sending-platforms

Sending Platforms

Overview

This skill picks the sequencer and turns a ScraperCity list into a running campaign without breaking anything on the way in. The deliverable is a platform plan: the requirements, the platform decision with dated vendor facts, the workspace and campaign settings, a column map from every ScraperCity source column to a platform variable with a fallback, the import procedure with dedupe and suppression, the webhook map back to the CRM, a test plan and the monthly cost.

The mistakes it prevents: buying a plan whose contact or send limit the first month (or, for stored-contact caps, month 5) already exceeds; importing with the block list or unsubscribe list overridden; one agency client's events or opt-outs landing in another client's CRM or campaigns; merge tags that render as {{firstName}} or "Hi ,"; the same prospect sitting in two campaigns; an unsubscribe in one tool that another tool keeps mailing; and feature claims copied from review sites instead of the vendor's page.

When to Use

• "Which cold email tool should we use?" or "Instantly vs Smartlead vs Lemlist".
• "How do I upload this ScraperCity CSV to Instantly?" or "map these columns to custom variables".
• "Leads are not uploading" or "why were 300 rows skipped?"
• "How do I stop duplicates across campaigns?"
• "Sync replies and unsubscribes to HubSpot" or "set up webhooks".
• "Can we just use HubSpot sequences or Apollo?"
• Moving from one sequencer to another.
• "We are an agency with 6 clients, what will this cost?" or "each client has its own HubSpot".
• "What settings should this Smartlead campaign use?", "which Instantly API call adds leads in bulk?", "why is the Smartlead block list not blocking?" or "our webhook stopped firing".

Use a different skill for nearby jobs:

Prerequisites

Ask these five first. When the user is not there to answer, write each one as a dated assumption in the plan's Assumptions and Open Questions section and carry on.

Then collect:

Authoritative Foundations

Workflow

1. Size the need

sends_per_month    = new_prospects_per_day x steps x sending_days
contacts_per_month = new_prospects_per_day x sending_days
mailboxes_needed   = ceiling(daily_sends / cold_cap_per_mailbox)   (1,000 / 30 = 34; per client 250 / 30 = 9)
stored_contacts(m) = contacts_per_month x m                        (Smartlead counts every stored prospect once)

Compare mailboxes_needed with the inbox-setup roster and write every number next to each plan's limits. A plan must cover month 2 as well as month 1, and a stored-contact cap must be checked against the cumulative total: 7,000 new contacts a month passes Smartlead Pro's 30,000 in month 5 (ask Smartlead whether client workspaces share that allowance).

2. Choose the platform

Run the decision matrix in references/platform-comparison.md section 3. Record for each option the plan, price, why it fits, its gaps, and the date you checked the vendor page. Never cite a feature from a review site. If a pricing page does not state a feature, write "Not stated" and confirm in the app or with the vendor before buying.

When prose compares two prices, compute higher minus lower and name the dearer plan, with the billing period. On monthly billing Instantly Hypergrowth ($97) costs $3 more than Smartlead Pro ($94); on yearly billing compare annual totals: Hypergrowth $931.20 is $8.40 a year cheaper than Pro $939.60 (the pricing page shows Pro as $78 a month). State the gap once, in the "Decision:" line, and end with a "Revisit if" trigger (a cap, a month, a client count).

Agencies: cost every client's workspace. Smartlead sells client workspaces from Pro upward at $29 a month each, and Unlimited Prime ($379) includes 3. Instantly needs a separate paid subscription for each workspace, and only Hypergrowth and above can invite staff or send webhooks. Lemlist has no client workspaces, so clients in one account share one block list and one webhook set. Link unsubscribes do not separate the two: lemlist's are account-wide and Smartlead's cover all future campaigns. With N clients:

Smartlead Pro:  94 + 29 x N               (N = 5: 94 + 145 = $239 a month)
Instantly:      N x workspace plan        (N = 5 on Hypergrowth: 5 x 97 = $485 a month)

Add the agency's own workspace if it prospects for itself. Full formula, Prime and Growth options, and the agency isolation rules: references/platform-comparison.md section 4.

3. Configure the workspace

Connect mailboxes by OAuth. Then set, per campaign, with the screen path and API field for each platform from references/platform-runbooks.md section 1:

4. Set up each client (agencies only)

Run references/platform-runbooks.md section 5 once per client before its first send: one Instantly workspace or one Smartlead client per client; client_id on every Smartlead campaign and mailbox; no mailbox shared between clients; the client's suppression loaded into its own scope; one webhook per client (Smartlead client-level, never a user-level webhook for the same events, because the user-level one overrides it); that client's HubSpot token in the credential store; a client login limited to its own campaigns. Keep one master list of contacted emails across clients: platform duplicate checks never see another client's campaigns.

5. Build the column map

For each source, map source column, normalized column, platform variable and fallback, using references/csv-mapping-and-import.md sections 1 to 3. Rename headers in the file to the platform's names before upload. Fill fallbacks in the file.

6. Import

Follow the 10-step procedure in references/csv-mapping-and-import.md section 4: normalize, dedupe, suppress, validate, split, rename, upload with duplicate checks on and every block list honored, reconcile counts, preview renders including fallback rows, enroll into a paused campaign.

7. Load suppression

Before the first import, load the company suppression list into the platform's block list: unsubscribes, bounces, complaints, customer and open-deal domains, competitors.

Agency accounts keep three scopes (references/platform-comparison.md section 4b):

8. Wire webhooks to the CRM

Subscribe to replied, interested, meeting request, unsubscribed, bounced and account error. Map each to a CRM action and a sequencer action using references/webhooks-and-crm-sync.md sections 1 to 4; creation calls per platform are in references/platform-runbooks.md. Receivers answer 200 at once, verify the signature, dedupe on the delivery id, and a daily job checks that no webhook was disabled after failures.

Pipeline stages follow section 2a of that file: a "Meeting Request" or "meeting booked" label is a reply, so it leaves the deal at Positive reply; Meeting booked comes from the calendar booking; Meeting held and No-show come from the meeting outcome; the client sits on the deal. Name one owner per client for Interested replies and measure reply-to-answer time; name campaigns {ClientCode}_{Segment}_{ListSource}_{YYYYMM} so the flow can write segment, list source and step onto the contact; leave created contacts non-marketing.

HubSpot writes that hold up: upsert contacts by email in batches of 100; log a reply as an incoming email engagement associated to the contact; opt a contact out with the communication preferences unsubscribe-all endpoint (writing a property is not HubSpot's documented route); record bounces in a custom cold_email_status property.

Decide the writer for each event before wiring anything:

Both writing the same event produces duplicate notes and a second deal. Test with one reply and one Interested categorisation: HubSpot must show one Email activity and one deal (references/webhooks-and-crm-sync.md section 3a).

One HubSpot portal per client. Create one private app per client portal and store each token under the client's name; the receiver picks the token from the client on the event and sends unknown clients to an error queue. Optionally mirror a thin record (company, client, meeting date, stage) to the agency's portal for billing (section 3b).

9. Test, then start

Send to seeds, preview renders with fallbacks, fire a test reply and a test unsubscribe, and confirm the CRM and every block list update once. After launch, reconcile the sequencer's Interested and Meeting Request leads against HubSpot deals every week. Agencies also prove that a test reply lands only in its own client's portal and that a client-scoped block skips only that client. Every import and test step gets an owner and a due date. Then start the campaign at the ramp row's cap.

10. Migrate from an old tool (only when the user named one)

Run this step only when the user named the tool they send from today. Otherwise the deliverable has no Migration Plan section. With a named tool, follow references/migration-runbook.md. Before committing, run the pre-commitment trial: a test domain with 2 to 3 mailboxes (Berman's inboxes per domain) connected only to the new tool, at least 2 weeks of warmup (Instantly's readiness rule), two seed placement tests, a small test import and webhook tests. Then, in order:

Salesforge, Reply.io, Outreach, Salesloft, Woodpecker, Mailshake and Close are CSV import paths, not ScraperCity integrations (references/platform-comparison.md section 1d).

ScraperCity Implementation

These calls are for the agent. In the plan, write each as a step a person runs: the dashboard at app.scrapercity.com or the REST call (GET /api/v1/database/leads with excludeDelivered=true, POST /api/v1/scrape/email-validator), never a tool name.

Wallet and count first. check_wallet(), then count every source before pulling.

Store Leads ($0.0039 per lead), counted with a 100-row test and paged with offset:

{"tool": "scrape_store_leads", "platform": "shopify", "category": "apparel", "apps": "Klaviyo",
 "countryCode": "US", "emails": true, "minEstimatedMonthlyVisits": 20000,
 "maxEstimatedMonthlyVisits": 200000, "totalLeads": 100, "offset": 0}

Lead Database (Growth plan, no per-row charge) for named people at those stores:

{"tool": "query_lead_database", "companyDomain": "northpaw.example, lumentea.example, ...",
 "title": "Founder, Co-Founder, CEO, Owner, Head of Ecommerce, Director of Ecommerce",
 "notTitle": "Assistant, Intern", "hasEmail": true, "excludeDelivered": true, "limit": 100, "after": "0"}

Local Business Database and Maps feed local campaigns with the columns in references/csv-mapping-and-import.md sections 1b and 1c.

Validate every email before import: validate_emails(emails=[...]) at $0.00169 per email; keep email_result, email_quality and sc_mx_provider.

Push by CSV upload, or by API for daily feeds (endpoints, batch sizes, dedupe flags and rate limits per platform in references/csv-mapping-and-import.md section 4). Instantly example, up to 1,000 leads a call:

POST https://api.instantly.ai/api/v2/leads/add
Authorization: Bearer $INSTANTLY_API_KEY          (scope leads:create)
{"campaign_id": "<campaign uuid>", "skip_if_in_workspace": true, "verify_leads_on_import": false,
 "leads": [{"email": "[email protected]", "first_name": "Dana", "company_name": "Northpaw",
            "website": "northpaw.example",
            "custom_variables": {"AppTrigger": "Klaviyo", "Category": "apparel"}}]}

Reconcile leads_uploaded and the skip counters (in_blocklist, skipped_count, invalid_email_count) against the rows sent before the next batch.

Apollo-sourced lists (scrape_apollo, $0.0039 per lead) usually take 15 to 30 minutes: register a ScraperCity webhook and let the pipeline in references/webhooks-and-crm-sync.md section 5 import into a paused campaign.

Worked cost estimate (Shopify apparel stores using Klaviyo, first month):

Growth at $47 is ruled out: 1,587 contacts to upload is over its 1,000-contact limit, and webhooks need Hypergrowth. The paid ScraperCity lines are $8.68; show them and the plan costs before pulling. The full monthly cost always carries mailboxes, domains, automation and verification next to the sequencer: here the sequencer is $97 of $359.17. Seat price from Google's Workspace billing help ($8.40 Flexible, $7.00 Annual), domain price from porkbun.com, both checked 2026-10-13.

Output Format

One Markdown plan following templates/output-template.md (model it on examples/worked-example.md for one team, examples/agency-example.md for an agency). Sections in this order and no others: Summary, Assumptions and Open Questions, Requirements, Platform Decision, Workspace Settings, Column Map, Import Procedure, Suppression and Dedupe, Webhooks to CRM, Test Plan, Migration Plan (only when the user named a current tool), Cost.

• Answer first. Summary is at most 110 words: platform and plan, its monthly price, the all-in total and the one reason it wins.
• Each fact once. The price gap to the runner-up appears only in the "Decision:" line; a cap step (month, next plan, new total) only in Cost, with "Revisit if:" naming the trigger, not the figures. No "Why X over Y" paragraph. No dollar figure with cents appears more than twice.
• One line under the title says what is assumed; the Assumptions table answers the five up-front questions, each marked answered or assumed. No "assumed until run" notes elsewhere.
• Every step is one a person runs: the vendor's screen path or a REST call. ScraperCity steps name the dashboard or the REST endpoint, never an agent tool name or a name() call.
• Workspace Settings gives the value to type or the API field, never "set in the UI" without a value. Source cells name a vendor page or an operator, never a mailbox-provider rule.
• Every vendor behavior the plan relies on (retry schedule, unsubscribe scope, seat limit) names its vendor page; if the page is silent, write "not stated on the pricing page; confirm in the trial".
• Start from what the user has. When they already send, existing mailboxes and domains are reconnected with no purchase: label them existing in Cost (their running cost stays in the monthly total) and buy only additions. With no named tool, Import Procedure carries three switch steps: the old tool's unsubscribes and bounces loaded first, mid-sequence prospects finish on the old tool, warmup on in one tool at a time.
• Cost Total = sequencer, client workspaces, mailboxes, domains and automation, and it equals the sum of its lines. Data and validation join the total only when the request includes building the list; otherwise list them under the table as optional.
• Webhooks to CRM: Meeting booked comes from the calendar booking, never a reply label; Meeting held and No-show come from the meeting outcome; the client sits on the deal. A "Positive replies:" line names who answers Interested replies and how reply-to-answer time is measured. Test Plan includes the weekly reconciliation.
• Import Procedure and Test Plan rows carry an Owner and a Due date (YYYY-MM-DD). Agencies get one row per client wherever the answer differs by client.
• Migration Plan has a dated row for each runbook step; with no named current tool, leave it out.
• No quality-check, framework or self-grading section, no vendor pitch, and no internal file paths, reference names or skill names: write the instruction itself ("warm each mailbox for at least 2 weeks"), not a pointer.

Quality Check

• Summary of 110 words or fewer; no section outside the list above.
• Every Platform Decision row has a check date; the chosen plan covers month 2, and a stored-contact cap shows the month it is reached.
• The price gap appears once, in the Decision line, with the dearer plan and billing period right (Hypergrowth $97 is $3 more than Pro $94 monthly; $8.40 a year cheaper yearly).
• Daily limit per mailbox under 50 (Berman); bounce auto-pause has a typed value; open tracking off; unsubscribe header on.
• Column Map has an email row and a fallback on every row; no import setting overrides a block, unsubscribe or bounce list.
• Webhooks cover reply, unsubscribe and bounce with one writer each; opt-outs through unsubscribe-all; replies as incoming email engagements.
• Meeting booked from the calendar; Meeting held and No-show stages; a "Positive replies:" owner; weekly reconciliation in Test Plan.
• Every Import Procedure and Test Plan row has an owner and a YYYY-MM-DD due date; Platform Decision ends with "Revisit if:".
• No agent tool names, name() calls, internal paths or skill names; every step runs in a vendor UI or by REST.
• Existing mailboxes labeled existing and reconnected, with no purchase; Cost Total equals its lines; data and validation outside the total unless the request includes the list.
• Agency plans: workspaces x clients in Cost; per-client block lists; bounces and complaints global; Client property is Multiple checkboxes; one webhook per client and no Smartlead user-level webhook for the same events; a cross-client contacted list.
• Migration Plan only for a named tool, with a dated trial and seed test, the same mailboxes reconnected and mid-sequence prospects finishing on the old tool.
• Any platform outside Instantly, Smartlead, Lemlist, EmailBison and ManyReach is a CSV import path, never a ScraperCity integration.
• python3 scripts/check-output.py plan.md prints PASS.

Common Pitfalls

Execution Artifacts

Related Skills

• domain-infrastructure and inbox-setup supply the mailboxes this skill connects.
• email-deliverability owns bounce and complaint monitoring once campaigns run.
• cold-email-list builds the validated file this skill imports.
• cold-email-copywriting writes copy against the variable names in the column map.
• reply-handling works the replies the webhooks route.
• scrapercity-integrations covers ScraperCity's own push connectors.
• crm-enricher fills CRM records the webhooks create.