Chicago’s bureau of detectives’ area technology centers – An exploratory case study of hidden values

Abstract

An exploratory case study was conducted on the Area Technology Centers (ATCs) in the Chicago Police Department. We investigated direct and indirect impacts of the organizational transformation on the homicide investigations. The analysis suggests that the ATCs might have had a role in improving the homicide investigation outcomes, and detective productivity. However, it is difficult to state categorically that the ATCs resulted in a systemic improvement on whether or not a case was actually solved or not. The study provides insights into how organizational transformation effects can be considered in the analysis data in such situations.

Introduction

In 2019, the Chicago Police Department’s (CPD) Bureau of Detectives (BOD) implemented Area Technology Centers (ATC) with the assistance of the University of Chicago’s Crime Lab (Cook and Berglund, 2021; Cook et al., 2023). The ATCs are specialized support groups in each of the BOD’s areas to help with the acquisition and processing of video and other digital evidence. The first ATC opened February 28^th, 2019 with the second and third ATCs opening August 1^st, 2019 (Cook et al., 2023).

The new units have trained specialists who acquire and process digital evidence as their main task instead of having the work done by detectives or other evidence processing units. The ATC introduction was an organizational transformation of reasonable importance as indicated by the Mayor’s and CPD Superintendent’s press release (City of Chicago, Office of the Mayor, 2019), stating the objectives behind the ATC introduction: “solve crimes faster and improve clearance rate” through the ability to “process videos faster from private surveillance cameras and cellphone footage”.

In 2021, a research project was initiated with the CPD BOD to review the ATC processes to better understand the contributions being made by the units. The methodology used was an exploratory case study approach, not descriptive or explanatory (Yin 2018). As noted by Yin, exploratory forms of case studies are useful when the situation is initially studied without any prior basis for investigating causality. This independent study was done in parallel to a formal review funded by the Department of Justice and conducted by the University of Chicago Crime Lab (Cook et al., 2023). The Crime Lab. evaluation focused on the period 2016-2020; analyzing video processing, the felony process, prosecutions, cases served by the ATCs, and fatal versus non-fatal shootings. The Crime Lab. study used a mixed-method approach incorporating multiple qualitative and quantitative analyses. In contrast, we used a slightly longer time window 2014–2021 focused on the initial stages of an investigation (e.g., evidence related activity in the first 48–72 hours of an investigation), the speed or trajectory of cases (e.g., time until the first evidence is inventoried, time until the first arrest is made), and understanding how the evidence profile and activity changed over the time period of 2014-2021. There was no formal qualitative component to our study and the Crime Lab. report is suggested as a key resource for comprehensive, descriptive insights. An overview of the Crime Lab. study is found in Appendix A.

We investigate how the evidence processing was potentially impacted by the ATC introduction by looking at the period of time following the homicide. As with any organizational transformation such as this, there are possible transitional effects to be aware, as well as the background story of what was happening before and during the transformation. The ATCs are supporting units and do not investigate the complete homicide cases, determine the use of the evidence, work with witnesses, identify suspects, or have a formal role of the felony approval process. This arm’s length relationship makes it problematic for robustly identifying causality relationships between what the ATCs do and outcomes such as arrests and clearances. However, it is possible to look at the situation pre- and post- introduction and analyze any changes in the status quo. The exploratory analysis suggests that the ATCs might have had a significant role in improving the homicide investigation outcomes in terms of solving crimes faster and more consistently. In addition, the study suggests that the ATC introduction might have allowed the detectives to be more productive and to be able to handle more cases. However, it is difficult to state categorically that the ATCs resulted in a systemic improvement on whether or not a case was actually solved or not. There are many individual examples of where the ATC activity aided in solving a case (Cook et al., 2023), but the larger view is not known.

The next three sections provide a focused literature review of digital evidence in homicide investigations and background information about the ATCs and are followed by sections on methodology and analysis. We conclude with a discussion about the ATC introduction in Chicago in addition to observations about doing such an analysis.

Literature review

The importance and potential contributions related to technology and digital evidence have been highlighted by Goodison et al. (2015); Strom (2017). The increased focus on digital evidence is not ill-founded and there are several factors influencing this focus. For example, technology appears to have an escalating presence in criminal activities, helping to inform criminals about intended crimes and targeting victims via social media stalking as noted by Recupero (2021) and Todd et al. (2021). The digital footprints and trails are important to understand, acquire, and interpret and as observed by Todd et al. (2021), can be overlooked in investigations. In addition, knowing when and where relevant events took place have long been noted as being important for solving homicides (Brookman and Jones, 2022; Ford 1953; Keppel and Weis, 1994), as well as how the probability of clearing a homicide decreases with the duration of the investigation (Regoeczi et al., 2008; Wellford and Cronin, 1999). For example, Wellford and Cronin (1999) note that “Of the closed cases in our study, 93.2% were solved within 1 year and 50% were solved within a week. Therefore, if a case is not solved within a year, the chances of it ever being solved are low.” p. 12.

Timely processing of digital evidence can provide the traditional insights into the when and where aspects of a case in addition to additional evidence that the geospatial data is associated with – for example, videos of what suspects are wearing/doing at the time, cell phone activity, movement tracking, social media content, and internet activity. All of this is important for interpreting the meaning and significance of the evidence and creating credible homicide narratives needed for prosecution (Brookman et al., 2022).

Brookman and Jones (2022) studied the use of CCTV footage in solving homicides and noted that CCTV was used the most to identify and charge suspects compared to other forensic methods. They also documented the challenges and issues faced by detectives in analyzing the evidence, such as the training, lack of skills and expertise, resource conflicts, inadequate processes, and issues with time conflicts resulting in interrupted and duplication of work. Allsop and Pike (2019), van Brunschot et al. (2023) echo these points regarding the challenges many detectives face, including the general lack of resources; having to deal with the increased volume of information that results from the acquisition and processing of digital evidence; and the requirements to stay apprised of changes in forensic science and technology.

Organizational transformations can be relatively large, with major changes in the organization to consider, or they can be smaller, similar to the introduction of the ATCs. The transformations can occur where specialized tasks are isolated (e.g., Smith, 1776), or dedicated to another unit within the organization (e.g., Robb, 1910; Skinner, 1974). Task specialization is associated with improved efficiency and effectiveness, and these are further improved with dedicated, specialized units. With transformations, there are individual factors to possibly consider: learning curves (Anzanello and Fogliatto, 2011) as personnel join the specialized unit and learn new tasks and processes, and transition periods called honeymoon and hangover (Boswell et al., 2009; Franklin et al., 2023). There are also organizational learning curves (Epple et al., 1991).

The individual learning curve period has increasing productivity (efficiency and effectiveness), but also has mistakes to be dealt with and the impact felt by those training and helping the personnel learn. Productivity during this period will be lower than when everyone is trained and working as per the task description. After the initial training, there is a possible honeymoon period and this often sees escalated, heightened productivity associated with high motivation and enthusiasm, which is followed by a rebound period called the hangover where productivity drops, followed by a slight kick in improved productivity once the hangover is dealt with, but not to the levels seen during the honeymoon. These effects can occur at the individual level, not always, but often enough (e.g., Epple et al., 1991) that they need to be considered when doing field work or interpreting data. At the individual level with task specialization, there are also possible effects associated with less multi-tasking and fewer interruptions which can impact both parties – the new area of specialization and the individuals no longer doing the moved tasks allows them to focus on the remaining tasks. The reduction in multi-tasking and time conflicts usually leads to greater productivity on complex tasks (Speier et al., 1999).

There are also organizational effects at the unit level during a transformation and these effects have been called the j-curve or j-effect, named because the shape has a dip and then a rise (Katz and Kahn, 1966). When units are created or roles significantly changed, units need to adjust and learn how to work together and this temporarily creates a dip in productivity as there are meetings, new processes to iron out, a protocol to establish, etc. After the units figure out how to work together and have reached a steady state, the expected or anticipated gains should arise.

If it is necessary to sample and study a situation during the transformation stages and before a steady state is achieved, caution should be taken with conclusions and extrapolations about the future. If possible, it is best to sample and study situations once they reach the stable state, not during the transformation stages.

ATC operations and task complexity profile

As noted, the Crime Lab. final report (Appendix A) provides a comprehensive description of the ATC development, deployment, and what the ATCs do (Cook et al., 2023).

The ATCs are in a support role and are largely directed by the detectives as to what digital evidence to acquire, process, and prepare. The ATCs acquire and process evidence; they do not actually investigate the case and make any summative decisions about suspects, how the evidence is to be used, and how the cases proceed through to possible prosecution. It is hard to directly link downstream outcomes such as arrests with what the ATCs do in a supporting role. This forces the process evaluation to look at the direct impacts which can result from the direct activities of the ATCs. Each ATC has a different workload and mix with variations during the year. This makes comparisons between ATCs and month-to-month situations difficult. There is also clumping within each ATC as each district is different with its own unique profile and it is possible to have one district in an ATC having a different clearance profile compared to the others. Depending on the questions being asked, a district level analysis is possibly needed to better understand the efficiency and effectiveness factors associated with each ATC.

The ATC personnel are trained in the acquisition and processing of digital evidence. This concentrated focus allows them to develop enhanced expertise without distractions related to other types of evidence collection and processing. The expertise extends to what can be collected and processed on scene, and how this be used by the detectives. Compared to the previous pre-ATC situation, the ATCs might collect the same, or different evidence, and if the same evidence is collected, it might be earlier, at the same time, or later than before. The quality of the gathering and processing may also be different. The evidence gathered may or may not be instrumental in discovering other evidence, a subsequent arrest, prosecution and conviction. This suggests that it is very difficult to draw strong conclusions about the actual evidence found and the subsequent processes. In any given month, it is also possible that the homicide profile will be different between districts and areas, and this complicates any time-based comparisons between districts, areas, or ATCs.

Each ATC specialist will have individual differences in skills, expertise, experiences, and approaches to the task which will impact what is done and how it is done. The ATC teams will also be different from each other in efficiency and effectiveness, aggregating to the ATC level where the ATCs are different from each other in various ways. Not all teams will operate at the same level of efficiency and effectiveness, nor will they have the same workload profile.

The task structure also has variability beyond the control of the ATC personnel. For example, the time it takes to unlock a phone is technology based and is random; the ATCs have no influence on how long this will take. In other instances, the homeowner or business owner might not be home when the address is canvased for potential video. There are many aspects of the ATC processes which have inherent variability beyond the ATC’s control, and this makes any process evaluation difficult when considering causality relationships which impact efficiency and effectiveness.

Teams and individuals will be different, even within an ATC unit. Differences will also exist in how the individuals and teams self-manage and take initiative, recognizing and following opportunities as they present themselves on-scene. For example, on-scene, real-time processing of videos by ATC personnel can help guide detectives to where fingerprints might be found. Or, based on what the videos divulge, the evidence will illuminate the path taken by suspects down streets, around corners, leading to other locations where evidence can be found. Or the ATC personnel might decide to collect extra video in the area that could be useful in the future, in anticipation of what the detectives might ask for, to save a trip back to the neighborhood. These types of ‘extra’ activities and contributions go beyond the normal tasks related to canvasing and pulling video from targeted locations as requested by the detectives. There is the formal task and then what the ATC personnel do. This individuality creates variability in the process, impacting the type and ‘quality’ of the digital evidence. There is not a standard, consistent task with a set script, processing time, or tasks. There is great uncertainty in each task and this impacts learning curves, task efficiency, and task output. For the most part, the specific tasks performed by the ATC units are not tightly scripted by the requesting detectives or procedures – they are opportunistic by nature.

Potential impact of ATCs

Year-over-year comparisons can be problematic when analyzing the Chicago Police Department data. For example, there was the turmoil in 2014-2019 leading up to the Consent Decree (Consent Decree, 2019), followed by mandated changes, additional turmoil, and then capped off with the COVID pandemic impacts during 2020-2021. However, it is possible to speculate about the types of impact that the introduction of the ATCs could have and then investigate their possible presence via an exploratory case study, while keeping the larger context in mind. The environmental scan (Appendix B) suggests that at the investigation level, there were potentially some impacts on actual processes and methods used by the investigators, and that in general, cases would be expected to be possibly harder to solve and that the processes would be slower over the time period analyzed. Though this perception may have been biased and inaccurate, it was held by all of the CPD individuals interacted with as part of the scan.

The analytical framework used to view the potential impact of the ATCs on the investigative processes has its foundations in Industrial Engineering work flow analysis. Concepts related to (i) value stream analysis (Rother and Shook, 2003), (ii) cellular manufacturing (Hyer and Wemmerlöv, 2002), (iii) focused factories (Skinner, 1974), (iv) parallel tasks and work sequencing (Kelley and Walker, 1959), (v) work specialization (Smith, 1776), (vi) bottleneck analysis (Goldratt and Cox, 1984), and (vii) critical path analysis (Goldratt, 1990) provide guidance about how to explore the task splitting and workflow changes associated with the introduction of the ATCs, and the impact on quality and flowtime.

Work sequencing, bottleneck, and critical path analysis provide insights into the impact on the key resource (e.g., detectives) when work is moved upstream or performed in parallel. In addition, the ATCs can be viewed as either a focused factory element or as a form of cellular work design leading up to the main investigative tasks performed by the detectives, both of which have been known to improve quality and flowtime performance. Specifically, using cellular work units is an Industrial Engineering method directly designed to improve flow time and quality, especially for high-mix, low-volume environments. Finally, value stream analysis provides a lens for identifying the way value added work may or may not shift with the impact of ATC introduction.

In summary, these Industrial Engineering topics informed the research activity as to what might be reasonable and plausible to observe during the study; what can be expected when work is specialized and performed in parallel. To understand the complexity of the task and understand what was specialized and split, two documents provided key information, describing the scope and nature of the ATC introduction in terms of tasks, task specialization, and support of the homicide teams (Chicago Police Department, 2021; City of Chicago, 2020).

Based on the literature relating to Industrial Engineering and work flow and design, six trends and patterns can be potentially expected to be observed in the exploratory study. Note, the impacts normally associated with the Industrial Engineering principles are not predictive or normative and are dependent on the quality of implementation. The possible implications of dedicated work specialists found in the ATC are:

(1) Digital evidence would be possibly available ‘earlier’ in a case to be used by the investigators

(2) More digital evidence would be likely acquired and made available

(3) An improvement in consistency and reduction of variance with respect to the acquisition and processing of digital evidence would potentially occur

Assuming that the digital evidence plays an integral and important part of the investigations, several indirect impacts may be expected to be observed:

(4) The timing and variance of processes leading up to arrests could be markedly different and more consistent

(5) The number of cases processed by the homicide detectives to the point of an arrest being made could increase when compared to equivalent time periods (e.g., first 90 days after a homicide occurs)

(6) More cases could possibly be cleared with arrests than with “exceptional means”

These six potential impacts are suggested by the literature, related to the impacts on flowtime and quality and will be investigated, using the CPD supplied data and exploratory case study methodology. The environmental scan and the organizational transformation dynamics will be applied in the interpretation of results.

Methodology

An exploratory case study approach (Yin, 2018) was used. As such, there are no propositions or hypotheses about causality. The goal was to better understand the nature of digital evidence processing before and after the introduction of the ATCs. Although direct causality relationships were not expected to be found, the exploratory approach expected to identify patterns of evidence processing and investigator activity. During 2021, interactions and discussions were held with CPD BOD personnel to understand what the ATCs were and how they operated. An environmental scan to understand the greater context within which the ATCs were operating was performed as part of the background analysis and is summarized in Appendix B. The scan focused on activities that would impact flowtime and quality. Though the scan has limitations and it is possible that factors could have been missed, it suggests that during the 2019-2020 period there were likely no work design factors which would have made investigations significantly more efficient or effective with the exception of the ATC introduction. The 2025 CPD report (Chicago Police Department 2025) on changes in the homicide investigative process 2019-2024 was also reviewed; no additional factors were highlighted in the CPD report that would potentially impact the individual work flows. The scan further suggests that a number of factors existed in this time frame which would have potentially slowed down the investigative processes and made cases harder to solve.

In April 2022, the CPD provided a variety of datasets to support the analysis. The time window for data extraction was January 2014, through March 31, 2022. The data for 2022 allowed for a 90 day window of analysis to be performed for homicides occurring in late December 2021. ATC ticket data was provided for April 25, 2019 to March 31, 2022. In addition to the ATC ticket data, the CPD provided data on violent crimes, shootings, homicides, electronic evidence records (ETrack), felony approval transactions, number of detectives per area per year, and status updates per case. In addition, the public portal for CPD data was used to capture a snapshot of homicide data in May 2022, as well as in May 2024 (https://data.cityofchicago.org/Public-Safety/Homicides/iyvd-p5ga/data). The May 2024 public portal data provided an additional 2 years of longitudinal information for the cases being studied. For comparative purposes, the CPD provided data was filtered to match the cases visible via the public portal. The 2022 versus 2024 portal data shows the effect of time as 1806 arrests are noted for the 2014–2021 period in the 2024 data versus 1670 in the 2022 snapshot.

Several data analysis tools were created to merge the various datasets, create timelines for each case, and timelines for evidence aligning with the ATC requests. The tools provided filtering capability to extract feature statistics (e.g., time of first digital, non-digital evidence) at the District, Area, and CPD levels, yearly and monthly.

Once the final Crime Lab. report was obtained, additional analyses were performed with the datasets to ensure that a number of the lab.’s analyses on homicide rates, arrests and evidence were possible to replicate. There were some differences related to data sources and focus, but otherwise, patterns, trends, and observations were matched. For example, felony approval records from CPD versus records from the state attorney’s office (Crime Lab.), the final ticket system versus the initial ticket system (Crime Lab.), focus on digital versus video only (Crime Lab.).

The focus was on homicides and digital evidence; the evidence records were analyzed with this perspective. From the 555,257 evidence records, 162,149 were identified with homicides. In the dataset, there are descriptive tags which we concluded could be used to indicate some kind of digital evidence processing or handling - “Cell Phone / Pager”, “Computer Hardware / Software”, and “Recordings – Audio / Visual”. These tags accounted for 20,934 records, of which 16,999 were in the recordings category. Three other tags (“Photographs”, “Other”, and “Worthless”) were processed using a text mining process keying on terms related to digital evidence and this was followed up with manual inspection of each record to eliminate false positives. Using the text mining, 9146 additional records were identified, of which 2744 were eliminated as false positives, leaving 6405 additional records. The mining provided 31% more records, compared to the three main tags. It is possible that false negatives exist in the coding, but visual skimming indicated that there would be relatively few and the analysis proceeded with the 27,339 evidence entry records.

The ‘first’ decisions or events were identified within the feature sets. Cases are cumulative and can have multiple victims, multiple suspects, multiple evidence entries, and statuses can change. The analysis was oriented to how quick the detectives obtained processed evidence to use and the first entry in the ETrack system was used, versus the recovery date. Similarly, a decision had to be made about the anchor date for the homicides. There is the injury date and date of death. While often the same, there can be a difference and when analyzing at the hour and day level, the timing anchor is important. We chose the date of death to anchor the statistical analysis – data collected and processed before this date would have been handled under the category of non-fatal. Evidence processed prior to the death date was assumed to be immediately available to the homicide detectives.

Homicide cases can have extended investigations lasting many years. For example, as of April 2022, the ATCs had processed requests for 50 cases from 2017, 93 from 2018, to 28 cases prior to 2017, going back in one case greater than a decade to 1993. Extended cases can create spikes in the data and distort the statistics. A general filter/cap, of 2 years was used to limit extreme cases from biasing the data when the data was not otherwise filtered by time. The mean for the time between the death date and arrest date in 2019 was 81 days and a general 90 day threshold was set, along with fences at 30 and 60 days for analysis purposes. The first month was also decomposed to days for the first week and then weekly for the remainder of the month.

We included a capacity analysis as part of the study – how productive the detectives were. Were the detectives handling more cases? Were the cases being dealt with more efficiently? Were there more homicides cleared per detective? It was not possible to get a precise number of detectives working homicides per area per month. However, the number of detectives associated with an area during a year was obtained. The number does not consider when detectives joined the area or left. Areas do not have the same number of detectives, and it is assumed that the assignment of detectives to homicide cases will vary with the homicide load. In the analysis, it was assumed that for the areas with a high homicide count, roughly the same distribution of detectives to homicide investigations was used by the areas. This assumption was based on discussions with CPD personnel regarding the number of homicide teams, and the personnel numbers per team in the high homicide activity areas; 20% of detectives dedicated to homicides appeared to be a reasonable estimate for trend analysis.

Analysis

The demand and supply profile is first reflected upon, followed by the six key ‘expected’ observations.

Homicides are an independent ‘demand’ factor in the analysis, they happen when they happen for a number of reasons and must be dealt with via the BOD capacity. Figure 1 shows the number of homicides per year for the study period.

Figure 1.

# of homicides per year.

To illustrate the variance during the year and variance between years, the charts below show the number of homicides by area for each month in 2020 and 2021 (Figures 2 and 3). The high variance suggests that a yearly cumulative view should be taken without ‘reading’ too much into monthly patterns or details at the month-to-month level of analysis (e.g., June 2020 vs June 2021 for Area 4).

Figure 2.

# of homicides by month by area for 2020.

Figure 3.

# of homicides by month by area for 2021.

Digital evidence would be available ‘earlier’ in a case to be used by the investigators

Figure 4 shows the first digital evidence activity for a case. Prior to 2019, digital evidence processing in the first 24-48 hours was reasonably stable with a ‘similar’ gap between the number of cases and the cases with processed digital evidence.

Figure 4.

# of cases with digital evidence activity within 24 and 48 hours.

Starting in 2019, there was a change in the pattern of evidence activity, narrowing the gap, and this is reasonable to expect with the dedicated resources for digital evidence being available. While it is not possible to know the precise impact of having the digital evidence available in the early stages of an investigation, it is plausible that earlier access to the evidence can significantly impact the investigations with respect to identifying suspects, piecing together the crime profile.

However, the first digital evidence activity is not the whole story. Another perspective is to look at the percentage of digital evidence activity for the initial 90 day window - what was done in the first 24 and 48 hours. Figure 5 shows a dramatic difference pre- and post- 2019 – what was available for the detectives to use.

Figure 5.

% of digital evidence activity within the first 24, 48 hours compared to 90 days.

More digital evidence would be acquired and made available

There was an increasing trend prior to 2019, with a dip in 2018 (66%) – Figure 6. There is then a marked upwards change in 2019 (87%) which could be related to having the additional ATC personnel available to acquire and process digital evidence. It can also be due to a greater number of available evidence sources (e.g., more doorbell cameras).

Figure 6.

% of cases with digital evidence per year.

The average number of digital evidence entries per case started to increase in 2016 (Figure 7) and peaked in 2020 (possible honeymoon effect), and then decreased in 2021. This change could be a hangover effect or be caused by other plausible causes such as the ATC personnel becoming more skilled and selective in what evidence to collect.

Figure 7.

Avg. # of digital evidence entries per case.

An improvement in consistency and reduction of variance with respect to the acquisition and processing of digital evidence would occur

The mining of the ‘Other’, ‘Worthless’, and ‘Photograph’ tags discovered significant digital evidence entries, combined what we called the ‘Other’ category compared to the specific tags: “Cell Phone / Pager”, “Computer Hardware / Software”, and “Recordings – Audio / Visual”. With specialization and training, it is expected that record keeping and processes should improve. This is also an indication of the learning curve and impact of standardized processes, shared terminology, and methods.

Prior to 2019, the number of ‘hidden’ digital evidence references in the ‘Other’ categories hovered around 27% (Figure 8). In 2020 and 2021 this dropped to 20% and 16% respectively.

Figure 8.

% of digital evidence entries found in other tags.

The timing and variance of processes leading up to arrests should be markedly different and more consistent

Visually, the mean and standard deviation in ‘making arrests’ started going down around 2017 with a relatively consistent decrease in both over time (Figure 9). It is not clear if the ATCs contributed to this trend since it started pre-ATC, but with the increased digital evidence in the early stages of the investigations, it is plausible to assume that the ATCs had some contribution to the ongoing improvement.

Figure 9.

# of days to make an arrest – mean, std. dev.

In many of the years, approximately 50% of the cases which will ultimately result in an arrest will be solved in approximately a week (the median) (Figure 10). This is similar to the 1 week threshold found by Wellford and Cronin (1999). 2020 saw a possible honeymoon effect followed by a higher median in 2021 (a low of 5, then a bounce to 11). If 2021 is considered to be more typical, and 2018 is excluded, there is no clear “visual” evidence that the introduction of the ATCs impacted the median time to solve a homicide. However, if we use the environmental scan, all things being equal, the median should have risen due to cases potentially taking longer to solve and hence the ATC introduction might have had a mediating effect – keeping the median time to a reasonable level. A deeper study would be required to distill any direct relationships and to understand why the median time did not increase as the environmental scan suggested would happen.

Figure 10.

Median # of days – time to arrest.

The number of cases processed by the homicide detectives to the point of an arrest being made should increase when compared to equivalent time periods (e.g., first 90 days after a homicide occurs)

To probe this possible effect, we estimated the number of detectives working homicide each year and conducted several analyzes using this estimation (Table 1). This estimatation allowed a homicide/detective ratio to be calculated. The ratio was at a low in 2019 possibly related to the new detectives added to the roster, and then increased in 2020 and then in 2021 again. Assuming that a resource is working close to capacity, further load is often linked to congestion, resource conflict and a reduction in efficiency and effectiveness.

Table 1.

Homicide cases - # of detectives.

Year	Cases	Est. detectives working homicide	Homicide/detective ratio	% of cases with arrests - 90 days	# solved per detective ratio – 90 days
2014	417	134	3.1	31.9	0.99
2015	464	145	3.2	23.7	0.76
2016	742	142	5.2	18.6	0.97
2017	613	174	3.5	16.7	0.58
2018	554	155	3.6	19.7	0.71
2019	480	177	2.7	21.9	0.59
2020	743	186	4.0	23.4	0.94
2021	763	167	4.6	20.2	0.92

The relative number of arrests within 90 days had decreased since 2014 to a low in 2017. There was then an increase in ‘more’ cases being solved within the 90 day window. This might be related to detective efficiency or not. It is possible that the increase was not related to efficiency or effectiveness. A ratio was used to investigate the ‘number of detectives’ possible effect: number solved within the 90 day window per detective. This has a slightly different pattern, with 2017–2019 being relatively low. In 2020 and 2021, there was marked increase in what could be called detective productivity, going from 0.59 to 0.94 and 0.92.

More cases should be cleared with arrests than with “exceptional means”

There are multiple states a homicide can be assigned as it goes through the investigation process. The states associated with arrests are of interest as are the states where the case is considered clear by exceptional means. Higher quality clearances should see an increase in arrests and fewer cleared by exceptional means. Figure 11 shows that 2017–2019 was reasonably stable, but 2020 and 2021 saw an increase in the clearances associated with arrests.

Figure 11.

% of clearances with an arrest.

Discussion

Six potential impacts related to work specialization and parallel work, three direct, three indirect, were identified as part of the exploratory case study. While causality links (how and the why) were not part of the study, based on the trends, something happened in 2019, an inflection point, where the status quo was changed. Digital evidence acquisition and processing was different.

We suggest that the three potential direct impacts can be seen in the changed patterns. More digital evidence appears to have become available earlier in the investigation processes. The record keeping associated with inventorying the evidence saw a marked improvement in consistency and standardization. Based on a process view of the tasks, these are plausible and are likely to occur if the introduction and implementation of the task splitting, parallel work does not have major flaws.

The three indirect impacts which could have been expected based on the Industrial Engineering principles can also be possibly seen in the data patterns. While the cause-and-effect relationships cannot be determined with certainty, the time-to-arrest appears to have continued to improve and more consistent from 2019 onward. Using an estimate for the number of detectives working homicide cases, it would appear that detectives were able to handle more cases, which is logical if the ATCs are doing tasks the detectives used to do, and cases are being processed quicker as a result of the earlier evidence availability. It is also possible that the ATCs impacted the quality of the investigations, not just the speed, since more cases were cleared with arrests than with “exceptional means”.

Taken together, considering that the environmental scan suggests that cases would be harder to solve and take longer, we suggest that the ATC introduction had a possible observable impact on the investigative processes. Something happened in 2019 and the next 2 years. The only thing of note that we were able to identify through this study that would have improved the processing was the ATC introduction.

It is also important to note the possible implications of the individual and organizational learning curves, the honeymoon and hangover periods. 2021 is likely close to a steady state situation and caution should be used when considering any trends or observations about 2019 and 2020. However, there are many moving parts to the situation, and it is likely that 2021 was not a ‘normal’ year when all of the factors are considered. There is possibly no normal.

Limitations and future research

Any evaluation of a situation similar to the ATC introduction is complex. It was not a controlled experiment. There are added challenges because of the variation per year, per month, per area, per district. There are no comparators, no baseline. It is not clear how much of the impact is due to the specialization of the tasks and/or the dedicated nature of the specialists. There would have been some gain from the basic specialization, and additional gain from the decentralization of resources, but it is not possible to isolate these two effects.

The exploratory case study was limited to what could be found in the electronic records. There was no access to the actual case notes. This limits the analysis to proxies or secondary indicators. For example, we know that a disc with digital evidence was inventoried, but we do not know what exactly was on the disc, number of clips, type of cell phone data, how the evidence was later used. We decided the best approach was to look at the situation as ‘digital evidence activity’ and not at a further level of evidence processing. We suggest that future research be longitudinal with researcher participation, awareness at the case level, where enriched data can be gathered and analysed.

We were unable to create a clean link between the CPD data and what is available via the state attorney’s public portal – linking the felony data to the homicide case. We decided to limit this aspect of the analysis and to direct interested parties to the Crime Lab. final report which focuses on the process and used data from the state attorney’s office. Future research should address this gap – linking evidence to actual use, value in a prosecution.

In a homicide case, there can be multiple victims, multiple suspects, multiple arrests, and the status of the case can proceed or revert as the case unfolds; decisions and status changes. For example, cases can be closed and then re-opened. Because of this, there were points in the analysis where a decision had to be made about what data to use and how the robustness of the data will impact the analysis. There is also variance in the data related to the long-time horizons for some cases, creating a partial snapshot in April 2022, when the datasets were pulled for 2014 onwards. It is suggested that a future analysis include data through 2025 to provide a 3 year window after the 2022 stabilization point.

We were able to obtain a rough number of detectives per area per year and create a rough rule-of-thumb for detectives assigned to homicide based on current practice in several of the areas, but the capacity analysis is limited in its robustness. A more granular approach is needed with personnel tracking per case if productivity is to be investigated a more robust fashion.

The Crime Lab. final report notes similar issues with attempting a quantitative analysis and, like our study, avoided a sophisticated statistical analysis or using strong methods in a weak situation, instead relying on general observations about the trends.

As noted above, future research could possibly include analyzing an updated dataset from the CPD to assess how the digital evidence processing performed since 2021. It would also be of value to do similar studies where specialization has not been undertaken, or where the specialized resources are centralized and not distributed. If other contexts were studied, a formal qualitative approach similar to the Crime Lab. study would be required, which provided a foundation and a comparator for our analysis.

Conclusion

We do not know if homicides are becoming harder to solve in general. In Chicago, there was a perception that they were, and this thinking is plausible considering the situation at the time of the study. These challenges appeared to put an emphasis on the digital evidence that is collected, as noted in the Crime Lab. study. While the CPD ATC introduction appears to have made a positive contribution to the solving of homicides in Chicago, we cannot claim that such an introduction would have similar results in other constituencies, as each agency faces different circumstances.

Based on our analysis, the specialization and dedicated resources resulting from the ATC introduction produced some of the outcomes initially set forth by the administration. Crimes appear to have been solved faster and more cases were solved with arrests in 2020-2021. Furthermore, it appears that process changes impacted the detectives; they were able to do more with less in more challenging situations. In 2021, there was an increase in homicides and a reduction in detective capacity, but the performance level remained high. Given Chicago’s situation, the investment and support of the ATCs appears to have been sound and appropriate.

The ATC introduction was a substantial initiative and innovation. While the cost per ATC is not substantial (Cook et al., 2023), there are personnel implications and ongoing costs associated with the operation. In addition to the detective complement, additional resources and specialized ATC personnel are required to maintain and run the ATC as designed, which must be factored into consideration.

For others considering similar studies where an organizational transformation occurs, where a specialized unit is created with the possible altering of workflow, it is important to consider the implications of the change. There will be some kind of learning curve and j-curve effects unless mitigating measures are taken (e.g., additional personnel for the period of transition), as the process evolves into a steady state. We hope that we have provided some insights to other researchers into these issues.

We would like to thank the Chicago Police Department, Bureau of Detectives for supporting the study, facilitating personnel access, and the Data Fulfillment and Analysis unit for providing the datasets. Chief Deenihan, Lt. West, Lt. Kinney, Sgt. Nichols, Detective Meehan, and others were instrumental in the study. We would also like to thank the paper reviewers for their constructive suggestions and guidance. The granular data supporting the aggregate analysis was provided by and belongs to the Chicago Police Department. Any further use of this data must be approved by the Chicago Police Department. Points of view or opinions contained within this document are those of the authors are do not necessarily represent the official position or policies of the Chicago Police Department.