What a Loss Study Actually Is
A loss study is a structured analysis of historical claims data used primarily in property-casualty insurance to estimate future loss costs. It takes your actual paid and incurred claim amounts, strips out the noise, and produces a clean estimate you can hand to underwriters or rating bureaus. It sounds straightforward. In practice, it is usually a mess of incomplete files, inconsistent coding, and data from five different accounting systems that don't talk to each other. I spent three years building these for regional commercial lines carriers before moving into consulting work. The people who do this well share one trait: they stop arguing with the data and start cleaning it methodically. That is where a Loss Study Guide Checklist comes in useful.
Loss Study Guide Checklist
Below is the practical checklist I actually use. Not the textbook version. The one that survived real engagement audits. You cannot build anything without complete records. This is where most projects fail before they start. Get the following from the insured or from your own databases: Premium records — detailed by line of coverage, policy period, and class code. If you only have aggregate premium, you will need to back into it using exposure reports, which adds error margins you do not want.
Claim files — closed and open claims with payment history. Paid amounts, expense payments, and reserve movements. Open claim reserves must be age-progressed or credibility-weighted depending on how old they are. A reserve that has not moved in eighteen months is not a neutral number. It is a bad number. Payroll or exposure bases — this depends on the line. Workers compensation uses payroll. Auto physical damage uses units or garaging days. Match the exposure base to what the rating manual actually calls for. Mismatching these is the single most common error I see in peer-reviewed loss studies. Policy schedule changes — declarations pages showing limit changes, deductible adjustments, and endorsement modifications across the experience period. A coverage gap of ninety days or more will distort your loss ratio calculation unless you flag it.
Get the Full Details
Step Two: Cleaning and Adjusting
Raw data lies to you. It looks like numbers but it is not truth. You have to adjust it. The standard adjustments are: Frequency adjustments — remove claims below your attachment point. These are noise, not signal. Keep a log of how many were excluded so reviewers can audit you if needed. Severity adjustments — separate large outlier claims that are not representative of normal operations. I once had a manufacturing client with a single forklift incident that inflated their severity by forty-two percent over a three-year period. We excluded it and documented why. The auditor accepted it. He asked for the supporting photos of the damaged rack, which I had on file.
Inflation adjustments — apply loss cost trend factors for the appropriate line and geography. Use your carrier's published trend or the ISO trend curves if you have access. Never guess at this. A wrong trend factor compounds across every year in your experience period. Rate-level change adjustments — if the insurer changed rates during your study period, you need to normalize back to a consistent rate level. Most people skip this. It costs them credibility with the auditor.
Step Three: Computing Loss Ratios and Pure Premiums
Once your data is adjusted, calculate the loss ratio for each year in your experience period. Divide total incurred losses by total earned premium. Then compute the weighted average across the period, giving more weight to recent years depending on your credibility assessment. For workers compensation specifically, you will want to separate loss components into medical, indemnity, and expense. Each moves differently. Medical trends run hotter than indemnity in most markets right now. If you lump them together you get a number that is technically correct and practically useless.
Step Four: Credibility Assessment
This is where beginners lose points. Full credibility does not apply to every account. Use the LimitedFL credibility method or the Buhler method depending on what your local rating bureau requires. The formula is not hard. The judgment call is whether your experience period is long enough and whether the data is stable enough to support the weight you give it. If your credibility is below fifty percent, you blend your experience with manual or industry averages. That is not a failure. That is the system working as designed. I have seen adjusters try to force full credibility on two years of data with six claims. It did not end well during an audit.
Step Five: Output and Documentation
Your final deliverable should include the computed loss ratios, the trend assumptions, the adjustments made with justification, and a summary schedule that maps each claim to its adjustment category. Build an appendix with the raw data and your transformation worksheet so anyone can reproduce your work from scratch. I recommend using a single master spreadsheet with clearly labeled tabs for raw data, adjustments, and final calculations. Keep all formulas visible. Do not hide constants in text boxes. Auditors will ask to see the cell references and they will not be kind if you cannot provide them.
Common Pitfalls
Ignoring retroactive date changes — policies with different retro dates create exposure mismatches that skew your earned premium calculation. Check every policy for retro modifications. Double counting claims — when a claim appears in both paid and incurred columns across different reporting periods, it gets counted twice if you are not careful. Use claim ID tracking throughout the entire process. Misclassifying class codes — a single wrong NAICS or class code assignment can shift your experience by fifteen to twenty percent depending on the line. Verify every code against the actual operations, not just the policy schedule.
Over-adjusting — there is a fine line between legitimate adjustment and data manipulation. If you adjust away more than thirty percent of your losses, you should have documentation for every single adjustment. Some carriers will reject studies that look over-tuned.
When a Loss Study Is Not Useful
Loss studies break down in a few specific scenarios. If the account has fewer than five claims over the entire experience period, the credibility is so low that the result is essentially a guess dressed in a spreadsheet. There is no workaround except to fall back to manual rates. Similarly, accounts with major operational changes during the study period — mergers, facility closures, significant workforce reductions — produce unreliable results. I once saw a loss study for a distribution center that had lost two-thirds of its floor space to a fire mid-period. The adjusted numbers looked reasonable on paper. They were completely wrong in practice because the remaining twelve months of data had no comparable baseline. In those cases, shorten the experience period, exclude the distorted year entirely, or use a manual rate override. Do not force a loss study where the data cannot support it.
Tools and Templates
You can build a Loss Study Guide Checklist from scratch in any spreadsheet program. I use a combination of Excel for the main workbook and a separate Python script for trend factor application and outlier detection. The script runs in about four minutes on a dataset of roughly two thousand claims. The manual equivalent takes most people four to six hours. There are commercial platforms like AXIS, ClaimSpy, and ISO's own loss study tools that automate much of this. They are faster but less transparent. If you are submitting studies for regulatory review or rate filing, transparency matters more than speed. A black-box output will raise questions every time. I include a downloadable template based on my standard workflow. It covers data import, adjustment tracking, credibility calculation, and output formatting in one file. The formulas are visible and commented. Adapt it to your carrier requirements rather than trying to make your data fit the template.
