Segment-Level Pricing Constraints

Why Pricing Guardrails Are Segment-Specific

A portfolio-wide average can look acceptable even when the proposed prices exceed a limit for one group of customers. Segment-level constraints address that problem by testing prices within defined groups, such as vehicle-use categories or property types. The constraint is evaluated for each segment using its own prices, weights, and baseline; a passing portfolio average does not make a failing segment pass.

The segment definition and weighting basis must be set before the calculation. In this tutorial, each policy has a fixed exposure weight $w_j$, such as the number of covered exposure units represented by that policy. Prices are annual premiums per policy, and weights and segment membership do not change between the baseline and proposal. These assumptions allow a direct comparison of weighted totals. If the business constraint instead defines a different denominator or weighting basis, use that stated definition.

Definition: For segment $s$, let $P_{sj}^{(0)}$ be policy $j$'s baseline premium, $P_{sj}$ its proposed premium, and $w_{sj}>0$ its fixed exposure weight. The segment's weighted average baseline and proposed premiums are $$ \bar P_s^{(0)}=\frac{\sum_j w_{sj}P_{sj}^{(0)}}{\sum_j w_{sj}}, \qquad \bar P_s=\frac{\sum_j w_{sj}P_{sj}}{\sum_j w_{sj}}. $$ The segment's average rate change relative to its baseline is $$ r_s=\frac{\sum_j w_{sj}P_{sj}}{\sum_j w_{sj}P_{sj}^{(0)}}-1. $$ A segment-specific rate-change guardrail with lower bound $L_s$ and upper bound $U_s$ is satisfied exactly when $L_s\le r_s\le U_s$. A proposed-price guardrail may instead directly limit $\bar P_s$, for example $A_s\le\bar P_s\le B_s$.

Calculate the Measure the Guardrail Actually Uses

A guardrail on average premium is a dollar constraint, whereas a guardrail on average rate change is a percentage constraint. They are not interchangeable. For example, a segment with higher baseline premiums can have a larger dollar increase but a smaller percentage increase than a segment with lower baseline premiums.

For a rate-change constraint, compare the proposed weighted premium total with the baseline weighted premium total for that same segment. Do not use the portfolio’s baseline as the denominator. Where the constraint is stated as a change in average premium, the segment's fixed exposure denominator cancels:

\frac{\bar P_s}{\bar P_s^{(0)}}-1 = \frac{\sum_j w_{sj}P_{sj}}{\sum_j w_{sj}P_{sj}^{(0)}}-1 =r_s.

The denominator is not necessarily the number of policies. It is the sum of the specified weights. If all policies receive equal weight, the result is the ordinary arithmetic average. If weights differ, calculating an unweighted average of individual premiums or percentage changes generally gives the wrong segment measure.

Worked Example: Testing Rate-Change Bands by Segment

A hypothetical insurer has two segments. In segment A, two exposure units have baseline premiums of $800 and $1,200; their proposed premiums are $880 and $1,320. In segment B, one exposure unit has a baseline premium of $600 and three exposure units have a baseline premium of $1,000 each; the respective proposed premiums are $660 and $1,080. The allowed average rate change is 5% to 12% for A and 0% to 8% for B. Test both segments.

Method. Calculate each segment’s change in weighted premium total divided by its own baseline weighted premium total. The exposure weights in A are 1 and 1; in B they are 1 and 3.

Segment A baseline total: $1(800)+1(1{,}200)=800+1{,}200=\$2{,}000$. Its proposed total is $1(880)+1(1{,}320)=880+1{,}320=\$2{,}200$. The increase is $\$2{,}200-\$2{,}000=\$200$, so $r_A=200/2{,}000=0.10=10\%$. Since $5\%\le10\%\le12\%$, segment A passes.

Segment B baseline total: $1(600)+3(1{,}000)=600+3{,}000=\$3{,}600$. Its proposed total is $1(660)+3(1{,}080)=660+3{,}240=\$3{,}900$. The increase is $\$300$, so $r_B=300/3{,}600=0.083333\ldots=8.3333\%$, to four decimal places. Because $8.3333\%>8\%$, segment B fails.

Check. Across both segments, the baseline total is $\$5{,}600$ and the proposed total is $\$6{,}100$, for a portfolio change of $500/5{,}600=8.9286\%$. That portfolio result does not alter segment B's failure. The common mistake is to check only the portfolio change; the correction is to test each segment against its own band.

Why the Weighted Rate-Change Formula Holds

The segment rate change can also be expressed as a weighted average of individual percentage changes, but the weights for that average are baseline premium amounts, not just exposure counts. This identity explains why an unweighted average of policy-level percentage changes can be misleading.

g_{sj}=\frac{P_{sj}-P_{sj}^{(0)}}{P_{sj}^{(0)}}, \qquad \frac{\sum_j w_{sj}P_{sj}^{(0)}g_{sj}} {\sum_j w_{sj}P_{sj}^{(0)}} = \frac{\sum_j w_{sj}\bigl(P_{sj}-P_{sj}^{(0)}\bigr)} {\sum_j w_{sj}P_{sj}^{(0)}} = \frac{\sum_j w_{sj}P_{sj}}{\sum_j w_{sj}P_{sj}^{(0)}}-1.

The first equality follows by substituting $P_{sj}^{(0)}g_{sj}=P_{sj}-P_{sj}^{(0)}$. The second follows by distributing the sum and subtracting the baseline total from the proposed total. Thus $r_s$ is an average of individual percentage changes weighted in proportion to $w_{sj}P_{sj}^{(0)}$. This result assumes positive baseline premiums and fixed weights. If a baseline premium is zero, its individual percentage change is undefined; use the segment’s total-premium ratio if that ratio is defined, or follow the constraint’s specified alternative definition.

Worked Example: Applying an Average-Premium Ceiling

A hypothetical insurer imposes dollar-average guardrails on two segments. In Urban, there are one and three exposure units with proposed premiums of $540 and $720, respectively; the segment average premium must not exceed $660. In Rural, there are two and one exposure units with proposed premiums of $420 and $950; the average must be between $550 and $620. Determine whether each segment passes.

Method. Calculate the weighted proposed premium total and divide by the segment’s total exposure weight. These are dollar limits, so no percentage-change calculation is needed.

Urban: the weighted total is $1(540)+3(720)=540+2{,}160=\$2{,}700$, and total weight is $1+3=4$. Thus $\bar P_{\text{Urban}}=2{,}700/4=\$675.00$. The ceiling is $660, so Urban fails by $\$675-\$660=\$15.00$ per exposure unit.

Rural: the weighted total is $2(420)+1(950)=840+950=\$1{,}790$, and total weight is $2+1=3$. Thus $\bar P_{\text{Rural}}=1{,}790/3=\$596.666\ldots=\$596.67$, rounded to the nearest cent. Since $\$550\le\$596.67\le\$620$, Rural passes.

Check. The combined weighted average is $(2{,}700+1{,}790)/(4+3)=4{,}490/7=\$641.43$. That portfolio average being below $660 does not rescue Urban: the Urban constraint applies to Urban alone. The common mistake is to average the two segment averages equally; the correction is to combine their weighted totals and divide by total weight only when a portfolio average is actually required.

Finding a Shared Adjustment That Meets Several Bands

A pricing proposal may be expressed as indicated premiums multiplied by a common adjustment factor $k$. For each segment, the factor must place its weighted proposed total inside that segment's allowed range. The same factor is feasible for all segments only if the permitted intervals overlap. This is a direct feasibility check; it does not require choosing or optimizing an objective.

Worked Example: Checking a Common Factor Across Segments

A hypothetical pricing team considers multiplying indicated premiums by a shared factor $k$. Segment C has baseline weighted total $2,000 and indicated total $2,300; its allowed average rate change is 5% to 12%. Segment D has baseline weighted total $3,000 and indicated total $3,180; its allowed rate change is 3% to 8%. Find the feasible range of $k$, then test $k=0.972$.

Method. With fixed exposure weights, multiplying all indicated premiums in a segment by $k$ multiplies its indicated weighted total by $k$. For segment $s$, the constraint is $L_s\le kI_s/B_s-1\le U_s$, where $I_s$ is the indicated weighted total and $B_s$ the baseline weighted total. Rearranging gives $(1+L_s)B_s/I_s\le k\le(1+U_s)B_s/I_s$.

Segment C: $1.05\le k(2{,}300)/2{,}000\le1.12$. Therefore $k\ge1.05(2{,}000)/2{,}300=2{,}100/2{,}300=0.913043\ldots$, and $k\le1.12(2{,}000)/2{,}300=2{,}240/2{,}300=0.973913\ldots$. Its permitted interval is approximately $[0.913043,0.973913]$.

Segment D: $1.03\le k(3{,}180)/3{,}000\le1.08$. Thus $k\ge3{,}090/3{,}180=0.971698\ldots$ and $k\le3{,}240/3{,}180=1.018868\ldots$. Its interval is approximately $[0.971698,1.018868]$.

Common feasible range: intersect the intervals by taking the larger lower bound and smaller upper bound. The result is $[0.971698,0.973913]$, so a shared factor is feasible.

Test $k=0.972$: Segment C's proposed total is $0.972(2{,}300)=\$2{,}235.60$, giving a change of $(2{,}235.60-2{,}000)/2{,}000=11.78\%$. Segment D's total is $0.972(3{,}180)=\$3{,}090.96$, giving $(3{,}090.96-3{,}000)/3{,}000=3.032\%$. Both satisfy their bands: $5\%\le11.78\%\le12\%$ and $3\%\le3.032\%\le8\%$.

Check. The selected factor lies within both intervals, and the direct rate-change calculations confirm both limits. The common mistake is to find a factor that works for one segment and assume it works for all; the correction is to intersect every segment's permitted interval before selecting a shared factor.

Practical Checks and Common Traps

A segment-level constraint is meaningful only when the grouping, baseline, weights, and measure match the stated rule. In a practical review, verify that the segment assignment has not changed between the baseline and proposed calculations, and that premiums and exposure weights use consistent units. If policies are reweighted, added, removed, or reassigned, a simple before-and-after comparison may mix price changes with changes in the portfolio composition.

Do not substitute a portfolio guardrail. A portfolio average can conceal a segment violation, as the first example shows.
Do not confuse dollars with percentages. A ceiling on average premium is not a cap on average rate change.
Use the stated weights. Averaging policy premiums equally is wrong when exposure weights differ.
Use each segment's own baseline. A rate-change denominator from another segment or from the full portfolio tests a different quantity.
Check both bounds. A proposal can violate a lower bound as well as an upper bound.
For a common adjustment, check feasibility first. If the permitted intervals have no overlap, no single factor can satisfy every segment's band.

Key takeaway: Define the segment metric exactly, calculate it with the segment's own baseline and weights, and test every segment separately. When one adjustment applies across segments, the permissible adjustment must lie in the intersection of all segment-specific feasible ranges.

Check Your Understanding

For each problem, use the stated weighting and test the segment constraints directly.

A segment has one exposure unit with a baseline premium of $1,000 and proposed premium of $1,100, and two units with a baseline premium of $500 each and proposed premiums of $540 each. Its allowed average rate change is 8% to 10%. Does it pass? (Answer: Baseline total $2,000; proposed total $2,180; rate change 9.0000%, so it passes.)
A segment has two exposure units with proposed premiums of $680 each and one unit with a proposed premium of $1,100. Its average-premium ceiling is $800. Does it pass? (Answer: Weighted average is $(2(680)+1(1{,}100))/3=\$820.00$, so it fails by $20.00.)
A segment has one exposure unit with baseline and proposed premiums of $400 and $440, and three units with baseline and proposed premiums of $800 and $840, respectively. Its maximum average rate increase is 5.5%. Does it pass? (Answer: Baseline total $2,800; proposed total $2,960; rate change $160/2{,}800=5.7143\%$, so it fails.)
Segment X has baseline and indicated weighted totals of $5,000 and $5,500, with a 4% to 8% rate-change band. Segment Y has totals of $4,000 and $4,200, with a 2% to 6% band. Find the common feasible interval for a shared factor $k$, and determine whether $k=0.98$ works. (Answer: X permits $k\in[0.945455,0.981818]$; Y permits $k\in[0.971429,1.009524]$. The intersection is $[0.971429,0.981818]$, so $k=0.98$ works.)
Segment A has a baseline total of $1,000 and a proposed total of $1,200; Segment B has a baseline total of $9,000 and a proposed total of $9,000. Each segment has a maximum allowed increase of 10%. Does the portfolio satisfy the guardrails? (Answer: The portfolio increase is $200/10{,}000=2\%$, but A's segment increase is 20%, so the segment guardrails are not all satisfied.)

What You'll Learn