Building a Data-Driven Digital Product with Claude Fable 5 + Ubersuggest MCP

You spend a week building a digital product. You polish the design. You list it. And then — almost no traffic.

The design was never the problem. The problem is that the niche was chosen by intuition, and intuition tends to land you in exactly the corner of the market where nine other people are already competing on price.

This is the written companion to the video where I build a $14.99 product for career changers, and the niche is picked entirely from real keyword data — not a hunch. The whole thing takes three prompts once the tooling is wired up.

Prefer to watch it on YouTube?

Watch on YouTube — the three prompts are in the video description.

What you need

Claude Code with the Fable 5 model. Fable takes noticeably longer to think than the faster models, but for research work — where the whole point is weighing options against each other — the extra thinking time is what you are paying for.
An Ubersuggest account (Google sign-in is fine).

Step 1 — Pick the right project first

Open the Code tab in Claude and choose the project folder you actually want to work in before you prompt anything.

This step looks trivial enough to skip, and skipping it is the single most common way this workflow goes wrong. If the wrong project is selected, generated files land in the wrong workspace and you end up debugging output that was never meant for that codebase.

Select the folder, open it, and confirm the workspace name is the one you expect.

Step 2 — Connect Ubersuggest as an MCP server

Ubersuggest is not in Claude's connector directory yet, so it gets added as a custom connector.

In Claude, open Add → Manage connectors and search for Ubersuggest. You will not find it — that is expected, and it is why the next steps exist.
Go to ubersuggest.com and sign in with Google.
From the dashboard, scroll to the MCP Integration section. It is fairly new, so it sits near the bottom of the settings list.
That section gives you per-client setup instructions — Claude, Cursor, Codex, and others. For Claude you get two options: a URL or a config file. The URL is simpler.
Copy the MCP URL.
Back in Claude: Add custom connector, paste the URL, and give it a name (Ubersuggest).
Click Connect. The browser redirects to Ubersuggest, asks you to grant access, and you click Authorize.

When you return to the connector list, Ubersuggest should show a checkmark. That checkmark means the handshake succeeded — it does not yet mean queries work.

Step 3 — Smoke-test the connection before you trust it

Always fire one throwaway query before building anything on top of a fresh MCP server. Finding out it is misconfigured in the middle of a research run costs far more than thirty seconds now.

Use Ubersuggest to check "digital planner" and limit to 5 results.

Claude will ask permission to call the tool — allow it. If Ubersuggest returns a real response (in my run, 27 keyword suggestions), the server is genuinely wired up and you can move on.

Step 4 — Prompt 1: research niches from real data

The first real prompt does the research. The important constraint is the last line: search, do not guess. Without it, the model will happily invent plausible-sounding keyword volumes.

You are a digital product researcher.

Use the Ubersuggest MCP to pull ACTUAL keyword data and find digital product ideas.
Search — do not speculate or estimate from memory.

Process (do not skip any step):
1. Research  — pull real keyword data via Ubersuggest
2. Filter    — apply the criteria below
3. Profile   — build a customer profile for each surviving idea
4. Rank      — order by opportunity and explain the ranking

Focus on product formats: templates, checklists, workbooks, planners.

Filter criteria:
- Prefer HIGH CPC (proves buyers exist and are worth money to advertisers)
- Prefer LOW competition / paid difficulty
- Do not select on search volume alone

Claude will ask which niches you want compared. I gave it a list to work across — digital planner, budget tracker, and neighbouring ideas — rather than letting it pick one and stop.

Then it runs, and with Fable 5 you wait. This is the slow part of the workflow.

Step 5 — Read the numbers like a market pulse

This is the part of the video worth rewatching, because the ranking logic is the entire value of the workflow.

Four columns matter:

Metric	What it actually tells you
Volume	How many people want this
CPC	How much a click is worth — i.e. whether buyers spend money
SEO Difficulty	How hard it is to rank organically
Competition	How saturated the paid market already is (0 → 1)

Here is what came back:

Product idea	Volume	CPC	Competition	Verdict
Digital Planner	High	High	1.00	Saturated — skip
Budget Tracker	High	High	0.96	Saturated — skip
Career Changer Resume	Low	$1.68	0.16	✅ The opening

The instinct is to chase the first two rows: big volume, high CPC, obvious demand. That instinct is what puts you in the price war.

A competition score of 1.00 and 0.96 means advertisers have already bid those keywords to the ceiling. Demand exists, but you are entering a market where everyone is spending heavily to be seen. That is not a market you win with a new $15 PDF.

Career Changer Resume looks worse on the metric everyone checks first — volume is low. But CPC is $1.68 and competition is only 0.16. Read together, those two numbers say: buyers here are worth real money, and almost nobody is bidding for them.

Low volume plus high CPC plus low competition is not a weak niche. It is an underserved one — and it is why Fable ranked it first.

The rule

Rank by CPC ÷ competition, not by volume. Volume tells you how many people are looking. CPC tells you whether they buy. Competition tells you whether you can be seen.

Step 6 — Prompt 2: turn the winning idea into a product outline

Feed the top-ranked idea straight back in, along with the fields the research step already produced — product format, target customer, and price.

Build the product for rank #1 from the table above.

Use the product format, target customer, and price from that row.
Deliverable: a PDF / template / checklist package.

Keep it simple:
- No design software required
- No coding required
- Output an outline I can fill in

Fable takes its time again, and produces outline.md — the full product structure. Slow thinking, good output.

At this point you have a real product skeleton. But it is still just text, which makes it hard to judge.

Step 7 — Prompt 3: generate the landing page

The third prompt turns the outline into something you can look at and react to.

Take outline.md and build a frontend landing page for this product.
Single page, visual, ready to review.

The result for the career-changer product:

Price: $14.99, including Word, Google Docs, and Canva formats
Positioning: for people switching industries who need to rewrite an existing CV
Case studies: teacher → corporate, corporate → teaching, nurse → new field
Key objection handled: not knowing the vocabulary of the new industry — the product ships with AI-assisted translation of experience into the target industry's language
Structure: benefits, format list, before/after comparison, summary blocks, copy-and-fill objective templates, FAQ, closing CTA

Seeing it rendered is what tells you whether the idea holds up. Text outlines hide weak positioning; a landing page does not.

Why this works

Three prompts turned Claude into a research → selection → build pipeline:

Research with real data, explicitly forbidden from guessing
Rank by buyer value against competition, not by raw volume
Build the product and its landing page from the winning row

The leverage is not in the generation step — plenty of tools generate PDFs. It is in step 2. Choosing the niche from CPC-against-competition is what stops you from spending a week building for a market that was already saturated before you opened your laptop.

One honest caveat

Low volume is still low volume. This method finds underserved niches, not large ones. It is well suited to a $15 product where a modest number of motivated buyers makes it worthwhile — and poorly suited if you need volume at the top of the funnel.