The example is Marcus Reed, who runs Fieldstone CRM Consulting (both made up). Every Friday the workflow reads the newest Trustpilot reviews of pipedrive.com and adds one row of themes to Google Sheets. In the lesson the ProLevelMax call used a placeholder key, and the later steps ran on pinned SAMPLE reviews.
What are the most common mistakes?
| # | Mistake | Fix |
|---|---|---|
| 1 | Clicking Run in Collect to test | Read the fields, click Close, let n8n call |
| 2 | Key in the JSON body or URL | Bearer Auth credential, Bearer Token |
| 3 | Reading invalid key format as a broken node | Paste the real key and run again |
| 4 | Guessing field names | Copy them from the Collect form |
| 5 | Default timeout | Set Timeout to 2 minutes |
| 6 | Not checking field names in a real run | Adjust Keep rating and text to match |
| 7 | A prompt that allows any reply | Ask for JSON only, with three fixed keys |
| 8 | Sheet headers that don't match | Row 1 must match the field names exactly |
Why should you never test with Run?
1. Clicking Run in Collect. In the lesson, Collect is only for reading the source's fields: companyUrls, maxReviewsPerCompany, stars, languages, sort, date and more. Run would charge your balance. Close the form and let the HTTP Request node make the call. The lesson never ran Collect, so nothing was charged.
Where should the ProLevelMax key go?
2. Key in the JSON body or URL. The body holds the source and its input, nothing else. In Get new reviews (ProLevelMax), set Authentication to Generic Credential Type and Generic Auth Type to Bearer Auth, then paste the key into the credential's Bearer Token. The lesson's credential is named ProLevelMax - add your key. See bearer token for how n8n sends it.
3. Reading invalid key format as a broken node. With a placeholder in the credential, ProLevelMax answered Authorization failed: invalid key format. That means n8n reached ProLevelMax with the right URL and header. Paste your real key and execute the node again.
Why does the request return nothing useful?
4. Guessing field names. The body uses the exact names from the Collect form, inside input: "companyUrls": ["pipedrive.com"] (a list), "maxReviewsPerCompany": 30, "sort": "recency" for the newest, and "languages": ["en"]. scraper is the source id, procli_trustpilot_a, and maxItems is 30. If you switch to another review site, read that source's own Collect form, because each one has its own fields.
5. Default timeout. Collecting reviews takes a little while. The lesson sets the node's Timeout option to 2 minutes. The workflow timeout entry explains why long calls need room.
Why are the AI themes or sheet columns empty?
6. Not checking field names in a real run. ProLevelMax returns the reviews in items. Keep rating and text reads each review's rating, title, text and date, and it checks a few common field names for each. The lesson's own note says to check one real run's output and adjust the names if ProLevelMax's differ. Since the lesson ran on sample reviews, do this on your first real run.
7. A prompt that allows any reply. Find the themes says: use only what the reviews say, invent no numbers, and reply with JSON only, with top_complaints, top_praise and talking_point. Without that fixed shape, Make the sheet row can't put each answer in its own column. That node strips a code fence if the model adds one, and if it still can't read the JSON, the raw answer lands in Top complaints. If you see a long answer there, tighten the prompt. More on this in structured output.
8. Sheet headers that don't match. Add this week's row uses Append Row with Mapping Column Mode set to Map Automatically, so each field lands in the column with the same name. Row 1 must read Week, Company, Reviews read, Average rating, Top complaints, Top praise, Talking point. The lesson's guide page has a CSV with these headers. Paste your sheet's URL in Document; the template's URL is a placeholder.
One habit, not a setup step: read a few of the reviews behind a summary before you quote it to a client. The lesson's themes came from 8 made-up sample reviews, so treat them as an example of the format only.
Try it free, step by step
n8n + ProLevelMax: Weekly Review Themes From Trustpilot to Google Sheets. Stop reading reviews one by one. Every Friday n8n asks ProLevelMax for a company's newest Trustpilot reviews, AI sums up the top complaints and praise, and one row lands in Google Sheets, for about… You do every step yourself in a practice copy of n8n, it checks your work as you go, and “Do it for me” finishes any step you get stuck on. Free, and nothing touches your real accounts.
Start the free lessonKey takeaways
- Collect is for reading fields. Run charges your balance.
- The key lives in a Bearer Auth credential, never the body.
- invalid key format means only the key is wrong or missing.
- Copy field names from Collect and give the call a 2-minute timeout.
- Ask the AI for fixed JSON keys and match the sheet headers exactly.
Frequently asked questions
- Do I need a new URL for another review site?
No. Every source uses
https://prolevelmax.com/v1/run. Change the scraper id and the input fields.- Why does my row show the whole AI answer in Top complaints?
Make the sheet row couldn't read the AI's reply as JSON, so it kept the raw text there to avoid losing it. Check the prompt still says JSON only.
- Why is Average rating blank?
Keep rating and text only averages reviews that have a rating it can read. If the field name in your real output is different, add it to the code.
- Where are more answers?
The n8n ProLevelMax review digest FAQ covers each node, and the ProLevelMax Reddit n8n mistakes cover the Reddit workflow. The cheat sheets page has a one-page version.
