Join the Community

Effective Data Sourcing Strategies for German SMEs with Clay

·

·

Bo (., I found this: https://community.clay.com/x/support/mmdb58wfmocw/data-sources-for-clays-find-companies-and-find-peo?utm_source=chatgpt.com Here is the data: https://app.clay.com/workspaces/670253/workbooks/wb_0sz6ugaxRpWDQC9QBfm/tables/t_0szfyi9esn5Ucne2Hm3/views/gv_0szfyi9PsonfgqfE2u6 I made a video to explain where I am at right now: https://www.loom.com/share/9464fac8d23141ecac03fa40cb5022cf Clay S. From what I understand, Clay’s “Find Contacts at Company” enrichment relies on external data providers like Clearbit, Apollo, HitHorizons, etc. We’re working almost exclusively with German SMEs, and I’ve noticed that domains like zuzusauber.de often return “Company Not Found.” Can you confirm whether this is because those domains simply aren’t present in the provider databases — and if so, that our current workflow (scraping Impressum/contact pages + using GPT to extract names, emails, and phone numbers) is indeed the most effective and cost-efficient approach for this type of data? This is a key decision point for us, since it will impact how we scale the process and optimize credit usage. Appreciate your guidance!

8 comments

Clay
APP
·
·
Yes, "Find Contacts at Company" can fail for German SME domains because they're often missing from provider databases. Our Find Companies/Find People datasets combine data from 100+ providers, but smaller companies without LinkedIn presence may not be covered. Your current workflow of scraping Impressum/contact pages with GPT extraction is indeed effective for German SMEs. We specifically recommend using Claygent to pull publicly available contact details from company contact pages, especially for smaller businesses. There's even a template for this exact use case that scrapes websites for email addresses and validates them. For companies without LinkedIn presence, this scraping approach is the recommended method for finding SMB contacts.
Sources: Find SMB contact info & write a compliment | Clay Find Companies [Clay 101] - Clay University Find Companies [Clay 101] - Clay University
Stefano
·
·
Hi, thank you for reaching out. If the domains are not being recognised by our Enrich Company feature, I suggest instead using a Claygent model to scrape the Impressum to try and find the emails and phone numbers from the website. I have seen different clients in the EU use this approach successfully and if the websites are already provided we see this work fairly well. Would you need help setting up a Claygent column that looks for the information on the website's impressum?
Philip H.
·
·
Sure, that would be great! This is my situation so far with the agent: https://www.loom.com/share/b2135c97510542c79fbf340ce087c439 This is the sheet: https://app.clay.com/workspaces/670253/workbooks/wb_0sz6ugaxRpWDQC9QBfm/tables/t_0sz6ugaAdPnxkdmdM5g/views/gv_0sz6ugahV3D6yAkmBS3
Stefano
·
·
Hi, unfortunately as I do not speak German I cannot really modify the existing prompt. However, I recommend trying to write the prompt in English as a lot of these LLM models are actually trained on the English language meaning this might result in a more accurate response. I believe any prompt that guides our Claygent/ OpenAI to the website and clarifies that you are looking for both an email and a phone number that should be in the "Impressum" section of your website should help you get the desired result. I do think using a more advanced form of Claygent (like Neon or Argon) will work more effectively as sometimes the impressums are harder to find on websites, however through trial and error on a small set of rows you should be able to find what you are looking for.
Philip H.
·
·
Ok, I see. Is there some way to inform myself about how to make better prompts with Clay? I already use S.P.I.C.E., but I do not know the different forms of Claygent that you mentioned. I read this: https://www.clay.com/university/lesson/prompt-engineering-crash-course-clay-101 And the rest of the Clay 101 University
Stefano
·
·
Hi, for prompt optimisation I recommend trying to approaches: 1) Utilize the "Help me" AI feature. This option can assist in crafting a more refined and effective prompt, which is likely to produce better results. The AI will analyze the context of your query and suggest wording or structures that are optimized for the system, helping to guide you toward more accurate and relevant answers. The button is shown below:
(33 kB)
2) Consider using an external tool like ChatGPT. You can explain to ChatGPT what you're trying to accomplish, and it can help by offering suggestions on how to improve your prompt or rephrase it in a way that is more likely to yield the best results. This method can be especially useful for understanding how to communicate your request more clearly or efficiently. Let me know if this helps!
Philip H.
·
·
Thanks for checking in! I will try this out and get back to you!
Channeled
APP
·
·

Effective Data Sourcing Strategies for German SMEs with Clay | Clay