almost certain that - fair or not - much of the blame for the unresolved cases would have been levelled squarely at police investigators.

However, sexual violence experts and advocates almost unanimously agreed that since Bazley's inquiry into police culture, there has a been significant change in the way police handle reported sexual assault.

The inquiry, commissioned after Louise Nicholas accused four police officers of rape in 2004, found major failings with the way police handled sexual assault cases, and a "culture of scepticism" about rape complainants.

It prompted an overhaul of police practice and policy, including better training, specialist sexual assault investigation teams, better communication with victims and more rigorous oversight of sexual assault files.

The changes were so widespread that the final police report, "A Decade of Change", released on the 10-year anniversary of the inquiry last year, ran to 40 pages. It covered each of Bazley's initial 47 recommendations, with specific details about each point.

For example, police noted they now have 105 trained investigators who specialise in investigating adult sexual assault complaints. In 2007, there were none. A "quality assurance" framework ensures files are audited both at district and national level. The K3 code was not mentioned in the report, although the Bazley inquiry is thought to have been partly behind the decline in its use.

The improvements have been noticed by victims as well as critics.

Grace, who was sexually assaulted as a teenager, said the first time she tried to report it to police in the 1990s, she was told her case would never get to court, and not to bother with a formal statement.

But when Grace complained again in 2015, prompted by news reports of allegations against high-profile entertainers such as Bill Cosby and Jimmy Saville, police doggedly tracked down her offender and charged him. While the charges were later dropped because of issues with the historic nature of the crime, she said the police had done a good job.

"The detective said to me, 'I believe you'," Grace said. "Even though there wasn't the outcome I wanted it was so much better that before, to be believed, to have [her abuser] put in a cell even for one night."

Victoria University criminologist Jan Jordan, one of the country's most experienced sexual violence researchers and an expert on police practice, said there had been a huge change in investigators' attitudes.

Her current study, a review of sexual assault files from 2015, was supposed to examine a year of case work, but she found the files so detailed she only needed three months' worth to see trends emerge.

While detectives still struggled at times with complex cases of intimate partner violence, or with victims who were intellectually disabled, they were much better at keeping an open mind, Jordan said.

"Cases which in the past would signal red flags, for example cases where victims were intoxicated, or suffered mental illness … police now suspend any judgement they might have while doing the investigation."

Jordan noticed discretion around decisions to prosecution were no longer in the hands of a single officer, and files were reviewed by more senior detectives. On several occasions, detectives were sent back to re-interview witnesses, in a bid to extract the most evidence possible.

"They really are trying a lot harder before making a decision on court," Jordan said.

A long process of generational change

Despite improvements, it is clear police are not perfect yet. Following publicity about the end of the inquiry's monitoring period last year, and the subsequent publication of a video series about "how to report a rape", a flurry of women came forward to complain to advocates and media about the way their cases were treated by police.

Rape Prevention Education executive director Debbi Tohill said five women approached her during May and June 2017 with complaints about unresolved sexual assaults. She passed their complaints to police.

Several more spoke to Stuff journalist Tony Wall, saying police hadn't tried hard enough to collect evidence, or had delayed work on their files.

Monique, a 22-year-old university student who complained to police that she may have been sexually assaulted in Dunedin, told the Herald she felt her case was one of those closed without enough investigation.

Records show that, at first, Dunedin police didn't even record her complaint as an offence, instead treating it as "information", which Monique believes was because she had limited memory of what had happened, and wouldn't use the word "rape".

"What had happened didn't match this idea that I'd been brought up to think what rapes looks like. My perception was only of violence, of stranger danger," she says. "My justification was it's just a bad experience, it's not that bad."

Monique laid the complaint after she participated in a blind date organised by the Otago University magazine Critic in 2015. The evening spiralled out of control, and Monique was so intoxicated she barely remembers getting to her date's house, or how she got into his room.

The first time she complained, police noted Monique "agreed" sex had been consensual, even though she says she doesn't believe it was. During university holidays, she complained again, this time at the Whanganui police station, where detectives did a full video interview - including 18 pages of handwritten notes - before referring the case back to Dunedin. The file was later closed as unresolved because of insufficient evidence, even though the Dunedin detectives didn't follow up advice from Whanganui to interview the taxi driver or get CCTV from outside the bar, or to talk to extra witnesses.

A lawyer who reviewed the file for the Herald noticed several clear red flags with Monique's file, such as her memory of climbing out a window at one stage, and saying "no" when asked for oral sex. Equally, her assertion she may have been anally raped was never looked into, after the alleged perpetrator denied it happened. Police also seemed to have delegated some aspects of the investigation - like interviewing the other men at the flat, some who allegedly exposed themselves to Monique - to the university proctor.

Police did not want to discuss specific issues about Monique's case, but said the investigation was "thorough" and it was reviewed appropriately. The university has since moved to set up a centre for sexual violence, and updated its sexual misconduct policy. Critic magazine now sends advice to students participating in the date to warn: "If your date is drunk, they cannot consent to sex. Any form of nonconsensual sexual contact is assault."

Two years later, Monique is still bruised by the experience with police, and suffers anxiety. Her graduation has been delayed twice. While the university has been flexible with her enrolment, its refusal to take disciplinary action against any of the students involved has added to her sense of trauma.