To automate sourcing into Greenhouse, sourced people should enter as prospects (not candidates) in a prospect pool and stage, carry a source value, and only become candidates when they say yes. You can do that with Greenhouse's own Sourcing Automation add-on, with the Chrome extension and bulk import, or by connecting an outside sourcing tool through the API. Whichever route you pick, the data model, source tracking and GDPR timing decide whether the pipeline stays clean.

This guide is written for talent acquisition teams that already run Greenhouse. It shows the object model, a step-by-step workflow, how to keep source attribution and reporting intact, the GDPR clock that starts when you contact someone, a cost-per-qualified-profile calculation, and the limits of automation. Where we describe Sprad, we only state what is published on the People Search page.

Prospect or candidate: what Greenhouse expects from sourced profiles

Greenhouse separates people you have found from people who have applied. A prospect is a profile in the Greenhouse CRM: it sits in a prospect pool and a stage, but is not in a job's interview pipeline. A candidate is tied to a job. Prospects land in the "Uncategorized Prospects" pool by default, and admins can change that in the organisation settings. Both CRM and sourcing add-ons need Job Admin permissions or higher with CRM access.

That has a practical consequence. If an automation writes every sourced profile straight into a job as a candidate, your job pipeline fills with people who never agreed to anything, your stage-to-stage conversion rates drop, and recruiters start ignoring the list. Prospects first, candidates after a positive reply, is the cleaner design.

Greenhouse describes what happens when a prospect later applies: if auto-merge criteria such as the same email address match, both applications sit under one profile and the prospect application counts as converted (Greenhouse Support, prospect conversion). One trap is documented there too: conversion only happens if the prospect form arrived first, and in that case the merged profile can overwrite the candidate application's Source and Credit To fields.

How do sourced candidates get into Greenhouse?

According to the CRM overview, prospects can be added in these ways (Greenhouse Support, checked 18 October 2026):

• the Greenhouse Recruiting Chrome extension, from a website, a resume or by hand
• a bulk spreadsheet import, including jobs, pools, stages and custom fields
• Maildrop, by emailing resumes to a dedicated address
• converting rejected candidates from closed jobs
• prospect posts on the career page
• LinkedIn Recruiter System Connect and other sourcing integration partners (Greenhouse lists more than 80)
• the API

The API route is the one automation uses. Two interfaces matter, and the documentation changed recently:

• Harvest. The Harvest docs state that Harvest v1/v2 was deprecated with removal on 31 August 2026 and point to v3. In the write endpoint migration guide, the old "add prospect" and "add candidate" calls both map to POST /v3/candidates, with the prospect or candidate distinction in the nested application object. Custom integrations use OAuth client credentials; partners use the authorization code flow. If an older script or tool still calls v1, check it now.
• Candidate Ingestion API. The Candidate Ingestion API is for sourcing partners that share customers with Greenhouse. POST /v1/partner/candidates creates prospects by default (a job ID is only required if the person is created as a candidate), and applicants who use a partner's tracking link get the partner as their source automatically.

Step by step: a sourcing workflow that keeps Greenhouse clean

1. Prepare the structure. Create one prospect pool per hiring theme (for example "Data engineers DACH") with stages such as Found, Contacted, Replied, Qualified. Name an owner for each pool.
2. Define the source. Add a custom source under the sourcing strategy "Prospecting" (Greenhouse has eight strategies, including Agencies, Referral and Third-party Boards) with a name per tool or campaign, for example "Prospecting - Atlas search". Do this before the first import, because source cannot be reconstructed afterwards.
3. Write the brief. Must-haves, nice-to-haves and exclusions in plain words. The better the brief, the fewer profiles you waste.
4. Run the search and review a sample. Check 10 to 20 profiles by hand before any outreach. Fix the brief, not the results.
5. Import as prospects. Pool, stage, owner and source set on creation. Do not set a job yet.
6. Handle the privacy notice. The Article 14 clock starts as soon as you hold the data (see below).
7. Contact and follow up. A positive reply moves the person to "Replied". A decline or silence after your cadence moves them to a closed stage, so nobody is contacted twice by accident.
8. Convert on yes. Only now attach a job so the person becomes a candidate in the interview pipeline.
9. Review monthly. Prospect pool activity and source quality tell you which pools and sources produce interviews.

Source attribution and reporting: what to measure

Greenhouse source tracking has two parts: a sourcing strategy and a source name. The eight strategies are Agencies, Company Marketing, In Person Event, Prospecting, Social Media, Third-party Boards, Referral and Others. Splitting them lets you compare single sources and whole strategies, for example in the Candidate Quality by Source report. Credit To assigns the credit for a candidate to an individual, for example the referrer.

For automated sourcing, track four numbers per pool and source:

• Reply rate: replies divided by prospects contacted.
• Qualification rate: prospects that reach "Qualified" divided by replies.
• Conversion to interview: candidates created from the pool that reach a first interview.
• Cost per qualified profile and per hire (calculation below).

Greenhouse lists prospect pool activity reporting and the Sourcing Automation add-on on its higher plans, so check what your contract includes before you design reports around them.

What does Greenhouse's own Sourcing Automation cover?

The Sourcing Automation add-on creates campaign pools, looks up contact information for identified prospects, sends personalised automated email, and tracks prospect engagement inside Greenhouse. It is available on the Plus and Pro tiers (Core customers buy seats as an add-on), and users need Job Admin rights with CRM access plus a Sourcing Automation seat. Prospects are still added with the Chrome extension, so finding the people remains a manual or tool-assisted step. If your bottleneck is campaign execution for people you already identified, this is the shortest path. If your bottleneck is finding and qualifying them, you add a search tool in front.

Where an outside sourcing tool fits, and what to ask before you connect one

Sprad's People Search page lists Greenhouse among the ATS it works with and says Sprad docks onto the system you already run, so qualified candidates and interview results flow back into it. It also says outreach runs on Sprad's own stack rather than your team's LinkedIn accounts. The page does not publish field-level details. Before a pilot, ask any vendor, Sprad included, these questions in writing:

• Are people created as prospects or as candidates, and in which pool and stage?
• Which source name is set, and can you choose it?
• Which Harvest or Ingestion version is used (see the v3 note above)?
• How are duplicates handled when the email already exists?
• Who sends the Article 14 notice, and when?
• What gets deleted, and when, if a person objects?

GDPR: the one-month clock when you source people who never applied

If you collect a person's data from a public profile or a data provider, you did not get it from them. Article 14 GDPR then requires you to inform that person, including the source of the data. The information must be given within a reasonable period, at the latest within one month (Art. 14(3)(a)), or at the latest at the first contact if you use the data to communicate with them (Art. 14(3)(b)). Article 14(5) has exceptions, for example where informing would take disproportionate effort, but you should not rely on them for individual outreach.

A common legal basis for sourcing is legitimate interest under Article 6(1)(f), which needs a documented balancing test, and the person can object. In practice:

• put the privacy notice, source and objection route into the first message
• set a retention period for prospects who never reply and delete on schedule
• record the source in Greenhouse, because you must be able to name it
• honour an objection by removing the profile and the outreach cadence together

Our article on GDPR in active sourcing goes deeper. This is not legal advice; your data protection officer decides the details.

Cost per qualified profile: a worked example

Sprad's published prices (sprad.io/pricing, as of 6 October 2026, net): web search across more than one billion profiles costs 2 credits per qualified profile, up to 50 per run, and only qualified profiles count. Web sourcing starts with an Atlas Base membership at 450 € per year, which includes 6,000 credits (list price 0.08 € per credit). The flat-rate alternative is Recruiter Lite at 79 € per seat and month, with 500 profiles per month.

Assumptions: cost is membership divided by profiles, credits included in the membership are used up first, and the Base figures stop at 6,000 credits. Outreach, voice interviews and other steps are priced separately on the pricing page. For comparison, assume a recruiter needs 5 minutes to find and check one profile by hand and costs 60 € per hour fully loaded: that is 5 € per profile, a figure you should replace with your own. The point is not that tools are cheap, but that cost per qualified profile is the number to compare, not the seat price. A qualified profile is not a hire, so multiply by your own reply and conversion rates to get cost per interview.

Limits and typical mistakes

• Automation prepares, people decide. A ranked shortlist is input for a recruiter, not a decision about a person. Keep review steps in the workflow.
• Dirty source data. Without a source value set at creation, you cannot report on any of this later.
• Candidates instead of prospects. It inflates job pipelines and distorts conversion metrics.
• Duplicates. Decide in advance how existing profiles are matched, and remember Greenhouse's auto-merge can overwrite source fields.
• Outdated API calls. Anything still on Harvest v1/v2 needs migration.
• Skipped privacy notice. The one-month clock runs whether or not your tool tracks it.
• Quality claims. Sprad's own benchmark (38 % against 9 % for the next best tool in three roles, 15 August 2026) measures overlap with other systems' finds, not hiring outcomes; see the sourcing benchmark for method and limits.

FAQ

Is Greenhouse sourcing automation the same as an external sourcing tool?

No. Greenhouse's add-on handles campaigns, contact lookup and tracking inside Greenhouse. An external tool searches and qualifies, then writes results back through the API. Many teams use one or the other, some both.

Should sourced people be prospects or candidates?

Prospects, in a pool and stage, until they reply positively. Then attach a job.

Which Greenhouse API do I use?

Custom integrations use Harvest v3, since v1/v2 was scheduled for removal on 31 August 2026. Sourcing partners can use the Candidate Ingestion API. Always check the current docs.

How do I keep source reporting accurate?

Set a custom source under "Prospecting" at creation, and watch for auto-merge overwriting source fields when a prospect applies later.

Do I have to inform sourced people under GDPR?

Yes. Article 14 requires information about the controller, purpose and source, within one month at the latest, or at first contact if you write to them.

What does a qualified profile cost with Sprad?

2 credits per qualified profile in web search, with web sourcing starting at the 450 € Atlas Base membership, or 79 € per seat and month for Recruiter Lite. Figures as of 6 October 2026, net.

Does Sprad work with Greenhouse?

Greenhouse is listed on the People Search page among the ATS Sprad works with. Field mapping is not published, so confirm it in a demo.