Hi support, whenever I try to run a longer prompt using HTTP API (Deepseek) from the Clay, I get a JSON parsing error. Oddly enough, when I try the same prompt using the classic API request, it works fine without any errors. Is there a way to avoid this error, or am I doing something wrong?
{ “model”: “deepseek-chat”, “messages”: [ { “role”: “system”, “content”: “You are a B2B sales qualification expert specializing in cybersecurity prospect analysis. Return analysis in the specified systematic format.” }, { “role”: “user”, “content”: “# Ethiack B2B Sales Qualification - Systematic Multi-Signal Analysis # INSTRUCTIONS MISSION: Achieve maximum classification accuracy through multi-signal analysis and systematic edge case detection. Use weighted scoring across multiple data sources to determine TRUE business model and appropriate tier assignment. CORE PRINCIPLE: Distinguish between companies that BUILD technology vs companies that USE technology vs companies that OPERATE traditional businesses. # DATA # ESSENTIAL FIELDS FOR QUALIFICATION You will receive prospect data with these 5 essential fields: ## Required Fields - Company name for business model signals companyName - LinkedIn industry classification (baseline): industry i - AI-enhanced industry classification GPT (Homepage) Industry - Geographic qualification requirement: companyLocation - Primary website research target content ## Backup Research Fields (when available) - **Primary Website (Clearbit)** - Alternative domain source - **Primary Website (DeepSeek)** - Alternative domain source ## Processing Protocol 1. Use `Final Domain` for website research 2. If `Final Domain` empty, use backup domains 3. `GPT (Homepage) Industry` overrides `industry` field 4. Geographic qualification from `companyLocation` required # SYSTEMATIC SIGNAL ANALYSIS FRAMEWORK ## TECHNOLOGY PROVIDER SIGNALS (SaaS/Software - Tier 3 Indicators) **Strong Signals (+3 points each):** - Company name contains: \“Software\“, \“Tech\“, \“SaaS\“, \“Platform\“, \“Solutions\“, \“Systems\” - Homepage describes: \“API\“, \“Integration\“, \“Dashboard\“, \“Analytics Platform\” - Product pages show: Software demos, pricing tiers, \“Free trial\” - Customer testimonials from OTHER companies in various industries **Medium Signals (+2 points each):** - About page: \“We help [industry] companies\“, \“Built for [professionals]\” - Careers page: Software engineers, DevOps, Product managers - AI Industry classification overrides traditional LinkedIn industry **Weak Signals (+1 point each):** - Modern website design with SaaS layout patterns - Blog content about technology trends - Integration marketplace or partner ecosystem ## TRADITIONAL OPERATOR SIGNALS (Tier 1/2 or Anti-ICP Indicators) **Strong Signals (-3 points from tech score):** - Physical store locations, branch finder - Direct consumer services (bank accounts, insurance policies) - Manufacturing facilities, production capabilities - Government/public sector entity (.gov domain, public funding) **Medium Signals (-2 points from tech score):** - Traditional industry website layout - Focus on end-consumer rather than B2B - Regulatory compliance emphasis over technology ## HYBRID SIGNALS (Combined Business Model) - Company operates traditional business AND provides technology - Example: Bank that also sells banking software - Classification: Higher tech score wins # GEOGRAPHIC QUALIFICATION ## TARGET REGIONS (Required for all tiers) - **IBERIA:** Spain, Portugal - **NORDICS:** Sweden, Denmark, Norway, Finland - **BENELUX:** Netherlands, Belgium, Luxembourg - **UK:** United Kingdom, England, Scotland, Wales # SYSTEMATIC TIER CLASSIFICATION ## TIER 3 - TECHNOLOGY PROVIDERS (Priority Tier) **Automatic Qualification Criteria:** - Technology Signal Score ≥ 6 points - Located in target geography - NOT in Anti-ICP industries **Business Models:** - SaaS companies (ANY vertical focus) - Software development companies - Technology platforms and APIs - B2B technology service providers **Industry Overrides (Always Tier 3 if signals match):** - \“Banking\” company that builds fintech → Tier 3 - \“Healthcare\” company that builds healthtech → Tier 3 - \“Retail\” company that builds e-commerce platforms → Tier 3 - \“Automotive\” company that builds auto-tech → Tier 3 ## TIER 1 - RETAIL & E-COMMERCE OPERATORS **Qualification Criteria:** - Traditional Operator Score ≥ 4 points - Technology Signal Score < 6 points - Located in target geography **Target Industries (Traditional Operators Only):** - Retail, Consumer Goods, E-commerce Marketplaces - Fashion, Luxury Goods, Consumer Services - Only if they OPERATE retail, not BUILD retail tech ## TIER 2 - TELECOM & COMMUNICATIONS OPERATORS **Qualification Criteria:** - Traditional Operator Score ≥ 4 points - Technology Signal Score < 6 points - Located in target geography **Target Industries (Traditional Operators Only):** - Telecommunications, Media, Broadcasting - Communications Equipment Manufacturing - Only if they OPERATE telecom, not BUILD telecom tech ## ANTI-ICP (Automatic Disqualification) - Government, Military, Law Enforcement, Political - Non-profit, Educational, Religious institutions - Traditional industries: Farming, Mining, Construction - Any location outside target geography - Technology Signal Score < 3 AND Traditional Score < 3 (unclear business) # SYSTEMATIC PROCESSING WORKFLOW ## STEP 1: Initial Data Analysis (5 seconds) - Extract all available field data - Check for obvious technology indicators in company name - Verify geographic qualification ## STEP 2: Multi-Source Signal Collection (20 seconds) - **Website Research:** Homepage + About + Products pages - **Technology Signals:** Score SaaS/software indicators - **Traditional Business Signals:** Score operational business indicators - **LinkedIn Analysis:** Company description and employee roles ## STEP 3: Systematic Scoring - Calculate Technology Provider Score (0-15 points) - Calculate Traditional Operator Score (0-15 points) - Apply industry-specific modifiers - Check for Anti-ICP automatic disqualifiers ## STEP 4: Classification Decision Tree 1. IF Anti-ICP industry → ANTI-ICP 2. IF Technology Score ≥ 6 → TIER 3 3. IF Traditional Score ≥ 4 AND Retail signals → TIER 1 4. IF Traditional Score ≥ 4 AND Telecom signals → TIER 2 5. ELSE → MAYBE (insufficient signals) ## STEP 5: Confidence Assessment - **HIGH:** Clear signals, no conflicts (Score difference ≥ 4) - **MEDIUM:** Some ambiguity (Score difference 2-3) - **LOW:** Conflicting signals (Score difference ≤ 1) # SYSTEMATIC OUTPUT FORMAT **COMPANY:** [companyName] **DOMAIN:** [Final Domain] **LOCATION:** [companyLocation] → [REGION] =============================================== **MULTI-SIGNAL ANALYSIS:** Website Research: [Key findings from homepage/about/products] Business Model Evidence: [Technology Provider / Traditional Operator / Hybrid] Target Market: [Who they sell to vs what they operate as] LinkedIn Signals: [Industry vs employee roles vs description] **SYSTEMATIC SCORING:** Technology Provider Score: [X]/15 points Traditional Operator Score: [X]/15 points Geographic Match: [YES/NO - specific region] Anti-ICP Flags: [None / Specific disqualifiers] **SIGNAL BREAKDOWN:** Strong Positive: [List key indicators found] Weak Positive: [List supporting indicators] Negative Indicators: [List contrary evidence] **CLASSIFICATION RESULT:** **ICP FIT:** [YES/NO/MAYBE] **TIER:** [1/2/3/ANTI-ICP] **CONFIDENCE:** [HIGH/MEDIUM/LOW] **ACCURACY SCORE:** [0-10]/10 **SYSTEMATIC REASONING:** [One detailed sentence explaining why this classification is most accurate based on signal analysis] **STRATEGIC RECOMMENDATION:** [Specific next step with tier-appropriate messaging and confidence qualifier] # PROCESSING RULES 1. **Always check both LinkedIn and AI industry fields** - use either for tier matching 2. **Technology signals take precedence** - any clear SaaS/software indicators override industry labels 3. **Geographic matching is required** - must be in UK/Iberia/Nordics/Benelux 4. **Scoring threshold enforcement** - Technology Score ≥ 6 automatically triggers Tier 3 evaluation 5. **Anti-ICP overrides everything** - excluded industries cannot be saved by technology signals 6. **Confidence drives recommendation** - LOW confidence requires human review 7. **Multiple signal validation** - minimum 3 data points required for HIGH confidence classification } ], “stream”: false }
Did that and same error.
I can only if you can do a “rollback” since I’ve deleted the HTTP enrichment since it didn’t work.
Would you like me to add a new column and to the same setup?
Hey Alex! The parsing error was because of the smart quotes in your prompt. I could also see some error around the JSON content is being truncated (cut off).
The error shows:
## Backup Research Fields (when avail}It should say "when available)" but it's cut off at "when avail". This creates invalid JSON because the content string isn't properly closed. I edited your prompt and changed the smart quotes and truncated content. This seems to have fixed the issue! :)
Oh cool! It’s working now. You’ve changed my prompt from templates or? Could you please paste here the new prompt or changes?
🔥 Thanks!!
