I spend most of my week around people who run outbound at a scale. So I brought two of them into a recorded debate. Janis Plume runs Outbound Pros, a volume agency, the wide net approach, that sends six million emails a month and is scaling toward twelve, plus 200,000 LinkedIn messages a month. Artyem Jurkevich runs Revsculpt, where he builds custom signal systems, the spearhead approach. They sit on opposite sides of the same argument, and I wanted the receipts on the table.
The argument is simple. Do you go for volume and reach your whole market, or do you spearhead a small list with a sharp signal? I run both inside my own outbound, so I care about which one books more meetings, not the LinkedIn version of the answer.
Volume outbound, often called the wide net approach, means you reach your entire addressable market on a schedule. Janis described it as an insurance policy. Everyone in your market hears about your offer every six to eight weeks, so the pipeline stays roughly the same each month. He puts the floor for this approach at around 300,000 emails a month, and closer to a million if your market allows it.
Signal outbound, often called spearhead, means you wait for a reason to reach out. Artyem watches for a buying signal, qualifies the pain, and sends a message that speaks to that exact moment. The audience is small. The relevance is high. His view is that a complex product needs this, because you cannot blast 5,000 emails a day when your total market is tiny.
Here is how the two approaches lined up across the debate.
| Factor | Volume (Wide Net) | Signal (Spearhead) |
|---|---|---|
| Audience size | 300,000+ contacts a month | Small, signal-qualified lists |
| Reply rate | Often under 1% | 2% to 10% |
| Main benefit | Predictable pipeline | High relevance and conversion |
| Best for | Large or fragmented markets | Complex products, small markets |
| What you optimize | Positive reply rate and demos booked | Signal strength and pain match |
| Sales cycle fit | Short to mid ACV | Longer cycles, 3 to 12 months |
| Time to first data | Fast, once warmup finishes | Depends on signal frequency |
Before you pick a strategy, you need to land in the inbox. Both Janis and Artyem said the same thing. Deliverability is the first problem and the one that never goes away. Janis had spent his last two weeks fixing a placement drop after Google tightened its filters, a change he said about 99% of in-house teams never noticed. He walked me through the three things that drive deliverability more than anything else.
1. Copy : He put copy at the top. The email cannot read like a hard pitch, even though it is cold email. A harsh call to action, like a demand for a 60-minute call, gets marked as spam, and the filters already recognize cold email patterns and push them out of the primary inbox. A conversational tone is what lands.
2. Domain batching : You cannot buy 50 similar domains at once anymore. He buys them in small batches of around two, with 15 to 20 minutes between each batch. The reason is the same principle as mailboxes. If one domain in a large brand-linked batch gets blacklisted, the whole batch tied to your brand can go with it, and you throw away the domain spend. Spacing them out protects the rest.
3. Pre-warmed domains : A pre-warmed domain is a domain that has sat in email warmup for a long time, so it carries trust and skips the batching problem. Warmup is where automated conversations between mailboxes build sender reputation before any cold email sent. Both run new infrastructure through around two weeks of warmup before the first real message, and can start within a couple of days when the domains are already pre-warmed.
One point Janis was firm on is that ESP matching does not help right now. In his experience Google struggles to deliver to Google, and even Google's own mail lands in spam some days. When sending from one provider to another lands fine, matching the sending provider to the recipient's provider stops being worth the effort.
Many cold email agencies refuse Fortune 500 targets because of strict anti-spam filters. Janis does send to them, and his answer was not about infrastructure. It is about what you say. Large companies have many departments and many employees, so every message has to be relevant to that specific department and its pain. He calls this hyper-segmentation with volume.
His multi-threading method for big accounts is worth copying. Instead of sending four follow-ups to one person from one mailbox, he runs one-step sequences and reaches the same person from different mailboxes across the month, starting the conversation fresh each time. Because a large mailbox pool makes the odds of reusing the same mailbox tiny, each touch lands with different deliverability and reads like a new conversation.
This is where most people misread the two approaches. On volume, the reply rate looks alarming. Janis said his reply rate could sit at 0.2% while his positive reply rate ratio ran as high as 80%. He optimizes for positive replies and demos booked, not raw replies, and told me reply rate on its own does not matter to him at all.
On signal, Artyem sees reply rates from 2% to 10%. The number climbs with how triggering the signal is. A signal tied to millions in financial exposure pulls a much stronger reply than a soft mention.
The channel gap is just as large. Janis's email reply rate averages about 1% across markets, while his LinkedIn reply rate averages about 11%. That is roughly ten times higher, which is why he says clients ask for LinkedIn outbound every week.
Artyem drew a hard line between public signals and custom ones. Public signals are funding rounds, job changes, and LinkedIn posts. Everyone buys the same data, so a Series A announcement can trigger 600 near-identical emails in a week. Janis avoids these for exactly that reason, and prefers reaching people completely cold with better data.
A real signal is one you build before anyone else sees it. Artyem gave two examples. For an environmental consultancy, his team analyzes public factory pollution reports, then reaches out to the chief risk officer or whoever fits the ICP, with the exact pain that data implies. For a ship repair client, they track deficiency portals, note when the next survey is due and where the ship is heading, then contact the company with the timing and a price before the ship reaches port. The reply is almost always the same question. How did you find this?
He also combines signals to sharpen the message. On the ship example, location plus survey date plus deficiency history together make an opener that a single data point could not. His point was that a signal without context is just a notification, and a notification is not a signal.
You do not need many signals to start. He said most companies deploy two or three at first, solve the low-hanging problems, then expand. The heavy 65-use-case systems come later, once the simple ones prove out.
When you are ready to find and enrich those contacts, lead search and waterfall enrichment through Leadsforge pulls from a database of over 500 million contacts and lets you build lists by ICP, lookalikes, or competitor followers.
Both agencies run heavy AI agent stacks, and this was one of the most useful parts of the debate. Janis runs four people and 40 AI agents, and is scaling toward 100. His agents handle the tedious, repetitive delivery tasks nobody wants to hire for, like running programmatic placement tests every week or flagging when a mailbox disconnects from the sending platform. An internal analysis showed the current 40 agents save about 180 hours a month, and 100 would save closer to 300.
Artyem builds a custom operating system for each client instead. Every reply, every CRM change, and every call is tracked and fed back into a learning loop, so the context compounds over time. That includes insights pulled from sales call transcripts that are not public anywhere.
The warning both gave is the one worth remembering. You cannot automate a task you have not done yourself. Janis said founders jump to agents before they understand the process, which never works. You do the task manually for a couple of months, build a real knowledge base, then hand the well-understood parts to an agent. It is the same reason I tell founders not to scale outreach before they have product market fit and a channel that already books meetings by hand.
If you want an autonomous way to run outbound, Agent Frank is the AI SDR inside Salesforge. He prospects on your ICP, writes the outreach, sends it, and follows up, on Auto-Pilot or with your approval in Co-Pilot mode, so you can scale without hiring a larger SDR team.
LinkedIn outreach changes the math because the reply rates are so much higher than email. The catch is scale. The only way to scale LinkedIn is more accounts, and each account carries a low daily limit of roughly 25 connection requests and 30 messages.
Two account types do the work. Employee accounts keep the cost down, but many employees will not give consent to connect them. Avatar accounts fill the gap, and Janis says they often outperform real ones. His avatar profiles styled as retired professionals run a connection acceptance rate above 60%. He runs more than 200 accounts for clients.
A profile that looks like a person sharing a vacation story converts better than a corporate page, even though it sounds cliche. Reputable software keeps employee accounts safe, while heavy scraping from a personal account gets it restricted. Some avatar posts have pulled 80,000 impressions, and an agent handles the posting while a learning loop studies what worked.
One feature most teams miss is event invites. You can invite up to a thousand connections a week to an event, filtered by location, and the limit resets on Sundays. That works for large markets and matters even more for small, niche events. I run this myself with more than 40 accounts hooked into outreach, inviting people to events and webinars and going after competitor followers programmatically.
Multi-channel outreach across email and LinkedIn from one platform is the core of multi-channel sales sequences in Salesforge, and every reply lands in Primebox so you keep the conversation in one place.
One of the sharper ideas in the debate is that volume can build its own signal. When you email the whole market, some people never reply but go visit your site. Janis captures that traffic with a few tools. Microsoft Clarity shows heat maps of where people click and drop off. Captivate calls each visitor or drops them into a sequence through an interactive chatbot. De-anonymization tools like RB2B enrich the visitors so you can follow up by email or LinkedIn without ever saying you saw the visit.
His logic is that you cannot reach a real signal without going wide first. Volume feeds the market, the market reacts, and that reaction becomes your own signal. The more you send, the more signal you generate.
Artyem was more cautious here. A blog visit alone does not mean someone is ready. He would rather track multiple steps, like a visitor who reads five posts and clicks two calls to action, then let an agent decide it is worth a reach-out. Both agreed a strong web presence matters. Weak content means no one wants a conversation, which is why they build custom landing pages and redirects for each ICP.
On data itself, Janis prefers custom scraping over databases like Apollo or ZoomInfo, because everyone uses the same stale records. His team scrapes where a target audience actually gathers, then enriches domains, LinkedIn URLs, and contacts through a multi-step tool. High volume does not have to mean low quality.
Both lean on lookalikes, just from different angles. Artyem reverse engineers why an existing customer bought, then builds a lookalike for each use case, so 65 use cases can produce 65 lookalike audiences. Each one still gets a two-step qualification pass before any send, so noise does not slip through.
Janis keeps it simpler. Take your best customers, build lookalikes of them, and send. He also builds lookalikes on positive replies, since that pool is large and already leans toward interest. Competitor follower scraping is his other favorite, because those people opted in to a market and pre-qualified themselves. He has scraped competitor and adjacent-company followers for clients and reached audiences in the hundreds of thousands.
Two reverse plays came up as well. When a deal closes, pull similar companies and open with the fact that you already work with someone like them. When a good-fit prospect says no, use them as social proof to a competitor who would want the edge. In some markets, naming a big player you work with is enough to make a smaller one sign.
My honest take is yes, most teams should run both. They serve different intent, the way Google ads and Meta ads serve different intent. Both agreed you can hire for both, and Artyem framed the split as expertise. Large-scale volume and custom signal systems are two different skill sets, so the real question is who owns each.
Asked which one to start with, Janis picked his own approach, and his reasoning held up. You cannot build a real signal-based motion without data, and volume is how you generate that data. His method is to reject the idea that a client's stated ICP is the whole picture, set up 50 more ICP profiles, run around 150 sequences a month across them, then score the results and expand only the winners. That scoring is what tells you where a signal actually lives.
Artyem's counter is that a complex product with a small market cannot go wide, and needs pain qualification per ICP from the start. Both paths meet in the same place. You end up qualifying pain by ICP either way. Volume just gets you the data faster.
The math is what convinces a CFO. Janis framed a worst case of 50 meetings in month one at a $10,000 spend, which is about $200 a meeting. On a lower ACV, close rates often run 70% to 80%, so by month two the pipeline compounds and the outbound pays for itself. Once it is profitable, you raise volume rather than stop.
The event case stuck with me. Artyem had a client fly to a three-day event, book 62 meetings, and close 13. With one ACV at $100,000, that is $1.3 million in closed business from a single event.
Pricing lands where you would expect. Volume starts around $10,000 a month, because roughly 40% goes straight to infrastructure and data. Signal-based systems run $5,000 to $7,000 for a proof of concept, then $10,000 to $15,000 a month once inbound and CRM work is included. Here is how the two compared on the numbers they shared.
| Cost factor | Volume (Wide Net) | Signal (Spearhead) |
|---|---|---|
| Entry point | Around $10,000 a month | $5,000 to $7,000 proof of concept |
| Ongoing | $10,000+ a month | $10,000 to $15,000 a month |
| Time to value | 2 to 3 weeks warmup, days if prewarmed | 3 months to prove or disprove |
| Typical sales cycle | Short to mid ACV | 3 to 12 months |
| Best justification | Predictable monthly meetings | Fewer, higher-value deals |
The comparison both reached for is paid ads. A cold email CPM can run ten times cheaper than Google or Meta, with more targeting control, because you can send a custom offer to a custom audience at a fraction of the cost per thousand. To start with Google ads in the US, Janis estimated at least $30,000 before management fees, closer to $50,000 all in once landing pages are built. Outbound lets you cut that several times over.
On in-house versus agency, his argument was time and cost. Hiring a go-to-market engineer for outbound runs around $8,000 a month, plus recruiter fees, so the first month can approach $30,000 before you build any infrastructure. An agency runs a prebuilt stack from week two. Artyem's version is simpler. Agencies ship faster because they have done this across many industries.
If you want to run either motion yourself, start with the infrastructure and warmup, and the sequences on top. Set up domains in Mailforge, warm them in Warmforge, find contacts in Leadsforge, and send through Salesforge. The stack stays connected end to end.
Volume, or wide net, reaches your entire market on a schedule, usually every six to eight weeks, for predictable pipeline. Signal, or spearhead, waits for a buying signal and sends a highly relevant message to a small, qualified list.
Reply rates on volume outbound often sit under 1%, sometimes as low as 0.2%. The metric that matters is the positive reply rate ratio and the number of demos booked, not the raw reply rate.
Public signals like funding rounds and job changes are commoditized, so everyone targets them at once. Custom signals you build from data before others see it still convert well, with reply rates from 2% to 10%.
A volume, or wide net, motion usually starts around 300,000 emails a month. To hit your whole market on a schedule, many operators aim for closer to a million a month.
Yes, but it takes hyper-segmentation by department and pain, plus multi-threading that reaches the same person from different mailboxes rather than repeated follow-ups from one.
Most teams benefit from both, since they serve different intent. If you start with one, volume is usually first because it generates the data you need to build a real signal-based motion.
New infrastructure needs about two weeks of warmup before sending. With pre-warmed domains you can start within a couple of days, and first meetings usually follow inside the first month.
