An alert backlog clears at the rate capacity exceeds new alerts, which is why a team that looks almost big enough can take three years to catch up.
An AML alert backlog clears at the rate the team's capacity exceeds the alerts still arriving, so the time to clear it is the backlog divided by capacity minus inflow. On an illustrative team clearing 6,250 alerts a month against 6,000 new ones, a 9,000-alert backlog takes 36.0 months, not the 1.44 months that backlog divided by capacity suggests. Clearing it in three months takes 19 contractors and costs 541,500 euros.
Worked in full in The EU AML Handbook by Julian R. Sterling, with every figure reproduced in a free workbook.See the book on Amazon →
A backlog is a queue, and a queue drains only by the difference between what goes out and what comes in. A team that is almost big enough for its daily flow has almost nothing left over for the backlog. That is why backlogs persist for years in firms that believe they are a quarter away from clearing them, and why a supervisor asking for a remediation date wants to see the inflow next to the capacity.
| Input | Value |
|---|---|
| Alerts in backlog | 9,000 |
| New alerts a month | 6,000 |
| Analysts | 25 |
| Productive hours per analyst a year | 1,600 |
| Level one review, minutes per alert | 20 |
| Alerts escalated to investigation | 8.0% |
| Investigation, hours per case | 2.5 |
| Contractor productivity against an analyst | 70% |
| Contractor output in the first month | 50% |
| Contractor cost a month, euros | 9,500 |
All figures are illustrative. The level one and investigation times follow the same structure as the false positive rate and staffing calculation, with report drafting left out on the assumption that a separate team handles it; replace them with your own measured times before quoting a date.
1. Hours per alert = 20 / 60 + 8.0% × 2.5 = 0.5333.
2. Hours per analyst-month = 1,600 / 12 = 133.3, so one analyst clears 133.3 / 0.5333 = 250 alerts a month.
3. Team capacity = 25 × 250 = 6,250 alerts a month.
4. Net burn = capacity − inflow = 6,250 − 6,000 = 250 a month.
5. Months to clear = 9,000 / 250 = 36.0.
In Excel: =IF(Cap>Inflow,Backlog/(Cap-Inflow),"never"). The IF matters: at 24.0 analysts the team only keeps pace, and below that the backlog grows whatever the plan says.
The same numbers give the cost of the backlog to every new alert. Reviewed first in, first out, an alert raised today waits for the 9,000 ahead of it: 9,000 / 6,250 = 1.44 months, about 43.2 days. That is the figure that matters to the financial intelligence unit, because a suspicion that should have been reported is reported six weeks late.
To clear the backlog in three months the team must remove 3,000 a month on top of the inflow. The permanent team already contributes 250 of that, leaving 2,750. A contractor at 70 per cent of an analyst's output clears 175 alerts a month, so the steady-state answer is 2,750 / 175 = 15.71, rounded up to 16.
Sixteen is not enough, because contractors are not productive on day one. With half output in their first month, only 87.5 alerts each, 16 contractors leave 1,250 alerts at the end of month three and finish in month four. Solving with the ramp: 9,000 = 3 × 250 + n × 175 × 2.5, so n = 18.86, rounded up to 19 contractors, who clear the backlog in month three.
| Plan | Month 1 | Month 2 | Month 3 | Month 4 | Contractor cost |
|---|---|---|---|---|---|
| 16 contractors | 7,350 | 4,300 | 1,250 | 0 | 608,000 |
| 19 contractors | 7,087 | 3,512 | 0 | 0 | 541,500 |
The larger team is cheaper: 57 contractor-months against 64, because the smaller one has to be paid for a fourth month to clear a remainder. With 19 contractors the team clears 9,575 alerts in a full month.
| Case | New alerts a month | Net burn | Months, no contractors | Months, 16 contractors |
|---|---|---|---|---|
| Base | 6,000 | 250 | 36.0 | 4 |
| New scenario adds 10% | 6,600 | −350 | never | 5 |
| Tuning removes 20% | 4,800 | 1,450 | 6.2 | 3 |
The backlog is far more sensitive to the inflow than to the team. A new detection scenario that adds 10 per cent of alerts turns a slow recovery into a backlog that grows by 350 a month. Tuning that removes 20 per cent, if below-the-line testing shows little is lost, clears it in 6.2 months with no contractors at all, as worked in how to set an AML transaction monitoring threshold. Launching scenarios and remediating a backlog in the same quarter is the combination to avoid.
After the surge leaves, the permanent team runs with only 250 alerts a month of headroom, 4.0 per cent of its capacity. Any rise in inflow restarts the backlog, so the plan should fix the steady state, not only the remediation.
The same volume-times-throughput method sizes the KYC back book in the free workbook for this book.
Divide the backlog by the net burn rate: monthly team capacity less monthly new alerts. A team clearing 6,250 alerts a month against 6,000 arriving burns only 250 a month, so a 9,000-alert backlog takes 36.0 months. Dividing by capacity alone gives 1.44 months and is the most common planning error.
Work out the extra monthly output the deadline requires, then divide by a contractor's output after a productivity discount and ramp-up. In the illustration, clearing 9,000 alerts in three months, with contractors at 70 per cent of an analyst's 250 alerts and half output in their first month, needs 19 contractors, costing 541,500 euros.
With first in, first out review, a new alert waits for everything ahead of it: the backlog divided by the team's monthly capacity. A 9,000-alert backlog cleared at 6,250 a month means a new alert waits 1.44 months, about 43.2 days, before anyone looks at it, which is what delays any suspicion report that should follow.
This article is one calculation from The EU AML Handbook. The book takes the same case from first principles to the decision, chapter by chapter, and every figure it prints is a live formula in the free companion workbooks.
Get the book on Amazon →Free companion files
Also on Amazon UK · Amazon Germany · Amazon France · Amazon Canada
Reading guide: regulation, compliance and banking → · All 1340 articles →
If this book helped, or didn’t, a few lines on Amazon are worth more than they look: they are what the next reader goes on. Write a review. The workbook stays free either way.