This is Troy. 

Troy has almost a decade of experience in enterprise sales for technical companies. He’s very good at what he does.

Earlier this year, an early-stage startup named WarpStream hired Troy as its first salesperson. After ramping up, an early hire like Troy would normally hire a few SDRs. The SDRs would be responsible for prospecting: manually handling inbound interest, researching prospects, sending personalized emails, and booking first meetings. 

But Troy didn’t do that. Instead, he built a network of automated pipelines that do the SDR work for him. His automated tools contact potential customers faster and more efficiently than human SDRs ever could–for a tiny fraction of the cost. 

Troy treats prospecting like an engineering pipeline, and you should, too. Here’s how. 

Tools of the trade

Before we get into the actual pipelines that Troy put together (the recipes), we should talk about the tools that make them possible (the ingredients). Things have advanced quite a bit over the past few years, and in particular the website visitor identification tooling has gotten good enough to enable things you wouldn’t have been able to do just a year or two ago.

Website Visitor Identification: Koala and RB2B

Koala and RB2B help you figure out who has visited your website. They take an IP address and use publicly available data to work backwards towards an identity, company, and, ideally, contact info (although in practice, you usually need to use a different email finder). Koala works better at the company level, and RB2B (somewhat controversially) works better at the individual level.

Email sequencing and automation: Apollo

Apollo is the home base for your high intent pipelines (hand-raisers, people who fill out forms on the website, people who submit contact us inquiries, etc.) and low intent pipeline (people looking at content on your website). Think of it as a better version of ZoomInfo. 

Programmable spreadsheets: Clay

Clay is a data enrichment tool for personalizing outreach. RB2B only gives you the LinkedIn profiles of your visitors, whereas Clay can go much broader, so it’s essential for building out your contacts lists and integrations with LLMs. It also connects to several different tools via API, so it’s a central node in this stack. 

Email finders: FindyMail + LeadMagic

FindyMail and LeadMagic allow you to find and validate email addresses from any company. You can collect leads from anywhere, get verified email addresses, and load the data via CSV or API to any tool in your stack. FindyMail tends to be better for smaller scapes and LeadMagic tends to be better for larger scraping jobs (2,500+), especially if you want to build lists to support ads. 

Mass email sequencing: Smartlead

Smartlead is a mass email sequencer. If you’re doing cold / mass emailing, you need a specialized tool other than your normal sequencer: with Smartlead, you can connect to as many mailboxes as you want and cascade your email sequences over numerous inboxes, domains, etc.. 

ChatGPT

ChatGPT is the text generation tool of choice. In this context, you’ll use it to automatically write a personalized email for the prospects, plus accomplish glue tasks like normalizing text data and extracting tags. But of course feel free to use whichever LLM you like.

Workflow automation: Zapier

Zapier is another “glue” tool that helps connect these other tools to each other through a low code interface. It’s popular for building automated workflows that, for example, send an email when someone submits a form.

High intent pipeline

Troy splits his efforts into three pipelines:

• High intent: Prospects who have actively signaled interest in your product.
• Low intent: Prospects who have visited the site.
• No intent: Prospects who haven’t indicated interest at all.

Prospects in your high intent pipeline submitted information to you – likely through a Contact Us or Book A Demo form – and want you to contact them. When you contact high intent prospects, you can often get an open rate as high as 90% and a response rate of 60%. 

With high intent prospects, it all comes down to speed. Troy gets the highest response and conversion rates when he responds to an inquiry within (literally) five minutes or so. 

In Troy’s high intent pipeline, the Contact Us form on the website sends the submitted information to Zapier, which automatically notifies Troy across Slack and email, plus triggers the creation of a templated email in Apollo. 

Because high intent prospects have already raised their hand, little personalization is necessary, so again, it’s all about speed. 

Apollo auto-populates the template and fills in all the variables that give it light personalization. The email only needs to say:

• I saw you submitted this information.
• Thank you for submitting this information.
• Can we set up some time to chat?

Given how quickly the email goes out (and the fact that they’ve already expressed interest), these prospects usually respond. 

Emails can always get lost (or catch people at the wrong time), so Troy sequences these prospects, too, to ensure he can easily follow up later. 

Depending on your tolerance for meetings with bad-fit potential customers (or, worse, spammers), the high intent pipeline can be completely automated and requires little to no personalization.  

Low intent pipeline

Troy calls this pipeline "low intent," but you could also call it a second touch or double check pipeline. People in this pipeline have engaged with your brand in some way; in this case, it's people who have visited and spent time on your website. If you reach out to them shortly after they visit your web property, there’s a good chance you can convert the nascent interest that led them to the site to a deeper interest that leads them to a call. 

In this pipeline, response and open rates are lower, with about 40-60% of people opening and 3-5% responding. Note that this range varies quite a bit because of personalization. More personalization means better response rates, but it takes more time and effort to implement and tweak. For getting started, start with simple pre-canned templates, and enhance from there. 

The tool that makes this possible is a website visitor identification lookup tool like RB2B. These tools automatically identify who is browsing the website (US only for RB2B) and capture whatever information they can about them.

RB2B pings a Slack channel that Troy monitors, but it also ports the contacts over to a Clay table via a webhook. Clay can do a lot of enrichment on the profile, providing more information such as demographic (age, gender, occupation, etc.), firmographic (industry, company size, revenue, etc.), and technographic information (tech stack, infrastructure, etc.) from data sources like ZoomInfo, Clearbit, and People Data Labs. But in this case most of the data is ported in from RB2B. 

Then, Troy uses Clay’s integration with FindyMail to validate the email and, assuming the email is valid, pushes the profile to Apollo via Zapier. 

From there, Apollo activates a low intent pipeline sequence that Troy already has built out. The template is very simple, very light-touch, and says, essentially:

• I saw you on the website. 
• Let me know if you need anything. I’m here to help.

Even though the prospects are low intent, you can assume they know the basic value proposition of the product because they’ve been on the website. They’ve already displayed some intent, so little personalization is necessary. You’re just giving them an easy way to get more information. It’s like saying, hey, I’m here if you want to talk, if not, all good.

The low intent pipeline is completely automated. The gold standard is a 50%-60% open rate and a 3%-5% response rate, but once you have this system set up, you can tweak it to get to that level and maintain those rates with almost zero effort.

No intent pipeline (cold outbound)

Prospects in your no intent pipeline haven’t interacted with your website and might not know about you at all. This outbound is cold, so the pipeline here has to be more complex than the other two in order to be effective. 

The first tool up to bat is LinkedIn Sales Navigator (note, you can swap in tools like Apollo for this as well). Use Sales Navigator to find people in a specific segment. This could be all the people who use a particular tool, such as Kafka, or all the people at a certain seniority level in a given industry. 

You could search for people in California who use Kafka and don’t work for competitors, as in the example above, but you could also filter out people who aren’t technical and people who are entry-level (note, Troy only pulls contact information from people in the United States due to GDPR laws and regulations). 

From here, you can use Boolean operators to slice the results more narrowly. LinkedIn doesn’t let you export profiles from Sales Navigator, so you’ll need to use something like Evaboot to get a CSV.

With Findymail, you can scrape those results into a CSV, perform another narrowly sliced search, scrape again, and repeat. You’ll eventually have thousands of contacts in a CSV, and each row includes a LinkedIn URL, first name, last name, email, and fully enriched job title. 

This is the most manual step of the process, but once you have a clean list of data, you can use it as the foundation for the rest of the automated work.

Take the CSV file, load it into Clay, and start using the available filters to hone the list further. With Clay, you can go beyond Boolean operators filter based on who uses which tools in conjunctions and which team they’re on, for example.  

Keep running this flow, and you can end up with tens of thousands of people. With mass cold emailing, you’ll probably want at least a TAM of 20,000 people to use this methodology, or you will burn through it all in a few months. 

Alongside these broad searches, Troy also uses company-specific searches. With Koala, you can see which companies are visiting the website the most, scrape all the engineers at those companies, and sequence them so that you’re emailing one or two well-targeted people in parallel with your broader sequences. 

Now that you have these CSVs, you can use Clay’s enrichment tools to get their LinkedIn information, which you can then scrape and normalize so that you have a pure text clone of their LinkedIn profiles. In Clay, you can then tag these clones with whichever technologies they use so you can attach each profile to the use cases you’re targeting. 

Once you tag the profiles, ChatGPT can create an intro based on the background information you’ve captured. This can take a little iteration. Troy, for example, uses prompts that include examples of how ChatGPT should respond and prompts that give ChatGPT a role to play. 

Each response costs as little as two cents to generate, so the real investment will come from tuning a prompt that consistently gives you results you like. There are some prompt libraries out there that can help – and Clay provides thousands of templates to start with – but there’s going to be some upfront trial and error.

The resulting intro can be fairly simple, which you can see below. 

It’s not quite in-depth yet – there’s way more personalization you can build in – but the foundational elements are set. Once you can reliably and automatically generate a personalized hook, you can iterate over time so that the pipeline and the mass emails it generates improve at scale.  

From there, Smartlead will sequence your emails en masse. Email sequences are necessary for mass email campaigns because large scale outreach from the inbox can get flagged for spam. It’s easy to only send 50-100 emails through the high and low intent pipelines – keeping you under any spam filters – but if you’re sequencing thousands of leads a month, you need a way to maintain numerous inboxes. 

The no intent pipeline is partially automated. The first stage requires some manual scraping work, but once you complete that step, you dump the information into your pipeline, and it takes care of the rest. 

The gold standard for response rates here is 1%, but with your entire pipeline visible, you can fine tune over time and iterate toward that rate. With enough tuning, you could get as high as 2%. 

Save money and scale better

It’s almost ridiculous how easy it is to get much of this setup, how cheap it is to run everything, and how lucrative the results are. 

The low intent pipeline, for example, takes as little as twenty minutes to put together, and the high intent pipeline even less. Given how much more effective it is to respond quickly, building both pipelines is a no-brainer. 

The no intent pipeline takes a little more effort, but when you have it all built, you can have one salesperson manning the cockpit and accomplishing the work of an entire team. 

These pipelines, all in, cost about $1000 to $3000 a month, depending on the volume and level of customization. The investment of money and effort almost immediately pays itself off when you don’t have to hire a small army of SDRs. For comparison, an entry level SDR in the Bay Area at the low end will cost you about $60,000. Once you add in fringe and benefits (assume a standard 17%) the cost of a system like this is roughly 2x - 6x cheaper than traditional SDRs. 

Even better, this system scales with you and improves as you iterate. SDRs of the future will likely look like conductors of these kinds of pipelines, tuning them and scaling them up over time. More efficiency and scale, at a fraction of the price with less operational overhead…a true win-win.