Why does every ACA quoting tool claim smart recommendations and then ship a spreadsheet wrapper that ranks Silver plans by premium? Because plan ranking in this category should be constrained, explainable, and boring. The interesting part is what the ranker reads, what it ignores, and how it explains itself.
The QuoteTurbo plan ranker is a rule-based scoring function with broker-readable reasoning. It reads the household, the doctors, the prescriptions, and the preferences a broker captured during intake. It runs subsidy math against every Silver plan in the rating area. It returns three plans that represent meaningfully different trade-offs, with a one paragraph trace of why each plan won its slot. The broker reads the reasoning before anything reaches the client. The ranker suggests. The broker decides.
Key Takeaways
- Three plans, not one. Surfaces the trade-off the household is making, instead of hiding it.
- One paragraph of templated reasoning per plan. Brokers read it before sending anything to the client.
- Subsidy math (APTC, CSR, SLCSP) is inside the ranking. Net premium drives the order, not gross.
- Client data is processed within QuoteTurbo infrastructure.
- Override is a first-class citizen. The broker can replace any recommendation; the override is recorded against the quote.
What the ranker reads
Inputs are what separate a useful recommendation surface from a ranking widget. Six categories feed the ranker. None of them require the broker to leave the quote flow.
Household
Sizes 1 to 8, ages and tobacco use per applicant, dependent relationships.
Income
Projected MAGI for the coverage year. Used for APTC and CSR eligibility math.
Location
ZIP plus rating area lookup. Carrier participation varies by rating area, not state.
Doctors
Provider names or NPIs. Match strictness is configurable per agency.
Prescriptions
Drug names and dosages. Each plan's formulary tier and prior auth flag is read.
Preferences
Premium versus out of pocket weight, HMO versus PPO bias, brand carrier preference.
What it returns
One recommendation card per ranked plan. The card shows the plan, the metal tier, the net premium after APTC, the deductible, and the reasoning paragraph. The reasoning is what makes the output useful. A net premium number without the explanation is just a spreadsheet row in better typography.
Sample household
Family of 4 · ages 39, 37, 8, 5 · MAGI $86,000
FL Rating Area 22 (Broward County) · 2 in network requirements (pediatrician, OB-GYN) · 1 maintenance medication (Atorvastatin)
Florida Blue myBlue HMO 1416
SilverBoth required providers are in network. Atorvastatin is Tier 1, no prior auth. With CSR eligibility at 219% FPL, the Silver tier is the strict winner for this household.
Ambetter Balanced Care 11
SilverLower CSR variant, slightly broader network. Pediatrician is in network but OB-GYN is on a tier 2 referral. Worth a look only if the family wants broader network optionality.
Aetna CVS Health Bronze 5000
BronzeLowest premium option. CSR does not apply at Bronze, so the household trades $2,200 in annual CSR value for $1,392 in lower annual premium. Send only if the family explicitly wants the lowest monthly bill.
The third recommendation in the sample is the interesting one. The ranker did not put it second or first. It included it because the family might genuinely prefer the lowest monthly bill, and the broker should see that option with the math made explicit. The reasoning paragraph names the $2,200 versus $1,392 trade. The broker reads it and decides whether to send.
How the ranking works
Score every plan in the rating area
Each plan in the household rating area is scored on premium, network match, and formulary match. The score reflects the household preferences, not a generic ranking.
Apply subsidy math
APTC, CSR, and SLCSP math runs against every Silver plan. Net premium replaces gross premium in the ranking.
Pick three diverse plans
The ranker returns three plans that represent meaningfully different trade offs. Three Silver plans with similar premiums get collapsed into one suggestion.
Generate the reasoning paragraph
Each recommendation gets a one paragraph trace of why the plan won its slot. The broker reads it before sending anything to the client.
None of the four steps requires generative text. The reasoning paragraph is templated from the ranker's own score attribution, not a separate language model freelancing on top of the numbers. That is the right shape for insurance. Brokers can audit the trace and point at the specific input that drove the suggestion.
Why most vendor “AI plan finders” are not
Three patterns show up in the category. Spreadsheets with an AI badge, where the ranking is a fixed rule and the label is marketing. Generative wrappers, where a language model writes free text on top of plan data and occasionally invents a plan feature. And black-box rankers with no override, where the broker cannot disagree with the system. None of these are what the work actually needs.
The shape that matches the workflow is a rule-based ranker with visible score attribution and a real override path. The broker should be able to ask “why this plan” and get a specific answer from the score, not from a paragraph that sounds correct. If the ranker cannot point at the input that drove the suggestion, it is wallpaper.
When the ranker gets it wrong
Two cases regularly. The client just changed jobs and the broker knows MAGI is about to move, so the recommendation that wins on current income loses on projected. And the client has a non-obvious preference (a specific specialist, a parent on the same plan in a different rating area, a known formulary issue) that the intake form did not capture.
Both are override moments. The broker swaps the suggestion for the plan they wanted. The override is recorded against the quote so the broker can review the decision later.
FAQ
What does the plan ranker actually read?
Household composition, projected MAGI, ZIP and rating area, doctor list (NPIs if provided, names otherwise), prescription list, preferred metal tier weight, and any client preference signal the broker captured during intake. Client data is processed within QuoteTurbo infrastructure.
Why three plans and not one?
Brokers do not want a black-box answer. Three plans surfaces the trade-off the household is actually making, usually between premium and out-of-pocket. The broker decides which trade-off matches the client. The ranker suggests; the broker decides.
Can I override the recommendation?
Yes. Every recommendation includes a one paragraph reason. If you disagree, you swap in the plan you wanted. The override is recorded against the quote so the broker can review past decisions on the household.
Is this a model or a spreadsheet wrapper?
It is a rule-based scoring function, not a generative model. The inputs are household economics, network match, and formulary match. The output is a ranked list with per-plan score attribution. The reasoning paragraph is templated from the score components, not free-form generated text.
Does the ranker handle CSR eligibility?
Yes. CSR eligibility is part of the household input. When the household qualifies, the ranker weights Silver plans higher because Silver is the only tier where CSR applies. The reasoning paragraph calls out the CSR boost so the broker can explain it to the client.
For the subsidy math underneath the ranking, the ACA subsidy calculator, APTC calculator, and SLCSP calculator are the standalone walkthroughs. For the full data pipeline (sources, refresh cadence, source citations), see the methodology. For the full quote against live plans, use the free plan finder. For ranking logic embedded into a private platform on your own data, Devkrest builds that.
Figures shown are illustrative. Actual amounts depend on Healthcare.gov eligibility determination and current CMS plan filings. The IRS reconciliation on Form 8962 is the final number. Not insurance, tax, or financial advice.
Plan ranking is generated by software based on the inputs supplied. It does not substitute for broker judgment. Verify recommendations before sending to a household.
