Constant-Sum Questions: How to Measure Trade-Offs by Asking People to Allocate

A constant-sum question asks respondents to distribute a fixed total — almost always 100 points or 100% — across several options, so the amounts have to add up to that total. Because giving more to one option means giving less to another, it captures the relative weight of each item in a way a simple rating never can.

The short answer on why they matter: constant-sum questions force genuine trade-offs and produce ratio-level data, so you learn not just that customers value speed over price, but roughly how much more — which is exactly what you need for prioritization, budgeting, and pricing decisions. When you need magnitude and not just rank order, constant sum is the tool.

How a Constant-Sum Question Works

The respondent sees a list of items and a pool of points to allocate. A classic example:

"You have 100 points. Divide them across these factors based on how important each is when choosing a project-management tool. The points must add up to 100."

• Ease of use: ___
• Price: ___
• Integrations: ___
• Customer support: ___
• Reporting: ___

If a respondent puts 40 on ease of use and 10 on reporting, you have learned that ease of use is roughly four times as important to them. Average those allocations across your sample and you get a clean, weighted importance map — far more decision-ready than five separate "rate this 1–5" questions that all come back as 4s.

Constant Sum vs. Other Question Types

Ranking tells you A beats B beats C. Constant sum tells you A beats B by a lot and B beats C by a hair. That magnitude is the whole point.

When to Use Constant-Sum Questions

• Importance weighting. Quantify how much each driver contributes to a decision — perfect for feeding a weighted scoring model or a prioritization framework.
• Budget and resource allocation. "How would you split a $100 budget across these add-ons?" mirrors a real spending decision and exposes willingness to pay.
• Share-of-preference and share-of-wallet. Ask people to allocate their next 10 purchases or their monthly spend across brands.
• Attribute trade-offs in pricing and packaging. Understand which features justify a premium when customers cannot have everything.

The Pitfalls (and How to Avoid Them)

Constant-sum questions are powerful but notoriously easy to get wrong:

1. Math friction. Forcing totals to hit exactly 100 is cognitively taxing, especially on mobile. Respondents abandon or fudge the numbers. Keep the list short (four to six items) and show a running total.
2. Too many items. Allocating across ten options is miserable. Beyond six or seven, switch to MaxDiff.
3. Anchoring and rounding. People gravitate to round numbers (25/25/25/25) and anchor on the first item. Randomize item order to spread this out.
4. No reasoning captured. A respondent gives price 50 points — but is that because price is a dealbreaker, or because their boss controls the budget? The number alone cannot tell you.
5. Low-effort splitting. Speeders dump everything on one item or split evenly to escape. Without a quality check, this noise sits in your averages.

Why Constant Sum Has Been Hard — Until Now

Traditional survey builders treat constant-sum as a finicky widget: you fight with validation rules, respondents fight with the math, and at the end you have a column of numbers with zero context. Tools like SurveyMonkey, Typeform, and Qualtrics can collect the allocation, but they cannot ask the respondent why they spent their points the way they did — and the "why" is where the strategy lives.

An AI-native platform like Koji reframes the whole exercise. Koji supports all six structured question types — open_ended, scale, single_choice, multiple_choice, ranking, and yes_no — and you can run a constant-sum-style allocation as a structured prompt inside a conversational interview. The difference is what happens next: Koji's AI interviewer reads the allocation and follows up automatically — "You put 50 points on price and almost nothing on integrations. What's driving that?" — capturing the reasoning a static survey throws away.

Because the interview is conversational, the math feels like a discussion rather than a spreadsheet, which lifts completion. Across your whole sample, Koji aggregates the average weights into a chart, themes the open-ended reasons behind the allocations into a codebook, and applies a 1–5 quality score so that even-splitting speeders do not skew your importance map. You get ratio-level trade-off data and the narrative behind it, with no moderator and results in hours — the kind of mixed quantitative-and-qualitative read that used to require a dedicated research vendor.

A Worked Example

A SaaS team wants to know what to build next. They give 200 prospects 100 points to split across five candidate features. The averages come back: Automations 38, Mobile app 27, Advanced reporting 18, SSO 11, White-label 6. Ranking alone would have said "Automations first" — but the constant-sum gap shows Automations and Mobile together command two-thirds of perceived value, while SSO and White-label barely register. The AI follow-ups reveal that SSO scored low only because most respondents were on small teams — a segment insight a raw average would have hidden. That is a roadmap decision backed by both magnitude and reasoning.

Constant Sum vs. MaxDiff: Which Should You Use?

Both force trade-offs, but they fit different jobs. Constant sum is best when your list is short (four to six items) and you want a direct, intuitive read of relative weight that stakeholders can grasp instantly — "price got half the points." MaxDiff (best–worst scaling) is the better choice when your list is long: instead of allocating points across fifteen items at once, respondents see small subsets and pick the most and least important from each, and a model reconstructs the full ranking. MaxDiff produces cleaner data on long lists and resists the round-number anchoring that plagues constant sum, but it is less transparent — the scores come from a model rather than directly from the respondent. Rule of thumb: six items or fewer, reach for constant sum; more than that, switch to MaxDiff.

Analyzing Constant-Sum Results

The headline number is the mean allocation per item across your sample — that is your importance weight. But the average can lie if the sample is bimodal, so always look at the distribution: an item averaging 25 points could mean everyone gave it 25, or half gave 50 and half gave zero. Those are completely different stories. Segment the allocations by audience to catch it, and cross the weights against the open-ended reasoning to understand why a segment spends its points the way it does. Feeding the resulting weights into a weighted scoring model turns soft preference into a defensible prioritization.

Related Resources

• Structured Questions Guide — how Koji's six question types combine in one study
• Ranking vs. Rating Questions — when order is enough and when you need magnitude
• MaxDiff Analysis — the better choice for long priority lists
• Van Westendorp Price Sensitivity Meter — pairing allocation data with pricing research
• Feature Prioritization Surveys — turning importance weights into a roadmap
• Weighted Scoring Model — feeding constant-sum weights into a decision matrix