The expensive mistakes with a new traffic source almost never happen at scale. They happen in the first few days, when you point real budget at a source you haven't vetted and only notice the problem after the spend. A short checklist up front is the cheapest optimization you'll ever run.

Know what you're actually buying

"Traffic" is not one thing. Before you fund anything, get clear answers on where the inventory comes from - the formats, the top sources or zones, and how much of it is mainstream versus the vertical you want. A source that's 90% the wrong geo or the wrong device is a bad buy at any price. Ask, and treat vagueness as an answer.

Make sure you can measure it

You cannot optimize what you cannot see. Before the first dollar goes out:

• Confirm your tracker fires correctly on this source, with postbacks or conversion pixels tested end to end.
• Make sure you can break performance down by placement, zone, geo and device, so you can cut the losers.
• Set a hard test budget and a stop-loss, and decide in advance what "working" looks like.

If the source won't give you placement-level reporting, you're flying blind, and blind spend is just a slower way to lose the test.

Read the terms before the campaign, not after

Minimums, funding methods, refund policy on disapproved spend, and what happens to unspent balances - these are boring until they cost you. The single most important term is also the softest: does this counterparty have a reputation for dealing straight? In an industry where non-payment and vanishing balances are real risks, who you're buying from matters as much as what.

Watch for the fraud signals

Bot and junk traffic has tells. Implausibly high click-through with near-zero conversions, traffic that spikes the instant you launch and flatlines in quality, conversions clustered on a single placement that looks too good - any of these is a reason to pause and dig, not to scale. A clean small test that converts beats a huge one you can't trust.

Then, and only then, scale

Run the controlled test, compare against your stop-loss, and let the numbers decide. The buyers who do well over years aren't the ones who found a secret source - they're the ones who refused to skip the pre-flight. The directory's reviews and the community are both there to shorten it: learn from operators who already ran the test so you don't have to pay for the lesson twice.