B2B prospecting software: a buyer's checklist

B2B prospecting software is sold in a demo, and a demo is designed to be watched rather than interrogated. The seller drives, the data is prepared, the example account is one where the tool looks good, and the forty five minutes are structured so that the difficult question arrives at minute forty three when everyone is already thinking about their next call.

This is the list of questions to send before the demo, in the order that saves the most time. Most of them can be answered in writing, and a vendor who will not answer them in writing has told you something useful.

Decide what you are actually buying

The category contains four different products that all describe themselves the same way. Before any vendor call, decide which of these is missing from your stack, because a tool that is excellent at one is usually mediocre at the others.

• A list. Contact data: who works there, what their address is, what the company does. You are buying coverage and accuracy.
• Signals. Something changed at the account: hiring, funding, a technology added or removed, a leadership move. You are buying timeliness.
• Execution. Sequences, timing, reply detection, reporting. You are buying reliability.
• The writing. Something produces the actual message. You are buying judgement, which is the hardest of the four to evaluate in a demo and the easiest to fake in one.

Teams that skip this step buy their third overlapping tool and then spend a quarter deciding which two to cancel. Write down which one of the four you are short of, and hold the vendor to it.

Questions about the data

1. Where does the contact data come from? Not "multiple sources". Ask for the categories: public web, contributed from customer address books, licensed from a partner, inferred. Contributed data is the one to ask about twice, because it means somebody's address book was uploaded, and it is worth knowing whether yours will be.
2. How old is a typical record? Ask for the median age of the records in your target segment, not the headline database size. A hundred million contacts is not a feature if the ones in your industry were last verified in 2023.
3. What is the bounce rate you will stand behind? If there is a credit refund for bounces, that is a real commitment. If there is only a marketing number, treat it as marketing.
4. What is your lawful basis for holding EU personal data, and how do you handle an erasure request from a person who never contacted you? This is the question that separates the vendors who have thought about it from the ones who have not. Ask it even if you only sell in the United States, because the answer tells you how the company thinks.

Questions about the sending

1. Does mail leave from my domain or from a shared pool? A shared pool is cheaper and makes your deliverability a function of the worst sender on the platform. That can be an acceptable trade. It should be a decision rather than a discovery.
2. What is the hard ceiling on daily volume, and can the tool exceed it? A throttle you can raise is not a ceiling. Ask specifically whether anything inside the product can send past the number you set, for any reason, including catching up after an outage.
3. How is suppression enforced, and at what moment? Checking a suppression list at send time means the message was already written, queued and counted. Checking it at draft time means the contact never enters the pipeline. Ask which one, and ask what happens to an opt out that arrives by reply rather than by clicking.
4. What happens when the complaint rate rises? The answer you want involves the sequence pausing itself. The answer you will often get involves a report you can look at.
5. Is one click unsubscribe implemented in the headers? Since February 2024 the large mailbox providers expect it from bulk senders, along with SPF, DKIM and DMARC. If the vendor is vague about this, they are not paying attention to the environment they operate in.

Questions about what it writes

This is where the demo is least reliable and the questions matter most.

1. Where does each fact in a generated email come from? Ask them to open a draft and point at a sentence. If the answer is that the model produced it from the research, ask which research, and ask to see the line. A tool that cannot show you the source of a claim cannot help you defend that claim later.
2. What happens when the research turns up nothing? The two possible answers are that the email gets more generic, or that the email invents something plausible. Only one of those is acceptable, and you find out which by running the tool on an account with almost no public footprint.
3. Can I see what it decided not to say? A tool that flags its own weak claims is showing you its reasoning. A tool that returns polished prose and nothing else is asking for trust it has not earned.
4. What is checked before a message is allowed to send? Ask for the list. Spelling and merge tags is a legitimate answer, it is just a much smaller answer than the pitch implied.

Questions about controls and records

1. What exactly is stored per send? The message is the minimum. What you want is the message, whatever the tool checked, and the result of that check, kept together.
2. Can I export it, and in what format? Ask for a sample export file before you sign. "Available on request" and "downloadable as a CSV whenever you like" are different products.
3. Can approval be enforced by role? Not "is there an approval setting", but can an individual turn their own approvals off. In most tools they can, and in most teams somebody eventually does.
4. What is the single control that stops everything? Ask them to show it, live. Then ask what is still in flight after it is pressed.

Questions about the contract

1. Is the price per seat, per contact, per send, or flat? Each shape creates a different incentive. Per contact pricing rewards you for uploading fewer, better contacts, which is good. Per seat pricing punishes you for putting the tool in front of the people who should be reviewing its output, which is not.
2. What happens to my data if I leave? Ask for the export window in writing.
3. Is there a minimum term, and does it auto renew? Ask for the notice period, and put the date in a calendar the day you sign.
4. Who is on the other end when something goes wrong at nine in the morning on a Monday? Deliverability incidents happen on Mondays.

Four answers that should end the call

Not because the vendor is dishonest, but because they are describing a product that does not fit anybody who asked these questions in the first place.

• "The AI handles compliance." No tool can establish that you had a lawful basis to contact somebody. A vendor who says this either has not thought about it or is hoping you have not.
• "You can send unlimited volume." Unlimited volume is not a feature of a sending tool, it is a description of the risk you are being handed.
• "Our customers see a 4x reply rate." Ask for the sample size, the period, the segment and whether it was measured or reported. The answer is almost always that it came from a case study of one enthusiastic account.
• "We cannot show you the source, the model just knows." Then you cannot defend anything it writes, and you are the one whose name is on the email.

How to run the trial

Three tests, in this order, and none of them takes more than an afternoon.

The obscure account test. Point the tool at a real prospect with almost no public presence: a family owned manufacturer, a regional logistics firm, a business whose website was last touched in 2019. A good tool produces a short, careful, slightly boring email. A bad one produces a confident paragraph about their digital transformation journey.

The known account test. Point it at an account you know intimately. Read every factual sentence and mark whether it is true, unverifiable or wrong. Count the middle category, because unverifiable is the category that eventually becomes wrong in front of a buyer.

The record test. Send one message. Then, a week later, ask the tool what it sent, what it checked and why it went. If reconstructing that takes more than a minute, it will take a great deal more than a minute when somebody complains.

Everything above is why we describe our own product as prospecting tools that cite the research behind the pitch, and why the demo shows the claim ledger rather than just the email. If a checklist like this makes a tool look bad, that is the checklist working.