Retention As An Optimization Objective

Why Treat Retention as an Objective?

A renewal price affects both the premium collected from a renewing policy and the chance that the policyholder renews. An insurer may therefore want to choose prices with the aim of retaining as many policies as possible. To formulate that aim, the insurer needs a supplied relationship between price and renewal probability, and a precise definition of what “retention” measures.

Retention is not automatically a measure of adequate pricing. If a lower price never reduces the renewal probability, maximizing retention alone points toward the lowest permitted price. That price might not cover expected claim costs, expenses, or a required margin. A retention objective should therefore be interpreted alongside a suitable premium or profitability constraint.

Definition: Let $q(p)$ be the supplied probability that an offered policy renews at price $p$. For a fixed set of $N$ policies, the expected number retained is $Nq(p)$, and the expected retention rate is $q(p)$. For segments $j=1,\ldots,J$, with $N_j$ policies and segment renewal probability $q_j(p_j)$, expected retained count is $\sum_{j=1}^J N_jq_j(p_j)$. If the total number offered, $N=\sum_jN_j$, is fixed, expected portfolio retention rate is $\sum_j N_jq_j(p_j)/N$.

The renewal probability may be provided directly for a set of price actions or through a function of price. Treat it as a probability: on the prices being considered, it must be between zero and one. A price outside the function’s stated range should not be evaluated by extrapolation unless that use is explicitly justified.

The expected number retained is a forecast, not a promise about the realized number. It is calculated from probabilities and does not require assuming that renewal decisions are independent: linearity of expectation gives the expected total even when individual decisions are dependent. Independence would matter for calculating the variance of the total, not for this expectation.

Formulating the Objective

With one price applying to all policies in a fixed portfolio, maximizing expected retained count $Nq(p)$ is equivalent to maximizing retention rate $q(p)$, because $N$ is constant and positive. If the price is chosen from a feasible set $\mathcal{P}$, the problem can be written as:

\max_{p\in\mathcal{P}} q(p).

For multiple segments, the corresponding objective is to maximize the weighted sum of segment renewal probabilities. Dividing by the fixed portfolio count gives the expected retention rate, so the two formulations select the same price actions.

\max_{(p_1,\ldots,p_J)\in\mathcal{P}} \frac{\sum_{j=1}^J N_jq_j(p_j)}{\sum_{j=1}^J N_j}.

The weights $N_j$ matter. An unweighted average of segment renewal probabilities gives a small segment the same influence as a large segment, and generally does not equal portfolio retention. A weighted average is appropriate when the objective is the expected fraction of the offered portfolio that renews.

Why the Expected-Count Formula Holds

For each offered policy $k$, define an indicator $I_k$ that equals 1 if it renews and 0 otherwise. The total number retained is $R=\sum_{k=1}^{N}I_k$. If policy $k$ has renewal probability $q_k$, then $E[I_k]=q_k$. Therefore:

E[R] =E\left[\sum_{k=1}^{N}I_k\right] =\sum_{k=1}^{N}E[I_k] =\sum_{k=1}^{N}q_k.

The second equality follows from linearity of expectation. In a segment where each of $N_j$ policies has the same supplied probability $q_j(p_j)$, the segment contributes $N_jq_j(p_j)$ expected renewals. Adding those contributions gives the portfolio formula. If the portfolio size is fixed, dividing the expected count by $N$ gives the expected retained proportion.

Worked Examples

Worked Example: Retention With a Minimum Premium

A hypothetical insurer offers renewal prices from $300 to $600 to a group of 2,500 policies. Its supplied renewal-probability function is $q(p)=1.04-0.0012p$, where $p$ is the annual premium in dollars. The insurer requires a premium of at least $420. Find the price that maximizes retention among the permitted prices, the resulting renewal probability, and the expected number retained.

Method. Optimize the supplied renewal probability over the feasible price range. The probability decreases as price increases: for any $p_2>p_1$, $q(p_2)-q(p_1)=-0.0012(p_2-p_1)<0$. Thus the least feasible price gives the greatest retention.

Step 1: Identify the feasible prices. The stated price range is $300 to $600, and the minimum-premium requirement is $420. The feasible range is therefore $420 to $600.

Step 2: Select the retention-maximizing price. Since $q(p)$ is decreasing on this range, select its lowest price, $p=\$420$.

Step 3: Calculate the renewal probability. Substituting the selected price gives $q(420)=1.04-(0.0012)(420)=1.04-0.504=0.536$.

Step 4: Calculate expected retained count. The expected number retained is $2{,}500(0.536)=1{,}340$ policies.

Check. The probability 0.536 is between zero and one, the selected price meets the $420 minimum, and $1{,}340/2{,}500=0.536$. The common mistake is to select $300 because it is the lowest price in the function’s domain; that price violates the minimum-premium requirement.

Worked Example: Weighted Retention Across Two Segments

A hypothetical insurer has 600 policies in segment A and 400 in segment B. It must choose a common action $x$ from $\{0,40,80\}$, measured in dollars. Segment A’s price is $p_A=500+x$, with supplied renewal probability $q_A=0.92-0.001x$. Segment B’s price is $p_B=800+2x$, with supplied renewal probability $q_B=0.80-0.001x$. The insurer requires a portfolio average premium of at least $640. Find the action that maximizes expected portfolio retention subject to that requirement.

Method. For each permitted action, calculate the weighted average premium and the expected retained count in each segment. Exclude actions that fail the premium requirement, then compare retention among the feasible actions.

Step 1: Calculate average premium for each action. The portfolio average is $(600p_A+400p_B)/1{,}000$. At $x=0$, it is $(600(500)+400(800))/1{,}000=(300{,}000+320{,}000)/1{,}000=\$620$. At $x=40$, the segment prices are $540 and $880, so the average is $(324{,}000+352{,}000)/1{,}000=\$676$. At $x=80$, the prices are $580 and $960, so the average is $(348{,}000+384{,}000)/1{,}000=\$732$.

Step 2: Apply the premium requirement. The action $x=0$ fails because $620 is below $640. Actions $x=40$ and $x=80$ satisfy the requirement.

Step 3: Calculate retention at each feasible action. At $x=40$, $q_A=0.92-0.040=0.88$ and $q_B=0.80-0.040=0.76$. Expected retained counts are $600(0.88)=528$ and $400(0.76)=304$, totaling 832, or $832/1{,}000=0.832$. At $x=80$, $q_A=0.92-0.080=0.84$ and $q_B=0.80-0.080=0.72$. Expected counts are $600(0.84)=504$ and $400(0.72)=288$, totaling 792, or $0.792$.

Step 4: Choose the best feasible action. Since $0.832>0.792$, select $x=40$, which yields a portfolio average premium of $676 and an expected retention rate of 0.832.

Check. The retained counts add to 832, and the premium calculation weights each segment by its policy count. The common mistake is to take the simple average of the two segment premiums or renewal probabilities; that would give equal weight to segments with different sizes.

Worked Example: Retention Subject to a Contribution Requirement

A hypothetical insurer is considering annual prices of $240, $280, $320, or $360 for a class of policies. The supplied renewal probability is $q(p)=1.10-0.0015p$. Expected variable cost is $200 for each renewed policy. For this exercise, define expected contribution per policy offered as $q(p)(p-200)$, and require it to be at least $100. Which price maximizes retention among the permitted prices that meet the contribution requirement?

Method. Calculate renewal probability and expected contribution at every permitted price. First identify the feasible prices, then choose the one with the largest renewal probability. The contribution amount here is a supplied constraint measure; it is not being used as the objective.

Step 1: Evaluate the probabilities and contributions. At $240, $q=1.10-(0.0015)(240)=0.74$, and contribution is $0.74(240-200)=0.74(40)=\$29.60$. At $280, $q=1.10-0.420=0.68$, and contribution is $0.68(80)=\$54.40$. At $320, $q=1.10-0.480=0.62$, and contribution is $0.62(120)=\$74.40$. At $360, $q=1.10-0.540=0.56$, and contribution is $0.56(160)=\$89.60$.

Step 2: Apply the constraint. The $240 and $280 prices fail the $100 requirement. The $320 and $360 prices also fail it: their contributions are $74.40 and $89.60, respectively. Therefore, none of the permitted prices is feasible.

Step 3: State the result correctly. There is no retention-maximizing feasible price because the feasible set is empty. A price cannot be selected merely because it has the highest retention among the options; every option violates the stated contribution requirement.

Check. All four contributions are below $100, so the constraint rules out every option. The common mistake is to select $320 because it has higher retention than $360 while meeting a mistakenly rounded or misread contribution threshold. With the stated $100 threshold, there is no solution among these discrete choices.

Interpreting the Constraint and Avoiding Traps

The examples illustrate the central limitation of a retention-only objective. If renewal probability decreases with price, reducing price increases or preserves expected retention. Without a lower price bound or another requirement, the objective may push the selected price to the bottom of the permitted domain, regardless of whether that price is adequate.

A minimum-premium constraint can exclude prices below a specified level. A profitability-related constraint can instead require a minimum expected contribution or other defined financial measure. The exact constraint must be stated in the problem: the meaning of “profit” depends on which costs, expenses, and other components are included. Do not silently substitute a premium floor for a contribution requirement, or assume that a premium above expected claims necessarily satisfies every profitability standard.

A constraint can also make a problem infeasible. As in the third example, if no permitted price meets the required threshold, there is no price that solves the constrained problem. The correct response is to report infeasibility, not to choose the least-bad option and describe it as feasible. Any change to the threshold or available actions would be a separate decision.

Use supplied probabilities only on their stated domain. Check that each evaluated probability is between zero and one.
Keep the portfolio denominator fixed when comparing retention rates. If the number of policies offered changes, the interpretation of a rate comparison may change too.
Weight segments by policy count. Expected portfolio retention is not generally the unweighted mean of segment probabilities.
Apply every constraint explicitly. First determine which choices are feasible; only then maximize retention.
Do not confuse expected and realized retention. A forecast of 832 renewals need not mean exactly 832 policies will renew.

Key takeaway: Formulate retention as expected renewed count or expected retained proportion using the supplied renewal probabilities. Because retention alone may reward an inadequate price, optimize it over an explicitly stated feasible set that includes any required premium or profitability condition.

Check Your Understanding

For each problem, identify the feasible choices before comparing retention.

A hypothetical class contains 900 policies. Its renewal probability is $q(p)=0.97-0.001(p-450)$ for prices from $400 to $550, and the minimum permitted price is $460. Find the retention-maximizing price, renewal probability, and expected retained count. (Answer: $460; $q=0.9600$; 864 expected renewals.)
A portfolio has 300 policies in segment A and 700 in segment B. Choose $x\in\{0,25,50\}$. Prices are $p_A=600+x$ and $p_B=400+2x$, with probabilities $q_A=0.90-0.001x$ and $q_B=0.85-0.003x$. The weighted average premium must be at least $500. Which action maximizes retention, and what is the expected retained count? (Answer: $x=25$; average premium $502.50; 805 expected renewals, or 0.8050 retention.)
A hypothetical insurer can charge $180, $210, or $240. The supplied renewal probability is $q(p)=0.95-0.002(p-150)$, variable cost per renewal is $160, and expected contribution per policy offered is defined as $q(p)(p-160)$. The contribution must be at least $45. Which price maximizes retention among feasible prices? (Answer: $240; contribution $61.60; renewal probability 0.7700.)
A class of 500 policies has a supplied renewal probability of 0.82 at a price of $700 and 0.76 at a price of $760. The permitted choices are those two prices, and the minimum-premium requirement is $720. Find the feasible retention-maximizing choice and expected number renewed. (Answer: $760; 0.7600 retention; 380 expected renewals.)
A portfolio has two segments, with 200 and 800 policies. Under action A, their supplied renewal probabilities are 0.90 and 0.75; under action B, they are 0.84 and 0.80. A minimum average premium requirement makes both actions feasible. Which action maximizes expected retention, and by how many expected renewals? (Answer: Action B; 808 expected renewals versus 780 under A, a difference of 28.)

What You'll Learn