# How the decision check works

The decision check is rule-based and fully visible. This page shows every number it uses. The weights are editorial judgment, not measurements: they encode the trade-offs explained across this site.

[Take the decision check](https://microfrontends.dev/decide) [Jump to the weights](https://microfrontends.dev/decide/how-it-works#weights)

## On this page

-   [The steps](https://microfrontends.dev/decide/how-it-works#steps)
-   [Questions and points](https://microfrontends.dev/decide/how-it-works#readiness)
-   [Readiness bands](https://microfrontends.dev/decide/how-it-works#bands)
-   [Rules](https://microfrontends.dev/decide/how-it-works#rules)
-   [Approach weights](https://microfrontends.dev/decide/how-it-works#weights)
-   [Limits](https://microfrontends.dev/decide/how-it-works#limits)

## The steps

1.   **Readiness score (0–100)** Add the readiness points of your answers to Q1–Q5, subtract the penalties of Q6, Q8 and Q10, and clamp the total to 0–100. Q7 and Q9 do not affect readiness. Single team: the score is capped at 64.
2.   **Band** The score falls into one of three bands, which sets the recommendation.
3.   **Rules** Three rules override or extend the band’s recommendation. Some answers also rule an approach out of the recommendation; it keeps its place in the fit ranking.
4.   **Approach fit** Each approach adds up its weights for your answers. The percentage shown is that sum divided by the highest sum any approach can reach, so all approaches share one scale. Ties go to the approach listed first below.
5.   **What would change it** Every other answer to every question is tried one at a time. Changes that alter the recommendation or its top approach are shown, preferring changes to the recommendation, then neighboring answers, then earlier questions.

## Questions and readiness points

1.  Q1 Readiness points
    
    **How many teams work on this frontend?**
    -   1 0
    -   2–3 +18
    -   4–7 +28
    -   8+ +32
2.  Q2 Readiness points
    
    **Do teams need to deploy independently of each other?**
    -   No 0
    -   Nice to have +12
    -   Yes, critical +25
3.  Q3 Readiness points
    
    **How often do release coordination problems block you?**
    -   Rarely 0
    -   Sometimes +8
    -   Often +15
4.  Q4 Readiness points
    
    **Do you need to use more than one framework?**
    -   No 0
    -   During a migration only +10
    -   Yes, long term +13
5.  Q5 Readiness points
    
    **Is this a new project or an existing one?**
    -   New 0
    -   Existing, healthy +5
    -   Existing legacy to migrate +15
6.  Q6 Penalty
    
    **How strict is your performance budget?**
    -   Relaxed 0
    -   Moderate −3
    -   Strict −8
7.  Q7 Approach fit only
    
    **How important is SEO / server rendering?**
    -   Not important 0
    -   Somewhat 0
    -   Critical 0
8.  Q8 Penalty
    
    **How mature is your CI/CD and platform tooling?**
    -   Basic −12
    -   Solid 0
    -   Advanced 0
9.  Q9 Approach fit only
    
    **How strong must isolation between parts be?**
    -   Low 0
    -   Medium 0
    -   High 0
10.  Q10 Penalty
     
     **How consistent must the UX be across parts?**
     -   Pixel-consistent −6
     -   Mostly consistent −2
     -   Can vary 0

## Readiness bands

1.  0–34 **You probably don't need micro frontends** Stay with a modular monolith: a modular monolith or a monorepo with enforced boundaries.
2.  35–64 **Maybe, but start small** Build-time packages, plus a pilot of the best-fitting approach other than build-time packages, and signals to watch.
3.  65–100 **Micro frontends are a reasonable fit** The two best-fitting approaches.

## Rules

-   **Single team, single framework**
    
    Applies when Q1 “1” and Q4 “No”
    
    One team using one framework gets the benefits of clear ownership from a modular monolith, without the runtime cost, whatever the other answers.
-   **Legacy migration**
    
    Applies when Q5 “Existing legacy to migrate”
    
    Replacing a legacy app piece by piece calls for a strangler fig migration with a temporary composition layer, whatever the readiness score.
-   **High isolation**
    
    Applies when Q9 “High”
    
    Third-party or untrusted code brings iframes in as a candidate, because they are the only approach with a hard browser boundary.

-   **Single team: score capped at 64**
    
    Applies when Q1 “1”
    
    One team has no release coordination between teams to remove, so its readiness stays below the reasonable-fit band.
-   **Not recommended: Module Federation**
    
    Applies when Q4 “Yes, long term”
    
    Its sharing model works best when every team uses the same framework.
-   **Not recommended: Native Federation / import maps**
    
    Applies when Q4 “Yes, long term”
    
    Its sharing model works best when every team uses the same framework.
-   **Not recommended: single-spa**
    
    Applies when Q7 “Critical”
    
    Its approach page lists SEO-critical pages under avoid when.
-   **Not recommended: Web Components**
    
    Applies when Q7 “Critical”
    
    Its approach page lists SEO-critical pages under avoid when.
-   **Not recommended: iframes**
    
    Applies when Q7 “Critical”
    
    Its approach page lists SEO-critical pages under avoid when.

## Approach weights

Weight added to each approach’s fit for each answer. Empty cells are 0.

BT

Build-time packages

MF

Module Federation

NF

Native Federation / import maps

SS

single-spa

WC

Web Components

IF

iframes

SE

Server and edge composition

| Answer | BT | MF | NF | SS | WC | IF | SE |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Q1 1 | +1 |  |  |  |  |  |  |
| Q1 2–3 | +2 |  |  |  |  |  |  |
| Q1 4–7 |  | +1 | +1 | +1 |  |  |  |
| Q1 8+ |  | +2 | +1 |  |  |  | +1 |
| Q2 No | +2 |  |  |  |  |  |  |
| Q2 Nice to have | +1 | +1 |  |  |  |  |  |
| Q2 Yes, critical | −4 | +3 | +2 | +2 | +1 |  | +2 |
| Q3 Rarely |  |  |  |  |  |  |  |
| Q3 Sometimes |  |  |  |  |  |  |  |
| Q3 Often | −2 | +1 | +1 |  |  |  |  |
| Q4 No |  | +1 | +1 |  |  |  |  |
| Q4 During a migration only |  |  |  | +2 | +1 | +1 |  |
| Q4 Yes, long term | −2 | −1 | −1 | +3 | +3 | +3 | +3 |
| Q5 New |  | +1 | +1 |  |  |  |  |
| Q5 Existing, healthy |  |  |  |  |  |  |  |
| Q5 Existing legacy to migrate |  |  |  | +1 |  | +1 |  |
| Q6 Relaxed |  |  |  |  |  | +1 |  |
| Q6 Moderate |  |  |  |  |  |  |  |
| Q6 Strict | +2 | −1 | −1 | −1 | −1 | −3 | +2 |
| Q7 Not important |  |  |  |  |  |  | −1 |
| Q7 Somewhat |  |  |  |  |  |  | +1 |
| Q7 Critical |  | −1 | −1 | −2 | −1 | −2 | +4 |
| Q8 Basic | +2 | −1 | −1 | −1 |  |  | −2 |
| Q8 Solid |  |  |  |  |  |  |  |
| Q8 Advanced |  | +1 | +1 |  |  |  | +2 |
| Q9 Low |  |  |  |  |  |  |  |
| Q9 Medium |  |  |  |  | +1 |  |  |
| Q9 High |  |  |  |  | +1 | +4 |  |
| Q10 Pixel-consistent | +2 | +1 | +1 |  |  | −3 |  |
| Q10 Mostly consistent |  |  |  |  |  |  |  |
| Q10 Can vary |  |  |  |  | +1 | +1 |  |

## Limits

This is guidance based on ten answers, not a substitute for a proper architecture review with people who know your system.

When the logic changes in a way that can change results, its version number goes up. A result link made with an older version shows the current result with a note saying so.

Source: https://microfrontends.dev/decide/how-it-works
