Upwork proposal templates for web scraping jobs
Updated
Clients hiring for web scraping on Upwork usually screen for risk before they screen for code. They want to know if you can identify the data source, handle JavaScript-rendered pages, avoid fragile selectors, respect access limits, deliver clean output, and explain when an API is better than browser scraping.
Hi {{client_name}}, I saw you need prices tracked for {{store_or_marketplace}} and exported to {{output_format}}. The important part here is not just pulling the price once, but keeping the scraper stable when product pages change.
I can build a scraper that collects {{fields_needed}}, including price, stock status, product URL, SKU, and timestamp if needed. I’ll first check whether the data is available in page HTML, a public API response, or a JavaScript-rendered request. If the site exposes a cleaner data source than browser rendering, I’ll use that to reduce breakage and cost.
For this job I’d suggest:
- A small test on {{sample_count}} products first
- Validation rules such as missing price checks and duplicate URL checks
- Output to {{csv_google_sheet_database}}
- A simple run log showing scraped, skipped, and failed URLs
- Clear notes on any blocked pages, CAPTCHA prompts, or pages that require login
Relevant example: {{portfolio_link_or_short_description}}. It shows {{similar_result}}, including how I handled {{similar_challenge}}.
A few questions before I quote the final scope:
1. How many product URLs are in the first batch?
2. How often should the data refresh: once, daily, weekly, or on demand?
3. Do you need historical price changes stored, or only the latest price?
4. Are all URLs public, or do any require a login?
If that fits, I can review {{sample_urls}} and send a short implementation plan before starting.Use this when the job mentions Amazon-like listings, Shopify stores, competitor monitoring, stock checks, or recurring price reports.
Hi {{client_name}}, I read that you want to extract {{lead_type}} from {{directory_name}} and save the results into {{output_format}}. I can help, but I’d start by confirming the fields and the source structure so the output is useful, not just scraped.
For a directory scrape like this, I’d collect {{fields_needed}} and include source URLs for review. I’ll check whether the listings are public, whether pagination or search filters affect coverage, and whether duplicate businesses appear across categories or locations.
My approach:
- Inspect {{directory_name}} for HTML, API calls, or JavaScript-rendered data
- Run a small sample scrape for {{sample_location_or_category}}
- Normalize fields such as phone, website, address, and category
- Add validation rules for missing email, invalid website, duplicate company, and empty address
- Deliver {{output_format}}, with columns named exactly as you need them
I’ve done similar work for {{similar_directory_or_industry}}, where the main challenge was {{challenge}}, and the final output was {{result_type}}. You can see a sample format here: {{portfolio_or_sample_link}}.
One note: if the directory uses CAPTCHA, login-only pages, or blocks automated access, I’ll flag that early and recommend a safer path, such as lower request volume, using a public export/API if available, or narrowing the scope.
Could you send 2–3 example listing URLs and your required fields? I can then confirm the estimate and timeline.Use this for B2B leads, local directories, marketplace sellers, clinic lists, real estate agents, job boards, or public company databases.
Hi {{client_name}}, I noticed your existing scraper for {{target_site}} is failing after {{recent_change_or_error}}. That usually means the page structure changed, the data moved into a JavaScript request, or the site added a blocking step such as rate limiting or CAPTCHA.
I can debug the scraper and either patch the current code or recommend a cleaner rebuild if the old approach is too fragile. I’ll start by reproducing the failure, checking selectors and network requests, and confirming whether the data is still publicly accessible without login.
What I’d review:
- Current code: {{language_or_framework}}
- Error logs or failed output examples
- Target pages that still work and pages that fail
- Whether the site has a public API or JSON response that can replace browser scraping
- Current output requirements: {{output_format}}
For the fix, I’ll aim to make the scraper easier to maintain by adding clearer selectors, retry rules, error reporting, and a small test set of known URLs. If the issue is anti-bot blocking or CAPTCHA, I’ll explain the limitation instead of pretending it can always be bypassed.
I’ve fixed similar scrapers where {{similar_problem}} caused {{impact}}, and the repair involved {{repair_summary}}. Example or code sample: {{proof_link}}.
Please send the repo or script, one failed run log, and 3–5 URLs that should be scraped. I can then tell you whether this is a quick patch or a rebuild.Use this when the client already has Python, Scrapy, Playwright, Puppeteer, Selenium, Apify, or a custom script that stopped working.
What web scraping clients screen for first
- You noticed the actual source. Mention if the target looks like public HTML, JavaScript-rendered content, an API response, a feed, or login-only data.
- You understand fragility. Clients worry the scraper will work once and break next week. Say how you’ll handle selectors, retries, logs, and validation.
- You can define clean output. Web scraping clients often care more about the spreadsheet or database than the scraper itself. Name the fields, file type, and error columns.
- You ask about limits early. Ask about record count, refresh frequency, target URLs, login requirements, CAPTCHA, and whether the client has the right to use the data in the way they described.
- You can explain API vs browser tradeoffs. If a public API or JSON endpoint exists, it may be more stable than rendering pages in a browser. If the page needs interaction, browser tools may be required.
- You don’t sound reckless. Avoid promising to bypass every block. Say you’ll identify blockers and recommend a practical path.
Use an opening line that proves you read the post
| Weak opening | Better opening |
|---|---|
| I can build a web scraper for you. | I saw you need daily price and stock tracking from {{site_name}} into Google Sheets, with product URLs already provided. |
| I have experience with Python scraping. | Your job mentions JavaScript-rendered listings, so I’d first inspect the network calls before deciding between Scrapy and Playwright. |
| I can scrape any website. | I can review whether the data is public, whether a public API exists, and whether the target pages have CAPTCHA or login limits before confirming the approach. |
| Please check my profile. | I’ve handled similar directory extraction where duplicate companies appeared across locations, so I’d include source URL and dedupe rules in the output. |
Mini risk matrix for web scraping proposals
| Source type | Risk level | Proposal angle |
|---|---|---|
| Public HTML pages | Lower | Mention selectors, pagination, validation, and clean CSV or Sheet delivery. |
| JavaScript-rendered pages | Medium | Say you’ll inspect network requests and use browser rendering only if needed. |
| Public API or JSON endpoint | Lower to medium | Propose using the API-like source first because it is often cleaner than scraping rendered pages. |
| Login required | Higher | Ask whether the client owns the account and has permission to use the data as described. |
| CAPTCHA or heavy blocking | High | Avoid promising a bypass. Offer a review, lower-volume approach, or alternate source. |
What to customize in each proposal
| Field | Selector or source | Validation rule | Output type |
|---|---|---|---|
| Product price | HTML price block or JSON response | Must be numeric and include currency | Decimal/text |
| Stock status | Button text, badge, or product JSON | Map to in stock, out of stock, unknown | Text |
| Business email | Listing page or public website | Flag missing or invalid email format | Text |
| Listing URL | Current page URL | Required for every row | URL |
| Scraped at | Run timestamp | Required for recurring scrapes | Date/time |
- Target site and example URLs. Name the site in the first two lines. If the client included URLs, refer to them.
- Exact fields. Don’t say “data.” Say title, price, SKU, phone, address, rating, listing URL, date posted, or whatever the job asks for.
- Volume. A 200-page one-time scrape is different from a daily tracker across thousands of URLs.
- Refresh schedule. One-time extraction, weekly update, daily monitor, and near-real-time tracking need different designs.
- Output destination. CSV, Google Sheets, Airtable, PostgreSQL, BigQuery, or an admin dashboard all change the scope.
- Failure handling. Mention missing-field reports, skipped URLs, duplicates, run logs, and sample validation.
How to estimate scope for scraping jobs
A simple rule: quote fixed-price only when the target source, fields, volume, and output are clear. Use hourly, a paid discovery task, or a small milestone when the site is unknown, JavaScript-heavy, login-based, or likely to change.
- Small fixed-price job: one public site, clear fields, one-time scrape, low volume, CSV or Google Sheets output.
- Milestone-based job: first milestone for source inspection and sample output, second for full extraction, third for scheduling or monitoring.
- Hourly job: scraper repair, unclear blocking, changing page structure, multiple target sites, login flows, or recurring maintenance.
- Paid sample: useful when the client wants proof before full build. Define the sample size, fields, and whether the sample code is included.
- Recurring scraper: include time for monitoring, selector updates, failed-run review, and output checks. The build is only part of the job.
Proof to include in a web scraping proposal
- A sample output file. Link to a redacted CSV, Google Sheet, or screenshot with columns, timestamps, source URLs, and error fields.
- A short code sample. If you can share code, link to a small repo or gist that shows structure, retries, parsing, and validation. Don’t expose a past client’s private target or credentials.
- Before-and-after repair proof. For scraper fixes, show a failed output and the corrected output, with private data removed.
- Monitoring proof. If the job is recurring, show logs, status reports, or screenshots of run summaries.
- A plain-language explanation. One paragraph about the hard part you solved is often stronger than a long list of tools.
When you link proof, make it easy to scan. Use labels like “ecommerce price tracker sample,” “directory lead output,” or “Playwright repair example.” If the portfolio item is not public, say what you can show under NDA or offer a small sample using the client’s public test URLs.
Mistakes that get web scraping proposals skipped
- Saying you can scrape anything. It sounds careless. Some sources are public and simple. Others have login, CAPTCHA, rate limits, or unclear usage rights.
- Ignoring the target site. A generic Python pitch does not answer whether this specific site is static, dynamic, API-backed, blocked, or paginated.
- Not asking about output. Clients do not want raw HTML. They want usable rows with the right columns.
- Promising perfect accuracy. Better: explain validation rules, sample review, and how you’ll handle missing or changed fields.
- Skipping maintenance. If the client needs daily runs, include logs, alerts, and a plan for selector changes.
- Sending a huge tool list. Scrapy, BeautifulSoup, Playwright, Puppeteer, Selenium, Requests, and pandas are not a proposal by themselves. Connect the tool to the site’s behavior.
- Forgetting application cost. The Connects required to submit a proposal depends on the job post. Upwork lists Connects at $0.15 each as of 2026-10-02, so check the post before spending them.
A short checklist before you send
- Did you name the target site, directory, marketplace, or data source?
- Did your first line mention a detail from the job post?
- Did you state the fields and output format?
- Did you ask about volume, refresh frequency, and example URLs?
- Did you mention public access, public API options, login requirements, or CAPTCHA if relevant?
- Did you include one proof link or a short description of a similar scrape?
- Did you avoid promising to bypass every blocker?
If you apply to many scraping jobs, use a saved base template, then rewrite the first lines and scope section for each post. BidGrow can help score Upwork jobs with visible rules and draft proposals from your own templates. You still review the proposal and press send yourself. BidGrow does not submit proposals, does not log in as you, and is not affiliated with or endorsed by Upwork.
Questions
What should I say in an Upwork proposal for web scraping?
Start with the target site and the data the client asked for. Then explain your likely approach, the output format, validation rules, and any risks such as JavaScript rendering, login, CAPTCHA, or changing selectors.
Should I offer a sample scrape for free?
Not always. If the job is small and the site is public, a tiny sample can help. For unclear or risky targets, propose a paid discovery or sample milestone with a defined number of pages, fields, and output rows.
What if the client hides the target site in the job post?
Don’t guess. Say you can confirm the method after seeing the domain and 2–3 example URLs. Ask whether the data is public, whether login is required, and what output they need.
Should I mention proxies or CAPTCHA in the proposal?
Mention them only if the job suggests blocking, high volume, or protected pages. Keep the wording careful: say you’ll identify blockers and recommend a practical path, not that you can bypass anything.
Is fixed-price or hourly better for web scraping jobs?
Fixed-price works when the source, fields, volume, and output are clear. Hourly or milestone-based work is safer for scraper repair, unknown sites, login flows, recurring monitors, or JavaScript-heavy pages.