When you aggregate claim data across three or more annual cycles, something shifts in how you read it. A single year looks like a snapshot. Three years starts to look like a story. This is the central observation that shaped how we built Toduba's analysis layer: the unit of meaningful insight in benefits data is not the individual claim or even the annual utilization rate. It is the pattern that emerges when you map the same workforce across time.
We have been running this kind of longitudinal exercise since early in Toduba's development, reading anonymized utilization data from mid-market HR teams in northern Italy and the broader EU market. What follows is what that work has taught us about total rewards design, and where it leaves some important questions still open.
The first year is almost always misleading
When HR teams see their benefits utilization laid out by category for the first time, the instinct is to rank categories by claim rate and cut the lowest performers. That instinct is understandable, but it is often wrong.
Some low-claim categories are low because employees do not understand they exist. Others are low because the eligible workforce segment is genuinely small. A 12% claim rate on a learning and development allowance in a workforce where two-thirds of employees are in operational roles is not a problem. A 12% claim rate on a commuter subsidy in a company based in central Turin, where public transit is the dominant commute mode, probably is.
The first year of data tells you what was claimed. It does not tell you why. And without the why, a ranked list of categories is close to useless as a redesign input. This is the single most common mistake we see in how HR teams approach their annual benefits review: treating utilization rate as a proxy for fit, when it is at best a signal and at worst a distraction.
The drift problem
Between year one and year three, utilization patterns shift. Categories that were well-subscribed in the first year sometimes decay. New categories that started slow can build momentum. These movements reflect changes in the workforce, changes in how employees' lives are structured outside work, and changes in broader economic conditions that affect what people prioritize.
Commuter subsidies saw real declines across 2021 and 2022 for reasons that were obvious at the time. Wellness categories grew across the same period and in most mid-market Italian employers have not returned to pre-pandemic levels since. Childcare support has been growing steadily as median workforce age creeps upward in most EU member states, particularly in engineering and professional services firms that built their teams in the early 2020s and are now watching those same employees move into their late 30s.
None of these drift patterns are visible in a single year's data. They only become visible when you have three or more comparison points. And once they are visible, they change the design question. You are no longer asking which categories to keep and which to cut. You are asking where the trajectory of this workforce's needs is heading in the next 18 months, and whether your current mix is pointed in the same direction.
Where segment divergence shows up
Within a single workforce, claim behavior diverges sharply by segment. Age is the most obvious axis. A 27-year-old software engineer and a 51-year-old operations coordinator have structurally different lives. They have different commutes, different childcare situations, different health profiles, different financial situations, different ideas of what constitutes a meaningful benefit.
What surprises HR teams, in our experience, is how invisible this divergence is in standard utilization reports. A single aggregate claim rate on the gym subsidy might read as 31%. That aggregate can mask a reality where the claim rate among employees under 35 is around 58%, while employees over 50 are claiming below 6%. The aggregate number is not wrong. It is just not the number that helps you design.
Segment divergence becomes especially important in mixed-tenure workforces. A company that has been growing quickly, bringing in cohorts of younger employees while retaining a layer of longer-tenure staff, often ends up with a benefit mix that was designed for the older segment and is weakly relevant to the newer one. Three years of claim data, broken down by hire-year cohort and age band, makes this visible in a way that no survey can.
What the longitudinal view actually produces
Three years of segment-level claim data starts to surface patterns that feel more like preferences than individual decisions. You can see which employees reliably claim the same categories year after year, which employees shift categories as their life circumstances change, and which employees have never claimed anything significant despite being eligible.
That last group deserves attention. A non-claimant is not someone who has no needs. They are usually someone whose category has not yet been offered, or who does not understand the value of what is available, or whose life situation changed in a way that the existing mix no longer reflects. Non-claim behavior is a signal of mismatch, not indifference.
In total rewards design, the goal is to close the gap between what you offer and what employees actually need. Three years of longitudinal data is a better input to that process than any single-cycle survey, not because surveys are the wrong tool, but because surveys measure stated preferences and claim data measures revealed preferences. The gap between what people say they want and what they actually choose when it costs something is nearly always instructive, and sometimes substantial.
What claim history does not tell you
We want to be precise here about where this kind of analysis ends, because overstating what data can do is its own kind of problem.
Claim history does not explain causation. You can observe that claim rates in a particular category dropped 40% in a 12-month period. You cannot read from the data alone whether that drop reflects employee satisfaction, the availability of a better alternative outside the program, employer communication failures, a change in workforce composition, or some combination of all four. The data identifies where to investigate. It does not replace the investigation.
The other constraint is time lag. Claim data is a record of past behavior. A workforce going through rapid change, a significant hiring push, a restructuring, or a major demographic shift can move faster than the model tracks. When that happens, the longitudinal picture needs to be read alongside the qualitative context that the HR team already holds. Longitudinal analysis is a complement to HR judgment, not a substitute for it.
We are not building something that removes the HR professional from the room. We are building something that makes the time they spend in that room more productive.
The practical implication for redesign cycles
If you are entering an annual benefits review with less than two years of clean claim data, the most honest thing you can do is not attempt a full redesign. The better investment is in building the data discipline that will make future reviews evidence-based rather than assumption-based. Good decisions about total rewards configuration compound over time. An HR team with three years of segmented, longitudinal, category-level utilization data will consistently make better design decisions than a team relying on last year's survey and the intuition of whoever is in the room.
This is not the answer most HR leaders want to hear in November when the renewal window opens. But it is accurate. Total rewards design built on longitudinal claim data is slower to establish and faster to iterate than design built on annual surveys. The setup cost is a one-time investment in data capture and segmentation discipline. The ongoing return is a redesign process that starts from evidence rather than convention.
The question we keep coming back to is not whether this approach is better. It clearly is. The question is whether the HR teams responsible for these decisions have the tooling to make it practical inside a normal annual planning cycle. That is the problem Toduba is working on.