Lead qualification is a choice you make before the lead exists

Lead qualification is the decision of whether a lead is worth a salesperson's time, and how much. It is also called sales qualification, or qualifying leads. Most teams split it: marketing marks a contact as a marketing qualified lead (MQL) once it clears a scoring threshold, and sales accepts it as a sales qualified lead (SQL) or hands it back.

The rubrics are well documented. BANT checks budget, authority, need and timing. MEDDIC goes deeper for larger deals: metrics, economic buyer, decision criteria, decision process, identified pain, champion. Both do the job they were built for, which is making one rep's judgment legible to another.

What both take for granted is the lead. Every question is asked of a record that already exists, produced by an earlier choice about who to contact and when. That choice is graded nowhere. A rubric can tell you the person in front of you is a poor fit. It cannot tell you the list was wrong before anyone looked at it.

Bradley Jacobs relayed a call with a market researcher who reported 0.1% reply rates on his connection outreach. Jacobs' reading was that reaching out cold, with nothing published behind you, gives the other person no reason to care. No qualification step catches that. By the time a framework runs, the expensive decision has already been made.

daedalus, who runs an agency, described letting an AI SDR, the software standing in for the rep who makes first contact, run on a client account for two weeks without checking the prompt. It booked 14 calls. Eleven were with interns or office managers. By his account it cost the retainer. He audits every sequence now: "Automation without audits is just fast mistakes."

The system did the job it was measured on: it booked meetings. A rubric would have graded an intern out in one question, but it never ran, because grading starts once the meeting exists and by then the money is spent.

Kyle Asay, who leads sales at LaunchDarkly, described scrapping an AI SDR a couple of months in. Emails written for prospects were reaching existing customers, and response rates were near zero across every campaign. His root cause analysis named several factors, in the vendor's model and in their own approach, but he put data quality first: duplicate accounts made a customer look like a stranger. The evaluation team, he reports, had been sold on the idea that the data did not have to be perfect.

Asay is explicit that he is not against the category, and believes the right technology implemented the right way beats a human-only team. That is what makes the account worth reading. The failure was not in the judgment applied to each lead. It was in what the system was pointed at, and nothing in the process was looking there.

By Dra's account, the infrastructure to send ten thousand emails fell from thousands of dollars a month to a couple of hundred, the copy comes out of the same three models, and the database is the same 275 million contacts sold to whoever pays. He puts average cold email reply rates at 8.5% in 2019 against 3.43% this year.

When the volume, the copy and the records are identical across every campaign, the only thing that differs is who you chose to contact. Whoever builds the list, he argues, decides the reply rate before a word gets written. The people who do well, he adds, are the ones who stopped blaming the copy.

Gohar Kabir, who sells a competing data product, reported reviewing 200 million bought contacts. Among those from the two largest sellers, roughly one in six were no longer at the company listed. About 15% came back catch-all, meaning a domain that accepts any address, so the record cannot be verified either way. Another 3 to 5% were the same person appearing twice under a different email format or an old domain. He cites a test by Eric Nowoslawski where 500 addresses that had already replied positively still graded 14.6% catch-all. A list is a snapshot of a world that keeps moving, and the decay is invisible downstream.

Aaron Shepherd, who sells an alternative, points at sourcing: when every sender builds from the same database with the same filters, they reach the same inboxes, which is why a list that worked last quarter goes quiet with nothing visibly changed.

Yuriy Zaremba, who launched a competing product in the same breath, said $30,000 to an outbound agency bought thousands of emails a month, exactly as promised, and zero meetings. His reading: they spent everything on the list and the copy, and neither is why a busy person replies. People reply when they have a real problem, something just proved it, and the message arrived at that moment.

Loic Jeanjean, who advises companies on how they go to market, piloted one of these systems against about 200 people in 45 days, by email and direct message. It booked one meeting; that person did not show. Warm intros in the same window produced five calls, all with people who already knew him or someone he knew. An account can match the profile exactly and still not be qualified, because nothing has happened to make now the moment.

Dra says the list decides the reply rate. Zaremba says the list is not why anyone replies. Both point at the choosing, which has two axes: who, and when.

One growth lead reported testing a client running 80% of volume through a professional network: same offer, same list, 0.8% reply there against 4.2% by email. Dan Rosenthal, who sells connection outreach, claims connection acceptance of 40 to 50%; Benou cites a study of more than 50,000 requests putting it at 30 to 37%, against the 60% that outreach ROI calculators assume by default. Identical targeting can still swing five-fold on delivery alone. That is an argument for fixing delivery, not evidence that the list was ever graded. Changing channel moves the denominator, not the decision.

Aubrey, who runs a lead generation agency, let an AI SDR run outbound for a client. It booked zero calls and burned a sending domain. Set to research only, with the send done by hand, the same system, same accounts, same pitch, produced nine calls in two weeks. Aubrey's conclusion: "Automate the research. Not the send."

Nick Abraham, who runs a lead generation agency, reported three reversals. Artisan spent $2 million promoting the slogan "Stop Hiring Humans", then retired it, repositioned its agent alongside human reps and started hiring its first human rep to do that outreach. Ramp, by his account, shut its internal AI SDR program in December 2025; it had once produced 30% of pipeline but could not support an increasingly complex sales motion. He also points to a March 2025 TechCrunch report that some 11x customers stopped seeing effective leads after about a month.

Abraham's own read is that these systems work when the offer already converts cold traffic, which few do. He would give the machine research and enrichment, drafting and next-step suggestions, organizing replies and surfacing intent, and routing important conversations to people. Targeting, offers, campaign changes and the reading of replies stay with people.

Cody Schneider set out what teams do instead: personalization landing on the wrong variable, a line about where someone went to college, and, when it stops working, buying more data and adding more mailboxes. That is volume applied to a judgment problem. Abraham's division is a shape an agent can hold: continuous research, a ranked shortlist, a person on the send.

Every number upstream of a held meeting can rise while nothing underneath is qualified. Sends, replies, leads generated, calls booked: each improves on its own. Schneider ends with one measurement instruction: meetings held and pipeline, nothing above it. He is not arguing against volume, he pairs it with a precise list. Reporting reply rate as the headline number is the habit he files under the mistakes.

Most teams stop counting at the booking, and the gap is large enough to swallow the difference between a good program and a bad one. John Magnor described a founder whose show rate moved from 48% to 79% in two weeks after he sent each booked lead a short personal video the day before. Magnor's line for it: "Nobody no shows a human. They no show a calendar invite." Ben Amaliri described businesses spending $500 to $1,000 a day on leads, taking a 60% show rate and doing nothing with the other 40%. His point is that the cheapest revenue available is usually not another ad. Neither gap is a qualification problem. Both open after a lead was already counted as qualified, which is exactly why a booked meeting cannot be the number you run on.

John Karsant, who runs an appointment setting company, refuses the brief of running a client's outbound so they do not have to think about it: an agency with zero client input and zero sales feedback fails, whoever's name is on it. Qualification is a judgment the business keeps making, delegable in pieces but not as a whole.

The same pattern holds after the meeting, where sales automation only pays on selling already written down and graded. Before it, the work is choosing: an agent doing the research continuously and handing over a ranked shortlist, with the decision left to a person. Count the meetings that happened and the pipeline behind them. Everything above that line measures effort.