Boost Your Prompt Score: Examples, Reasoning, Model Fit

Same instructions, three small add-ons, a noticeably better answer — that's Day 4. Today we chase the 20 Enhancement points.

You've done the heavy lifting. Days 2 through 4 fixed Completeness, Specificity, and Structure. Those are the bones of a strong prompt.

Today we add the polish. Enhancement is the smallest category at 20 points, but it's the fastest to win. Three small moves, and your output jumps in quality.

Here's the plan. We'll add one example, ask the model to reason, and match the prompt to your AI. Then you rewrite one prompt and re-score it.

What Enhancement actually rewards

Enhancement is the category that asks: did you go beyond the basics?

A prompt can be complete, specific, and well-structured and still feel plain. Enhancement is the layer that makes the model perform at its best.

The scorer looks for a few signals here. The three that matter most today are an example, a reasoning request, and model fit. Let's take them one at a time.

Move 1: Add one example

This is the single most powerful upgrade you can make. It's called few-shot prompting, which means showing the model an example before asking it to work.

Why does it help so much? Describing what you want is hard. Showing it is easy. When the model sees one good example, it copies your pattern instead of guessing.

Think about a new teammate. You could explain your email style in three paragraphs. Or you could forward them one great email and say "like this." The second way works better. Models are the same.

Here's a prompt with no example:

Write a short product description for a stainless steel water bottle.

Now here's the same prompt with one example added:

Write a short product description for a stainless steel water bottle.

Match the style of this example:

Product: Wool hiking socks
Description: Built for long trails and longer days. These merino
socks wick sweat, fight blisters, and stay cozy from sunrise to
summit. Your feet will thank you at mile twelve.

See the difference? The second prompt locks in tone, length, and rhythm. You're not hoping the model guesses your voice. You're handing it the pattern.

How many examples?

Start with one. One clear example often does the whole job.

If your results still wander between runs, add a second or third. That shows the model the range you'll accept. Past three, you rarely gain enough to justify the extra length.

Move 2: Ask it to think step by step

The second move is chain-of-thought prompting. In plain terms, you ask the model to reason out loud before it answers.

This matters for any task with logic in it. Math, planning, comparisons, decisions, debugging. When the model works through its steps, it catches its own mistakes instead of jumping to a wrong answer.

The phrase is short. You add something like "Think through this step by step before giving your final answer."

Watch what it does for a real question:

A team of 4 finishes a project in 6 days. We add 2 more people at
the same pace. Roughly how many days will it take now?

Think through this step by step, then give your final answer.

By asking for the steps, you let the model show its work. If the logic is off, you can see exactly where. That's far better than a confident wrong number with no explanation.

A bonus: asking for reasoning often reduces a hallucination, which is when the model states something false with full confidence. Made-up facts slip through less often when the model has to justify each step.

Move 3: Match the model you're using

The third move is model fit. The same prompt can land differently across LLM tools. An LLM is the large language model behind ChatGPT, Claude, Gemini, and the rest.

You don't need a separate prompt for every tool. A clear, complete prompt works reasonably well everywhere. But small tweaks help.

Some models respond well to structured sections and a clearly named role. Others prefer plain, conversational instructions. The honest way to learn the difference is to test.

The practical rule: write your prompt, run it in the tool you actually use, and keep the version that works. Then save that version so you never redo the test.

The Day 4 rewrite

Now you do all three. Grab the prompt you've been improving all week. Add these layers in order.

Here's a before-and-after to copy the shape of.

Before:

Help me write a cold email to a potential client about our
bookkeeping service.

After:

You're a sales copywriter helping me in ChatGPT.

Write a cold email to a small bakery owner offering monthly
bookkeeping. Keep it under 120 words, warm and plain, no jargon.

Match the tone of this example:
"Hi Sam, saw your new spring menu — looks great. Quick question:
who's handling the books while you handle the croissants? I help
local shops like yours stay tax-ready without the spreadsheet
headache. Worth a 10-minute chat?"

Think about the bakery owner's daily pain points first, then write
the email.

The after version has an example, a reasoning nudge, and a named model. All three Enhancement signals, in one short prompt.

Re-score and lock it in

Time to check your work. Paste your rewritten prompt into the free prompt scorer and watch the Enhancement number.

If you added a real example, that line should climb. The total band may tick up too, from Good toward Excellent. Read the suggestions it gives. They'll tell you which signal is still missing.

If your score barely moved, check three things. Is your example specific enough to copy? Did you skip the reasoning line on a task that needed it? Did you name your tool?

You don't have to nail this from memory either. When you build prompts in our template builder, example slots and reasoning steps are part of the structure. The Enhancement points come built in.

That's Day 4 done. Your prompt now has bones and polish. Tomorrow we measure exactly how far you've climbed.

Keep going

Next → Day 5: Re-Score and Measure Your Jump

Or see the full Score Your Prompt: The 7-Day Challenge series.