Lieben and Nowoslawski publish lists databases cannot build
Two list workflows landed the same day, one arguing for 34 ICP questions and one for asking an API 100 questions per company.
Two list-building workflows landed on the same day, both arguing that the list a database hands you is not the list you want.
Michel Lieben says his team went through 1M+ cold emails in Instantly and found that tight lists of 500 to 1,000 people with a reason to reply got 20-30% reply rates, while 100K blasts got 2-3%. His fix is an ICP broken into 34 questions across four parts, company fit, user persona, buyer persona and signals. Company fit starts with exporting two lists from the CRM, the companies that paid and the ones that paid then left, and checking every answer against both. If a best customer drops out of your filters, one filter is too tight. Whatever the churned accounts share becomes a do not target line, which in his worked example is teams with only one SDR.
The signals part is the one most senders will recognise. Lieben's point is that a signal alone only means the timing is good. An account becomes Tier 1 when the company passes part one, your user works there, and you can name who sits in each buyer seat. Those accounts get the first emails, opening on the signal that fired and tying it to the number the user is judged on, which for an SDR Manager running four reps is meetings booked per rep.
Nowoslawski rebuilt a churned client's list for $40
Eric Nowoslawski posted a four step build for a customer who churned because no tool would let her build a list of companies selling payment processing, ERP, accounting software, accounts receivable solutions, point of sale and other financial software. He says no database can answer that because of how databases are structured.
The build used Clay CLI to enrich existing customers and find the industries and keywords they appear under, then a deliberately loose audience search that pulls in companies that do not fit, then the Parallel Web Systems decisions API to sort them out. He says the whole thing cost $40 of Parallel Decision credits, and that for .001 you can ask 100 questions in one request. Claygent then finds the page proving the company sells the thing. The result is an always on audience that qualifies new companies, with HubSpot synced in so closed deals add new filters.
For .001 you can ask 100 questions in one request and get really great answers back.
For people sending, Lieben's reply spread compares two different sending motions rather than two lists against the same offer, so treat it as an argument for targeting, not a measured lift. Nowoslawski's build is the more concrete claim, and it is consistent with his earlier point that usable lists come from stacking filters, not from one search.
A full ICP has 34.
Know what is landing before you burn another domain
One email a day. Under three minutes. What the operators found out yesterday.