Every welfare aziendale programme generates claim data. Most of it sits in platform export files or vendor-provided annual reports, used for accounting purposes and then filed away. It is rare for an HR team to read the same data as an analytical resource rather than an administrative record.
The distinction matters because the data contains patterns that are not visible when it is read as a ledger. This post describes five specific signals that are present in typical claim history from a mid-size Italian employer. They do not require sophisticated tooling to detect. They require reading the data with a different question in mind.
Signal 1: Budget stranding by category
The first signal is the simplest. For each benefit category in your welfare catalogue, what is the ratio of budget allocated to budget actually claimed?
Most welfare programmes have categories where the gap between allocation and utilization is large. The category carries budget but employees are not spending it. This can happen for several reasons: the category may be poorly specified for your workforce's actual needs, the claims process may be too cumbersome for the benefit amount, employees may not understand that the category is available, or the benefit may be structurally less relevant to your current workforce composition than it was when the catalogue was originally designed.
A category with a 60 percent budget utilization rate is allocating 40 percent of its budget to a pool that goes unclaimed and then expires. In a welfare aziendale arrangement, expired credits that employees did not spend are typically recovered. That recovery is not a cost saving. It represents benefit value that was nominally offered but never delivered.
The stranding rate is the first signal because it is the most direct measure of misalignment between what HR is offering and what employees are actually using. A welfare programme with average category budget utilization below 70 percent has a configuration problem that claim data can help diagnose.
Signal 2: Monotonic decline in a previously stable category
The second signal requires time series data rather than a snapshot. Take any category that had stable utilization for two or more years and then began declining. A monotonic decline, three or more consecutive periods of falling claim rates, is a signal distinct from normal variance.
Normal variance in welfare utilization looks like fluctuation: a quarter up, a quarter down, no consistent direction. A monotonic decline looks like a trend. Something changed. The most common causes are: a change in the employee population that reduced the share of employees for whom this category is relevant; a change in the external availability of the underlying benefit (a competing category became more attractive, or an external subsidy program changed the economics); or a change in the claims process or documentation requirements that made the category harder to use.
Each of these causes points to a different response. A population change suggests reconsidering the budget allocation for that category. A competitive change might mean redesigning the category to be more attractive. A process change suggests fixing the claims friction. Reading the decline signal without diagnosing its cause leads to removing categories that could be fixed rather than replaced.
Signal 3: Segment divergence above 30 percentage points
The third signal requires segmenting the utilization data by at least one workforce variable. Age bracket is the most commonly available and often the most informative. Job family or department is the second most useful variable when age data is not granular enough to be meaningful.
When utilization for a given category varies by more than 30 percentage points between the highest-utilizing segment and the lowest-utilizing segment, the uniform budget allocation is significantly misallocating value. Thirty percentage points means a segment claiming the category at 55 percent rate sits alongside a segment claiming at 25 percent rate or lower. At that scale of divergence, the allocation is serving one part of the workforce at the expense of another.
This signal is most actionable when the high-utilizing segment and the low-utilizing segment are both large enough to matter. A category that 60 percent of your 28-to-35 employees use and only 8 percent of your 48-to-55 employees use is a different configuration problem depending on the relative size of those segments in your workforce. If 60 percent of your employees are in the older bracket, the budget weight of this category is serving the minority.
Signal 4: Claim timing concentration
The fourth signal is less commonly analyzed but reveals something about claim behavior that average utilization rates do not show: when during the year are claims being submitted for a given category?
A category where 80 percent of annual claims are submitted in November and December is a different situation from a category with even monthly distribution. December-concentrated claims typically indicate that employees are submitting claims to use up expiring credits rather than because they organically needed the benefit in December. This is a form of forced utilization rather than genuine preference expression.
Forced utilization inflates the apparent popularity of a category. It can make a category look like a strong performer in terms of budget utilization rate while masking the fact that most employees are spending credits there as a last resort rather than as a genuine first choice.
When you see strong budget utilization concentrated in the final two months of the benefit year, the question to ask is whether employees are using this category because it is genuinely useful to them, or because it is the most frictionless place to spend the credits they have not used elsewhere. The answer changes how you should weight the category in next year's allocation.
Signal 5: New-hire versus tenured-employee utilization divergence
The fifth signal requires cross-referencing welfare claim data with hire date. It asks: do employees hired in the last 12 months use the welfare programme differently from employees who have been with the company for more than three years?
There are legitimate reasons to expect some divergence. New employees are learning the platform, may not fully understand what is available, and are making claims decisions in the first year without the experience of having used the programme before. These onboarding effects are real and expected.
What is not expected is a persistent large divergence that does not normalize after six or twelve months. If employees hired in the last year are claiming at substantially different rates across different categories than employees hired more than three years ago, and this pattern persists across multiple new-hire cohorts, it suggests something about how the programme is communicated or how the categories are designed that disadvantages newer employees.
This matters for two reasons. First, the new-hire experience affects retention at the early stages when turnover risk is highest. A welfare programme that new employees find confusing or poorly matched to their situation is a friction point during a period when the employment relationship is being established. Second, if your company has been growing through recent hiring, the new-hire cohort may be large enough that their divergent claim patterns are pulling aggregate utilization numbers in ways that mask the underlying issue.
What these signals do not tell you
This is the point where an honest account of data analysis requires drawing a clear boundary. These five signals identify where your current configuration is likely misallocated or underperforming. They do not tell you what the correct configuration is.
Claim history is a record of choices made within a constrained option set. Employees who never claimed an education benefit may not have wanted one, or may have wanted one that was specified differently than what was offered. An employee who claims meal vouchers consistently may be doing so because it is the most useful thing available, not because it is the benefit they would most want if the catalogue were different.
The signals in claim history are real and informative. They narrow the search space for configuration improvements. They do not substitute for the broader judgment about what your workforce actually needs, which requires combining the claim data signals with qualitative understanding of your workforce, the direction of your hiring, and the competitive benefit landscape you are operating in.
Using these five signals well means treating them as inputs to a decision process rather than answers to a configuration question. The data gets you to a better starting point. The HR judgment that evaluates that starting point determines whether the final configuration actually serves your workforce.