# Describe Manually Mode

> Source: https://docs.trailspark.ai/docs/icp-manual-mode

## How Manual Mode Works

Describe Manually lets you create an ICP entirely from free-text descriptions based on your team's knowledge. No historical data, CSV uploads, or CRM integration required. TrailSpark processes your natural language descriptions into a structured scoring model.

This is the fastest way to start and works well for early-stage companies, teams entering new markets, or anyone who prefers a qualitative approach.

## Prerequisites

- **Owner, Admin, or Editor** role
- Available model capacity on your plan

## Wizard Steps

Manual mode has four steps (no Opportunities step since there is no data upload):

| Step | Required |
|------|----------|
| **Accounts** | Yes |
| **Contacts** | Optional (can skip) |
| **Signals** | Optional (can skip) |
| **Review** | Yes (unlocks after Accounts) |

The Review step unlocks after Accounts is completed.

<img src="/api/images/872ef1ab-b980-4ada-83fe-4599e6599d90/file" alt="Trailspark ICP Wizard Manual Mode" width="600" />

## Step 1: Accounts

Describe your ideal customer companies and account-level detractors in two text areas.

**Ideal customers** -- Be specific about industries, company size (employee count, revenue), geography, technology stack, business model, and growth stage. The more dimensional your description, the better the resulting model.

Example:
```
Mid-market B2B SaaS companies in North America with 200-1000 employees.
Use Salesforce as their CRM. Industries: technology, professional services,
healthcare technology. Have established sales processes but need better
lead qualification.
```

**Account detractors** -- Describe companies that are not a good fit: too small, wrong industry, lacking required infrastructure, etc.

## Step 2: Contacts (Optional)

Describe the key personas involved in purchasing decisions and contact-level detractors.

Cover job titles, seniority, departments, responsibilities, and buying committee composition. For detractors, describe contacts without budget authority or from irrelevant departments.

If skipped, the ICP focuses on account-level characteristics only.

## Step 3: Signals (Optional)

Describe positive engagement signals indicating buying intent and negative signals indicating poor fit.

If skipped, the model relies on firmographic and persona matching only.

## Step 4: Review & Create

1. Review the status of all steps and your descriptions
2. Click **Begin Analysis & Preview** -- TrailSpark processes your descriptions into structured ICP criteria and generates a model preview
3. Review the model preview, optionally add notes to refine it, and name the model
4. Click **Create Model** to save and activate the model


## Writing Effective Descriptions

**Be specific, not generic.** "Companies with 200-1000 employees and $10M-$100M annual revenue" is far more useful than "medium-sized companies."

**Indicate priority.** "Must be in technology or healthcare" vs. "Ideally 200+ employees, but 100+ acceptable" vs. "Prefer companies using Salesforce" -- these distinctions help the model weight criteria correctly.

**Cover multiple dimensions.** Each description should address size, industry, geography, technology, and business model where relevant.

> [!TIP]
> Manual mode is a good starting point that you can refine later. As you accumulate data, create new ICP versions using SparkSense Agent or Upload & Analyze to validate or replace your manually-defined model.

## Next Steps

- [SparkSense Agent Mode](/docs/icp-sparksense-mode) -- AI-powered CRM analysis
- [Upload & Analyze Mode](/docs/icp-upload-mode) -- Data-driven CSV approach