How to use the TRIPSpeed Shape tab and controls

Use this tab to decide whether the modeled speed shape behaves enough like your FIT-derived baseline across grade.

This is usually the first real shape-review step after ingest.

What this tab is for

The TRIPSpeed Shape tab helps you answer this question:

Does the modeled speed shape match the observed FIT baseline well enough to trust, or is there a specific mismatch I need to fix?

In practice, that means:

• reading the main chart first
• checking the summary diagnostics
• using the controls only when the mismatch is specific and believable

Do not start by moving controls just because they are available.

Step 1: Open the Speed Shape tab

From the Calibration Workspace, click Speed Shape.

The tab has two main jobs:

• the left side tells you how the current shape compares to FIT data
• the right side lets you adjust the shape deliberately

Step 2: Read the main chart first

Start with the large chart before touching any controls.

What you want to know first:

• does the modeled curve broadly follow the observed baseline?
• is the mismatch mostly uphill, mostly downhill, or near the center?
• does the fit look trustworthy across the observed range?

If the chart already looks directionally believable, that is a strong reason not to keep editing.

Step 3: Use the top metrics as a quick screen

Before drilling into the lower diagnostics, check the summary metrics at the top.

Use them like this:

• strong score and strong coverage usually mean fewer edits, not more
• poor alignment means the baseline shape is missing the observed center story
• obvious peak disagreement can mean the shape is centered wrong or behaving badly near the middle

When these look good, the safest move is often to leave the shape mostly alone.

Step 4: Use the lower diagnostics only after you already have a hypothesis

The lower diagnostics help you localize the mismatch.

They are most useful after you already noticed a problem in the main chart.

Use them to answer:

• is this a real shape problem or mostly noise?
• is the model systematically high or low in certain grade bands?
• is the problem broad or confined to one region?

These diagnostics should confirm what you already suspect from the main chart. They should not replace it.

Step 5: Make only small, targeted adjustments

The controls are there to help you adjust the shape, but they should be used carefully.

Before changing anything:

• make sure the mismatch is clear enough to describe in plain language
• make sure the problem is broad enough to matter for planning
• avoid treating tiny differences as something that must be fixed

Examples of a good reason to adjust:

• the shape is clearly shifted
• the uphill side is broadly wrong
• the downhill side is broadly wrong

Examples of a bad reason to adjust:

• one odd-looking patch of noisy data
• a good fit that is not mathematically perfect
• curiosity about what every control does

Step 6: Treat optional refinements as optional

Some controls are best treated as later-stage refinement tools rather than first-pass fixes.

Use them only when:

• the main shape is already broadly believable
• the remaining problem is local and specific
• you can explain why the adjustment improves realism rather than just making the chart prettier

If the main curve is still broadly wrong, do not spend time on small refinement controls yet.

Step 7: Recheck after every small change

Do not make a pile of edits and only review at the end.

The safest rhythm is:

1. identify the mismatch
2. make one small change
3. recheck the main chart
4. recheck the diagnostics
5. stop if the result now looks believable

If one change helps one region but makes another region worse, back out and reconsider.

When not to keep tuning

Stop adjusting if:

• the chart already looks believable
• the score and coverage are already strong
• you are solving cosmetic differences instead of meaningful planning differences
• every change improves one area only by making another area less believable

This tab is about reaching a believable shape, not forcing a perfect overlay.

A safe order of operations

For most users, the safest sequence is:

1. Read the main chart.
2. Check the summary metrics.
3. Use the lower diagnostics to localize the mismatch.
4. Make only small targeted adjustments if the mismatch is obvious.
5. Recheck after every change.
6. Stop as soon as the shape looks believable enough for planning.

What to do next

After the TRIPSpeed Shape looks believable, continue with:

• How to use the COT Shape tab and controls
• Review calibration outputs
• Apply calibration to a route