Compare every option with one baseline, criterion by criterion: better, the same or worse?
Ask a team to give an idea a score out of ten and you get a shrug. Ask whether it's better or worse than what you already have, and people answer quickly and agree more often. Concept screening builds on that. You pick one baseline, often today's solution or a strong competitor, and compare every other option with it, one criterion at a time.
Each cell in the grid gets a simple mark: + for better than the baseline, S for about the same, − for worse. Add up the pluses and minuses and each option gets a net score. The scores don't make the decision for you. They show which options are strong, which are weak, and which weak ones a single fix could rescue.
Use it when you have three to ten real options and need to narrow them fairly, when a team keeps arguing from gut feel, or when you need to show others how a choice was made.
Each option is marked better (+), same (S) or worse (−) than the baseline. A leads; B might be fixable.
Which options are genuinely in the running, described at the same level of detail?
What must the winning option do well?
What will you compare everything against?
On this criterion, is the option better than the baseline, about the same, or worse?
Which options come out clearly ahead, and which clearly behind?
Could a cheap change turn a minus into a plus, or could two options merge?
Concept screening is simple arithmetic, but it's built on how people actually make good judgments.
People are much better at relative judgments (is this better or worse than that?) than at absolute ones (is this a 7 or an 8?). Anchoring every judgment to the same baseline turns a hard question into an easy one, and makes different people's marks far more consistent.
Agreeing on criteria before marking forces the team to say what success means. That debate is often the most valuable part of the session, because it surfaces hidden disagreements about goals while they're still cheap to resolve.
When every option faces the same questions, in the same order, against the same reference, a pet idea can't coast on charm. And because each mark has a reason, everyone can see how the result was reached.
The grid shows exactly why an option scores badly. Often one or two minuses can be fixed, or the best parts of two options combined. Screening is as much a tool for improving concepts as for ranking them.
Write every option up at the same level of detail, ideally one card or slide each. Unequal descriptions skew the marks: a polished plan nearly always beats a napkin sketch.
List what the winner must do well for users, the organization and the team. Merge overlaps and keep the list short. Treat true must-haves, such as legal or safety rules, as pass/fail checks before marking.
Choose the reference: your current solution, a competitor's, or the option most people already favor. A strong baseline makes the test meaningful.
Work across one criterion at a time, marking every option +, S or − against the baseline. Record a short reason for each mark. Where people disagree, discuss briefly and note it.
Count the pluses and minuses for each option to get a net score. Look for clear leaders, clear losers and surprises. Ask whether the ranking makes sense, and what would change it.
Fix the cheapest minuses, combine strong partial ideas and drop clear losers. If the race is close, run a second round with the leader as the new baseline.
A whiteboard or spreadsheet grid with criteria as rows and options as columns, a one-page description of each option and about an hour. With AI, start from the prompt below.
Subject: a three-person team's meal-planning app with 40,000 monthly users. Baseline: today's banner ads. Criteria: revenue, user trust, build effort, retention, brand fit.
Free for everyone, a small but reliable income, and a steady stream of complaints in app-store reviews. Every concept is marked against this.
$3.99 a month to remove ads. + revenue, + trust, − build effort, S retention, − brand fit. Net 0.
One-off packs such as “budget week” or “dairy-free”. S revenue, + trust, − build effort, + retention, + brand fit. Net +2.
Send shopping lists to a delivery service for a referral fee. + revenue, − trust, − build effort, + retention, S brand fit. Net 0.
Pay what you want, if you want. − revenue, + trust, S build effort, S retention, + brand fit. Net +1.
Packs led at +2 and the tip jar came second at +1. A packs-plus-tip-jar hybrid, rescreened against packs alone, came out ahead.
The subscription looked like the obvious choice until the grid showed two minuses: months of billing work for a team of three, and a poor fit with a brand built on budget cooking. Packs led, and the tip jar nobody took seriously came second. Together they let the team drop ads without a paywall.
AI helps most at the two points where concept screening usually goes wrong: choosing the criteria and sticking to them. It can suggest criteria you'd miss, argue each mark from both sides and notice when the grid and your gut disagree. It can't know your users, budget or strategy, so the marks and the choice stay with you.
Get a fuller, cleaner list of criteria before anyone starts marking.
Here are our 7 criteria. Which overlap, which are missing, and which are really pass/fail must-haves?
Have AI attack and defend a mark so you set it for good reasons.
Option B got a minus on build effort. Argue for an S, then for a minus. Which case is stronger?
Find cheap fixes for the minuses holding back a promising option.
List Option C's minuses and suggest the cheapest change that could turn each into an S or a plus.
Check whether the ranking holds up when you change the reference point.
Rescreen all the options with our current leader as the baseline. Does anything now beat it?
List your options, pick a mode, then copy the prompt into Claude, ChatGPT, Gemini or Copilot. In coach mode the AI helps you set criteria and a baseline, then works through the grid one criterion at a time, asking for your marks first.
The “Open in” links put the prompt in the web address. If your challenge includes anything sensitive, copy and paste instead.
After the session, click one to copy it, then paste it into the same chat.