How the IRS Decides Which Collection Cases Matter Most

The IRS has more collection cases than people to work them. So it has to choose. The choice is not a mystery. It is made by a combination of written policy, routing rules, predictive models and annual management priorities.

Understanding how those pieces fit, and where TIGTA found they did not, explains why some balances get immediate attention and others sit for years.

The policy foundation

Two policy statements in IRM 1.2.1 (rev. 2026-06-23) frame the question. Policy Statement 1-236, Fairness and Integrity in Enforcement Selection, says selection processes use scoring mechanisms, data driven algorithms, third party information and other sources to identify the highest potential noncompliance, under checks and balances, so that no one individual can control enforcement selection decisions.

Policy Statement 5-134 says collection operations will be concentrated on collecting the taxes that will produce the greatest revenue yield, based on analysis of revenue potential in each category of tax and local conditions. It also says that in disaster areas the IRS will concentrate on current taxes, while continuing enforced collection of delinquent accounts where taxpaying potential is substantially unimpaired.

Put simply: rank by risk and yield, and let systems, not individuals, do the ranking.

The routing models

The first ranking happens when a case leaves the notice stream. IRM 5.1.20.2.1 (rev. 2024-10-07) says the Inventory Delivery System uses predictive modeling techniques, case grade and routing indicators to prioritize cases for routing. Model scores indicate the likelihood that a predicted case outcome will occur, and IDS passes those scores to each function's case management system for final prioritization.

IRM 5.1.20.1.4 says headquarters analysts review the business rules annually for continued effectiveness and evaluate the predictive models annually. IRM 5.1.20.2.2 says IDS also shelves cases predicted to be inactive or unproductive because of resource limitations.

Inside ACS, IRM 5.19.5.1.8 (rev. 2025-03-03) describes priority programs including federal employee and retiree delinquencies, high income non-filers, international cases and payroll trust fund taxes, with risk categories A to Z and priority codes 0 to 99.

In the field, IRM 1.4.50.8.4 (rev. 2025-04-01) says the priority levels shown to group managers reflect elements such as balance due, return types, tax period, model score and selection code, and IRM 1.4.50.10 says the scoring was developed with an emphasis on recent business trust fund taxpayers.

The annual priorities

On top of the models sit management priorities. TIGTA Report No. 2017-30-069, issued September 25, 2017, explains that IRS Collection management sets goals for closing cases in certain priority areas each year and communicates them in the Collection function's annual program letters.

TIGTA compared two years. For fiscal year 2015, the priority areas it listed included delinquent employer payroll taxes, federal employees and retirees, global high wealth, high-income nonfilers, large business taxpayers with delinquent accounts, large dollar accounts, return preparer penalties and withholding compliance. For fiscal year 2016, the list shifted mostly to measures of how assigned cases were worked, such as federal tax deposit alert timely contact, lien timeliness, offers closed within nine months, trust fund penalty timeliness and timely proof of claim filing.

TIGTA reported that IRS management attributed the change to new SB/SE concept of operations themes focused on early intervention, prevention and timeliness.

What TIGTA found

The report's title states the conclusion: prioritization of collection cases is inconsistent. TIGTA found that from year to year the priorities changed for reasons not always articulated or based on objective data, that the priority areas were not all explicitly tied to the prioritization and risk rules that drive routing and assignment, and that cases in priority areas accounted for a small percentage of assigned cases. It also found limited data available to assess whether the goals for priority cases were met.

TIGTA reported that IRS management said measuring performance in priority area workload focused on how cases are closed, not how many are assigned or whether the next case worked is a high-priority case.

TIGTA recommended that the IRS develop case priority areas with data-driven approaches, make sure downstream manual selection by group managers favors higher priority inventory and set goals for all high-priority areas. The report records that IRS management agreed to take corrective action on four recommendations and disagreed with one, mainly due to limited resources.

Shelved and queued cases got less automated attention

The same report connected prioritization to automation. TIGTA found that the IRS did not use its automated collection tools on all modules awaiting assignment, and that unassigned modules in the queue or shelved were not subject to the same level of systemic processing as modules in ACS. It pointed to systemic actions such as automated levy source searches as tools that can improve compliance.

TIGTA recommended that the IRS consider expanding systemic processes to cases in the queue and in shelved inventory, and consider updating levy source details in real time and using them in routing and prioritization. Those recommendations show where the low-priority end of the inventory sat in 2017: ranked, but largely untouched.

What changed in the manual since

The current IRM puts heavy emphasis on the downstream piece TIGTA flagged. IRM 1.4.50.10 tells group managers that assignment must align with Policy Statement 1-236, that professional judgment plays a limited role in case selection, that it is generally inappropriate to assign medium or low priority accounts when high priority accounts are available and that a lower priority case should not be chosen over a higher priority case of the same type just because it has a better predictive value.

It also tells field compliance managers to run assignment and priority level reports quarterly for each group to find patterns that deviate from guidance on assigning priority inventory. I cover those rules in How Collection Cases Are Assigned to Revenue Officers.

I cannot tell you from public sources how well those changes resolved TIGTA's concerns. What I can tell you is that the manual now says, in plain terms, that the highest priority available case should be assigned first.

What makes a case rank high

Pull the pieces together and some patterns are clear from the IRM. Business trust fund taxes rank high, which is why payroll tax cases reach Revenue Officers quickly. Large balances rank high, and IRM 5.1.20.3.1 sends certain large dollar cases straight to the field. High income non-filers and federal employee delinquencies have dedicated programs. Recent liabilities and accounts with a predicted productive outcome rank above old, low-yield ones.

Time matters in both directions. IRM 1.4.50.10 says modules with less than six months left on the collection statute normally will not be assigned to a Revenue Officer, though high priority cases, high income non-filers and potentially egregious repeat trust fund cases should normally be assigned anyway. And the newest liabilities get attention because the IRS's policy, as Policy Statement 5-134 puts it, gives priority to current tax collections.

And some patterns rank low. Older, smaller individual balances with no known levy source are the ones most likely to sit in the queue or be shelved, as I explain in The IRS Collection Queue.

What this means for you

Do not mistake low priority for no priority. Every year a case is shelved or queued, penalties and interest continue, and a new balance or new information can push it back to the top. And do not take high priority personally. A payroll tax case gets fast attention because the system is built to give it fast attention.

The useful question is not whether the IRS will come. It is when, through which office, and what you will have done before it does.