The most expensive mistake in data collection is not picking an expensive source. It is picking a source designed for a different job — using an enrichment tool for discovery, or a discovery tool for enrichment.
The price difference between those mistakes and the right choice is routinely thirty times. This post is the decision tree.
Start from what you already have
Every workflow begins in one of four states, and the state determines the source.
**You have a description of a person.** A role, a seniority, an industry, a company size. This is the most common starting point and the cheapest to serve: **Leads Finder** at one credit per result turns a specification into contacts with verified emails.
**You have a place and a category.** "Every dental practice within 40km." A contact database will serve you badly here because local businesses are poorly represented in firmographic data. **Google Maps** at ten credits per result is the right tool, or **Maps Email** at fifteen if you need email addresses rather than phone numbers.
**You have a list of URLs or handles.** Profile URLs, Instagram handles, company pages. You are enriching, not discovering: **LinkedIn Profiles**, **Instagram Email**, or **LinkedIn Company** depending on the identifier type.
**You have a topic but no people.** You know the problem you solve, not who currently has it. This is intent territory: **Reddit Pro** at three credits per result, or **Twitter** at one.
Getting this first classification right eliminates most waste before you spend anything.
Discovery versus enrichment
The distinction that costs teams the most money.
**Discovery sources** answer "find me people matching this description." Leads Finder, Google Maps, Reddit Pro, Twitter, YC Directory, Google Search.
**Enrichment sources** answer "tell me more about these specific entities." LinkedIn Profiles, LinkedIn Phones, Instagram Email, Social Email.
Enrichment sources are more expensive per result because each one involves visiting and parsing a page rather than querying a structured index. That is entirely reasonable for enrichment and completely wrong for discovery.
The concrete failure: using LinkedIn Profiles to build a prospect list. At thirty credits per result, a 200-person list costs 6,000 credits. The same list from Leads Finder costs 200 credits and arrives with verified emails already attached. The correct sequence is Leads Finder to discover, then LinkedIn Profiles only on the shortlist that survives qualification.
The LinkedIn family, disambiguated
Five sources on one platform causes more confusion than anything else, so:
- **LinkedIn Profiles** (30 cr) — you have URLs, you want the person's detail. Enrichment.
- **LinkedIn Comments** (15 cr) — you have a post, you want everyone who engaged with it. Discovery, and the highest-intent discovery on the platform.
- **LinkedIn Posts** (25 cr, Pro) — you have a keyword, you want the conversation and who is driving it. Research, and the input step for Comments.
- **LinkedIn Company** (20 cr, Pro) — you have companies, you want organisational detail. Account mapping.
- **LinkedIn Phones** (8 cr) — you have profiles, you want a second channel. Run last, on a shortlist.
The pattern most teams eventually settle on: Posts to find the conversation, Comments to get the engaged people, Profiles to qualify them, Phones on whoever survives.
When two sources look interchangeable
**Reddit Pro (3 cr) versus Reddit Lite (10 cr).** Counter-intuitively, Pro costs less per result and does more — multiple subreddits, community discovery, comment threads. Lite is only cheaper because you collect fewer. Use Pro for anything serious.
**Google Maps (10 cr) versus Maps Email (15 cr).** Same underlying collection. Maps gives you phone numbers from the listing. Maps Email additionally crawls each business website for published addresses, which is slower and costs more. Choose by channel: calling or emailing.
**Instagram Email (12 cr, Pro) versus Social Email (18 cr, Pro).** Instagram Email has a much better hit rate on Instagram accounts. Social Email covers several platforms with a lower per-profile hit rate. For mixed lists, run the focused one first and the broad one on the remainder.
**YC Directory (10 cr) versus YC Scraper (55 cr).** Directory for filtering the universe, Scraper for depth on the survivors. Running Scraper across a whole batch is almost never the right economic call.
Reddit or Twitter for intent?
Both, usually, because they surface genuinely different people with less overlap than you would expect.
**Twitter** is faster and cheaper at one credit. People complain there first and in real time, which makes it right for launch reactions, competitor outages, and anything where speed matters. The limitation is depth: a tweet tells you someone is frustrated, rarely why or what they have already tried.
**Reddit** is where reasoning lives. Someone will write three paragraphs explaining what they need, what they rejected, and why. For understanding a market or building genuinely warm outreach, it is substantially better material.
Rule of thumb: Twitter for reacting, Reddit for understanding.
Sizing the run
Two habits prevent most billing surprises.
**Narrow before you widen.** Because billing is per result returned, a tight filter costs less *and* produces a better list. The instinct to cast wide and clean up later is wrong on both axes.
**Watch for multiplied counts.** Google Search returns pages times results-per-page. Reddit with comments enabled returns posts times their threads. Facebook Ad Leads bills on ads processed, and one advertiser runs many ads, so final lead count is a fraction of the ad count. These three are where runs come back larger than expected — run a small test first to learn the ratio.
The cheapest thing you can do
Score before you enrich. AI Lead Scoring at two credits per row typically removes half a list, and running it before any thirty-credit enrichment step means the expensive operations only touch contacts you have already decided are worth it.
Most teams that feel their credit balance is too small are not running too many searches. They are running expensive operations on unqualified lists, in the wrong order.
