Morning, like your hat Willie look festive.
Yeah, I just, I just got done seeing a friend and she was festive, I was festive, you know, it was all good.
Good morning. Listening to Christmas music this morning. So I'm starting to get into the spirits of the law.
Yeah.
Let's see. We'll give it another minute or so. But yeah, thanks for joining.
Yeah, no, thanks for setting this up. I appreciate it. I'm super excited.
Cool. All right, cool. All right, cool. And I'm just going to give a quick overview right now. I Might do it again in a minute if some others join, but yeah, today what I'm gonna do is give an overview of kind of like our recommended flow. There's, as you've probably all started to see, there's thousands of ways you can use Clay. This is just one of our recommended flows. So if you have any questions about either workflows you've tried to create before or workflows that you're interested in, just either save your question for the second half or drop into the Zoom chat and I'll get to it as I can. But yeah, The first part will be pretty quick, probably maybe 20 minutes or so. I'll just be showing the workflow, and then for the rest, we can really dive into any specific questions you have about the workflow I showed, or if you've been trying something else, we can go into that as well. But I'm gonna go ahead and actually get started. Give me one sec. Cool. So the first off, just like the way I like to explain Clay overall, when you're first starting to use it, our goal is to create creative tools for sales teams to be able to automate and experiment with different ICPs, different messaging, really trying to make this as easy as possible. And So when we think about like the typical SDR or BDR role, or really any role that involves a lot of this manual work, there's kind of five core functions, finding companies, finding people, finding research, auto-generating, or not auto-generating, I guess manually writing emails to those contacts. And then finally is finding contact information. All five of those pieces are things you can fully automate within Clay. Today I'm gonna go through all those five. I'm gonna mainly go through, starting with finding companies with our LinkedIn database. There's some trigger-based databases that you could use that I'm gonna mention as well, just for ideas. We'll probably cover these in later sessions as well. But let me share my screen. So like I said, I'm gonna start with finding companies. So if you click new table, you'll see we've got all of these sources here. So these sources are pretty flexible. Like I said, I'm going to do find companies, but if you're interested in more trigger based sources, ones that I would recommend checking out that I just want to call out really quickly. First two are going to be importing companies from your CRM. So both HubSpot and Salesforce we connect to natively. If you choose a specific list that you want to import from HubSpot or Salesforce, these will also update every 24 hours. So we have a lot of users who will set a specific list. Maybe it's like their inbound website form list in HubSpot. Every 24 hours, those leads will come in, and then they'll probably follow a similar flow to what we're about to show right now. One other, I think, Really interesting one that a lot of users have been using lately is our find jobs from LinkedIn flow. I think a lot of people initially assume this is just for recruiting use cases. The reason this works so well though is, you know, let's say you are hiring Or let's say you're selling technology to help with like the management of creative processes. Let's say you're ClickUp actually for a good example, right? So you help create tools that allow these marketing teams to better manage everything that they're doing. If you use find jobs, I'm not going to go fully into this, but if you use find jobs, you could say, okay, look for creative directors or campaign strategists or account managers, look for these jobs. And then suddenly you get access to buy signals that, okay, these companies are investing resources into roles that likely will be managing these different tools or managing this creative process. And then once you find jobs, you can really skip the rest of the jobs flow and you can just go straight to enrich companies. This is like, it works pretty much across every single use case. If you want to use hiring as your signal as a potential buy trigger, yeah, find jobs is super easy. But like I mentioned, I'm going to do find companies. So find companies, this is going to be similar type filters to what you see on LinkedIn sales now. The database is of 70 million plus companies. So what I'm going to do is I'm just going to filter this down really quickly to software development companies. I'm going to do 50, we'll do 200 to 1,000. And then I also highly recommend, make sure you choose this company type filter. Most people don't use this when they're prospecting on LinkedIn sales nav, but this is gonna really help filter the results. So in this case, let's just do privately held. It will leave out for us like the non-profits or the partnerships or self-employed smaller companies that we don't necessarily want. You can filter by these keywords if you want to. I'm gonna leave that out and I'm just gonna go to New York. And then I'm going to just limit this to 50 just for this example. So something to note here, we do have, as you can see, If you're on the Explorer plan, up to the Explorer plan, it's 2,500 records per search. If you're on the Pro plan, it's 5,000 records per search. This does not mean that you're capped at 2,500 or 5,000 records overall. It's just per query. So you can still do multiple queries and quickly add up to as big of a list as you want. But we've got a preview here. These look good to me. If you want to at any point, you can adjust with keywords or filters and go as you want. But I'm going to go ahead and import these companies to a new table. Perfect, okay. So now that we've got our companies, there's a few things you could do, a few routes you could go. If you have a really broad ICP, you could jump straight into finding people. In my case, I'm going to do some research and actually filter these companies down a bit more. I'm gonna show off Claygent too. It's like one of our most powerful new features. I've honestly been using it as just like a replacement for some of the more complex workflows I've been creating in the past. But let's say in this example, and for the example of this entire workflow, let's say I sell technology that helps you automate demos. So instead of having to hop on a call to give demos, you can send a personalized video that maybe mentions this specific person's use case and their name, and it allows me to just send out these pre-recorded demos that feel personalized. So for this, you know, I have a list of companies, but the first thing I want to figure out is just if the company even offers a demo. So if you go to enrich data and you search Claygent, we'll use this to figure out if they offer demos. Just really quick, you do need your own API key right now to use Claygent. But once you connect your API key, you have full access to this. And so what Claygent does is it essentially combines OpenAI's ChattyBT with web scraping. But you can think about it almost like it's self-aware. So if we were to give it a goal of, you know, figuring out, for example, in this case, figure out if this company offers a demo link on their website, What it could do is it could search for company name plus demo. It could search for, it could then do another step if it doesn't find that, and it could just look on the company's website. You can give it multiple different instructions and it can essentially complete multiple steps until it finds an answer to your question. So, in this case, and with prompting in general, I recommend really be as precise and clear with your prompting as you can. Any extra words that are not really needed can sometimes cause results to be worse than what you're looking for. So in this case I'm going to say you're researching companies to see if they offer a demo of their product. Okay so you're researching companies to see if they offer a demo of their product. I'd recommend searching for, and then I'm gonna give the domain of the company. So I'd recommend searching for company domain and seeing if they mention Keywords like book a demo or schedule a demo. Okay. So this is the general prompt. What I want to do, though, is make this even tighter. I don't want the cases where, you know, it might see the word demo and just assume that it has a demo. So I'm going to say, essentially, be extra certain. Be certain that there is a CTA to book a demo website. And I'm going to add in one more rule just to be extra certain. I'm going to say whatever you find must be on domain. Can't be on any other website. So I've actually found this is a pretty important line. Sometimes what it might do is it might look for a company name and then Crunchbase. It might accidentally go to Crunchbase. And then on Crunchbase it will find a demo link. Which in its mind, right, will be like, okay, this is spot on. This is right. But it's these little areas where things can get a little bit mixed up that we really want to try to limit. And then lastly, I'm going to say if you find a demo CTA, yes. If not, output no. And that's all we're going to do. You can select the model. You can select the max cost. Because this does use your own OpenAI API key, sometimes it is helpful to set a max cost. But for the most part, these should be pretty low cost anyways. But I'm going to go ahead and run this for the first ten rows. So let's check out some of these results. So we've got diligent.com. If we go to their website. Okay, cool. We've got talk to sales request a demo. Let's check out a no.
Cool.
Free product. Let's see what instant help goes to. Yep. Okay, So this looks solid so far. And let's check out one more no actually. Okay. So we have a get in touch link. Nice. This just goes to a contact form. So this is perfect. So far, we're three for three. If you are brand, brand new to Clay, for many of you, you may already know this, but if you're brand new to Clay, you'll see I've only run the first 10 rows. If you want to run the entire column, just click on the column, go to run column, and then you've got a few options. You can choose a number of rows to run if you just want to test another batch. If you want to just run all like I'm going to do now, just click force run all 58 rows. And so as we start to get these answers to, the purpose of running this and running Claygent was to filter down my companies. Something to note that's pretty important. If I were to go to filter right now and I were to try to filter based off of demo availability, You'll see I only have the option to do has error or does not have an error. So if I remove this, the reason for that is that this yes or no right now is just a preview. If you want to filter by it, you need to map it to your table. And so if you click on just any of the cells, you can see everything that's in here. So first I'm just going to add the result as a column and I'm going to say, I'm gonna name this hazard demo link. And then the one other thing I want to show you in terms of Claygent, let me find a good example of this. Okay, cool. So this is always really interesting to look at. If we open up this Claygent column where it said yes to silver line CRM this time actually, so it did change its answer. We can see exactly what it did. So okay, search for Google, search for the site plus, book a demo or schedule a demo, and it did find availability scheduling, which we could check this link. Okay, so this is actually wrong. This looks to be just a feature. It's not an actual link to schedule a demo. This is another feature page. This is another feature page. But it looks like it actually visited that first page. And it did find slash demos. Okay. So this is pre-recorded page. We can see if we go to the reasoning, Serverline offers the option to schedule a demo for the calendar anything product. This indicates that there is a demo option. So this is actually incorrect, right? It found schedule a demo for one of the products and what we want to do. And the reason why I also would typically recommend just running things 10 at a time and really checking all your results is we want to add into this prompt, make sure schedule a demo is not a feature of the company. It must be a it must lead to a link for me to actually talk to someone. We can make this more clear. But yeah, essentially I would just go through these different iterations until let's say until you're at like 10 out of 10. And then at that point, you can run it for the full company. So now that I have run this, and I've also added this column, you'll see I can filter this to has a demo link. And I'll say has a demo link must be equal to yes in my case. All right, so I filled it down. So we went from 58 companies to 34, which is good. So the last thing I'm going to do on this people table, or sorry, on this company table, I'm actually going to write a first line about the company based off of the company information. So a lot of people really try to do personalization based off of, you know, for example, someone's personal LinkedIn profile. That can work and it can work really well. The issue with that is let's say you wanted to personalize a first line based off of someone's LinkedIn, like about, which is kind of the top of their LinkedIn page. That's going to be missing for a lot of people. And if that's missing for a lot of people, it's not something that you can necessarily scale, or you might have a lot of gaps as you do this. The other thing just to mention with personal LinkedIn profiles, which I'd say is especially important with Google's new spam rules, a best practice we always try to keep in mind is how can we make an email so valuable that people actually want to read it? It gives them information that is helpful. Whenever you're doing the personalized LinkedIn lines, if for example you are mentioning a pizza restaurant that's in their area, that can work really well. And it's a really cool little touch that oftentimes people don't notice are automated. At the same time though, let's say you were sending a thousand emails a month. One, like the number of people who will judge that as valuable is likely going to be less than something that truly addresses whatever problem they might be having. And second, when you're sitting at that scale, there's just more variability for it to go wrong. And for that reason, I really like to set up campaigns that have either a mix of those two variables or focus primarily on company-level variables. And so for a really easy company-level personalization, you'll see we already have these LinkedIn descriptions, which explain what companies do. And A first line that seems extraordinarily simple, but I've seen work very well, is if we go to use AI, and we browse our templates, we have a template that is use a company's mission to write email first line. And I'm just going to use this template. Honestly, like you, this is very, very basic. I definitely would recommend adding more personalizations, but this also kind of seems like a no brainer to me in the sense that like, if you're going to send a personalized email, at the very least, have something like this. And what this is going to do is it's going to generate a sentence based off of the LinkedIn description and it's very simply going to say I was on your site and saw you dot, dot, dot. This is just, again, a very, very easy one that you can add in. And we can try this on five rows for free, so you can see some of these results. Cool. So you can see here are some examples. A lot of people will use the templates as starting places. So if you want to guide it specifically on tone or let's say you wanted a sentence that was like seven words max, you could make these adjustments within the prompt. Right now I'm saying complete the output in under eight words. So I'm giving eight words on top of these seven that I've given here. But I'm going to go ahead and run this for the first 10 rows. And something also important to know here, you might see in the output, sometimes there will be quotation marks from ChachiBT. In this case, actually, they all came out correct, but let me go ahead and run this for the whole column. And this is actually coming out clean. But okay. There we go. We've got one. So something I'd always recommend doing as just like a extra safety gap or safety measure, whenever you're using Chatshift, there are cases where it could hallucinate, it could output something that was not just along your guidelines, it could output extra characters. Even if you say specifically in the prompt don't include quotation marks, this could happen. A really easy way you can fix this and make sure nothing ever slips through the gaps is using formula columns. With a formula column, we're able to manipulate this data pretty much however we want. A really easy thing we could do is we could say remove quotation marks from and then when you type forward slash you can select a column. And so we'll do from this response column. And so this is very easily oops very quickly going to remove quotation marks. Another thing you could do, let's say you really want to make sure that Chatshubt never outputs a sentence that's accidentally like 30 words long and just makes no sense. Another way we can use formula columns is to count the number of words in this column. Cool. So we're seeing the number of words here. Let's see if there's any that are really large. These actually all look really good. But just to show you, okay, if let's say, for example, one of these is way too large. What you could do is then use AI again, and this is a really common practice, using AI to check AI. We could say, shorten this sentence so that it is no longer than, let's say, 13 words. Here's the sentence. And so what we're going to do is we're going to input the first sentence that AI generated. We're gonna shorten it to 13 words. And then If you click on settings, you can set run conditions. These run conditions, these conditional formulas, they're really helpful because now what we can do is we can say only run this column if and then we can select that count variable that we just created. So only run this column if the number of words is greater than, let's say, 15. The reason this is extremely important, and you'll probably find this more and more important as you continue building things in Clay. This is going to really help optimize your credits and make sure that you're not spending on places where you already have a good answer. So you'll see, right, 14 minutes, it's not gonna run, you're not gonna spend credits. And let's actually adjust this prompt and make sure that it starts with what we want. Make sure it still starts with, I was on your site and saw you at the bar. Let's try this really quick. Yeah, overall, like, you'll see for people who are, I'd say, true super users of Clay, these sort of conditional formulas on almost every column are very, very, very important. The cool thing too, so we've got an output in this column and we've got an output in this column. The goal is to, let's say, push this to our CRM in our email sequencer as just one variable. The easiest way to do that, we have something called merge columns. So if we go to merge columns, this allows us to really just choose two columns. And so it's gonna first try to pull a value from company summary. And actually I'm gonna reverse this. So you'll see if this is ever run, it means that the sentence was too long. This will not have run if the sentence was under 15 words. And so when I create a new merge column, I'm actually first going to try a shorter sentence. And then if that is empty, I will pull the result from company summary sentence. Cool. And now I've got one column and I'm going to call this email first line. Cool. I spent a lot of time on the prompting. There's so, so much more you can go into here. And I'm also going to go ahead and send this guide in our chat. There's a ton of information on this guide. Probably don't have to read it all. But if you're really interested in becoming this stuff, This was written by Eric Novoselovsky. You may have seen him on LinkedIn or YouTube. He sends a million plus emails a month, generated with AI, his response rate to like over 10%. So this was written by him and has like a lot of his tips and tricks. I highly recommend checking this out. Anyways. Okay. So now I have filtered my, I found companies, I filtered my companies, and I have already generated my first line. The next thing, pretty simply, that I'm gonna want to do is go to find people, and I wanna find people that work there, and I want to find their contact information. So if we filter this down, let's look specifically for people that are in growth, marketing, rev ops, marketing ops. We'll see what this looks like. Now I'm gonna filter the location to the United States. And then also very important, especially if you have really large companies in your list or enterprise companies in your list, If you were to look for just like titles like marketing, there could be a hundred different marketing people at this company. And we will return all those people by default unless you add a limit per company. So I'm gonna limit this to two people per company just to make sure that I'm not sending and also spending credits on potentially 50 people per company. So I'll give this sec. Okay. We've got a preview here. We can check the results and These all look good. I'm actually going to remove associate. So we also have exclusion job titles. So I'm going to remove associate, intern, And I think that's it. I'll create people again. Okay, and then I'm gonna go ahead and click import to the table. So this is a new modal that we have. There's some import options. What this allows you to do, instead of having to create waterfalls to find this contact information yourself, you can just toggle these on if you know that these are things that you want to get. So if you know for sure that you want all the information from their LinkedIn profile or their work email, mobile phone number, et cetera, just toggle these on. And when you click continue, we'll import the people and we will automatically create and run whatever you select here. I'm actually not gonna use this for now because I want to filter down the list I have a little bit more. So let's click continue. And the reason I left this off as well, if you are planning to filter down your lead list at all, and you toggle that on, you'd be spending credits on people that you'd potentially filter out. So again, like on the topic of credit optimization, just make sure you're really paying attention. There's definitely, I've seen cases where people just wanna pull a list, they toggle it on, and then maybe they imported 5,000 leads of which most of them they plan not to reach out to. Just really be careful with those. It's a super helpful feature, but just really gotta be aware of what you're clicking there. Okay, so I've got these people now. The one thing I want to do to filter down this list a little bit more is I want to look for people that are relatively new to their job. You can go both ways with this, both looking for relatively new people or looking for people who have been in their role for quite a while. But it's also like a really nice and easy thing that you can reach out to about in your email. So, you know, if I'm looking just for new people, I could say, I see you're new to your job. I'm assuming like you might be wanting to come in and make some really impressive improvements. This is how, and then you can give a case study, this is how James did it within four months of starting their role at this other company. Similarly, if someone's been in their role for six years, So you've been there for a while. I'm sure you get pitched on these things all the time, but I imagine you're spending a ton of time on demos and you've seen this problem before. So anyways, to figure out if they're relatively new, let's go to Enriched Person from LinkedIn profile And let's run all of these rows. All right, and I know this has already been 30 minutes. I appreciate you all hanging on. I know it's not always the most exciting, but I promise we'll get to questions in just a second. What I wanna do, like I mentioned, is I want to see if anyone is relatively new. So if you click on any of the cells here, scroll down to latest experience. I see people actually miss this pretty common, so I really want to call this out. You'll see that there's experience and there's latest experience. Now let me find someone with maybe some more experience to show this with. Latest experience. Okay, perfect. All right, this person has five experiences. Zero is also going to be their latest but if you just want to easily access whatever their latest experiences, we already have that pulled out for you. It's called latest experience. I'm going to pull out their start date. And we've got all of these here. So, the first thing I'm going to do is let's just filter this filter this down to where start date contains 2023. Cool.
Okay.
So, already we've cut the list down. Let's say we wanted to filter this even more. Just another cool case where you can use formula columns. So we wanted to filter this to anyone who started their job, let's say in the summer or after that. You'll see in here we have the month. So what we can do is we can go to a formula column and we can say, pull out just the numbers between the two dashes in start date. Okay, And so you'll see we've got 2023-04-01. And so what this is going to do sorry, we're having a little issue with this sample outputs. But what this is going to do is pull out just a month. The one thing to call out here, so for the AI formula generator, what it's doing is turning text into a formula. Some people, if they were to try this, they might say pull out just the month from this start date. That could work, but we also have to think here, it may not necessarily understand that this data type is even like year, month, date. And so To make it extra clear, I'm just saying pull out specifically, I'm looking at it and saying the pattern is year-month-date. I'm saying pull out the number between the two dashes. If you were to say pull out just a month, it could work. It may not. If it doesn't work, I would just try it again and see what else could. But we could say pull out just a month from start date. Let's see how this does. But think of this formula generator similar to okay, So this did almost work. This is similar, though, to using chat GPT. It's an iterative process. You can go back and forth. It's not this type in this exact thing, and you'll get it exactly right. But I'm going to go back to this. I believe this may have worked better. Yeah, there we go. So then I'm going to add another filter. And I'm going to filter to where oops. So this is actually another very good learning moment. So if I go to try to filter to where this date is greater than six, when I click on formula, you'll see that I don't get like greater than or equal to. This is because if you look at the column type, it's a text type. What we need to do is change this to a number type to use number operators. If we change this to a number, go back to filter, and then we click on this column, you will see, perfect, we now have these different number types. So it's something small, but definitely important to pay attention to. Like places this could come up, let's say you have a number column and you were trying to use the formula generator to add text after the number. It may struggle with that because you're trying to put text into a number column. Anyways though, I'm just gonna say greater than four. Cool. And then for these leads here, the last thing I wanna do is find contact information. So if I click Enrich Data and go to, oops, Enrich Data and then go to Work Email, you'll see we've got two options, waterfalls or single providers. Waterfalls are kind of like what makes ClayMagic, So I typically will recommend using this. What Waterfalls do is it allows you to try multiple different providers to find an email. Our entire thesis essentially is why rely on a single database. You can work with us and get access to all of these. Also, just like a little, not super well-known secret, but a little secret. A lot of the larger data provider companies, all they do is they are buying data from these smaller providers anyways. And so what we are doing is we're giving you direct access to these smaller providers And it's at a lower cost because we're just connecting you directly to them. And then lastly, there's also transparency on where all these things are coming from. So if at any point you're like, okay, drop contact works so well for me, I'm just going to get my own API key. You can always do that. All you need to do is fill in the inputs. So we've got full name, and we have company and domain. And you'll see we are also running a validation provider on every column. And I'm going to go ahead and run this. That's good.
Oops. That's good.
There we go. Okay. All right. So it's queued up. And something to note here as well. So you'll see in just a second, this waterfall is going to start running. And you might be wondering where did those validation providers go? We hide those by default. So if you go to click on your column view, scroll down, you'll see that there's all of these validation providers. We have these by default, so it's not very confusing. What sometimes happens is people are confused because they'll see an email's found, but another provider runs. That's not an error, it's actually working correctly. The reason the second provider will run is because the email will come back as invalid. So if you're ever kind of confused around the results of your email waterfall, just double check here. Last thing to note. So if you remember at the beginning of this, we made a company table and then we made a people table. These two tables are connected and data kind of passes back and forth between them. So once we make our people table, you might be like, okay, like I wanna access that first line that we wrote. To find that, all you need to do is click on any of these cells in the company table data column. Hello? Hello? Sorry. So yeah, when you click on- Naomi, Is that you? When you click on the company table data, you're able to then add any of this information as a column. So I'm gonna add this company summary sentence result. And I think this is actually our merge column. Okay, And we could add has demo link as well if it's helpful. Cool. And this is the entire flow. The last step from here would be to image data and push this to your CRM or your email sequencer. But, yeah. I hope this was helpful. If anyone has questions, feel free to unmute or just drop them into the chat. And yeah, happy to chat a little bit more about them. Oh, Willie, for some reason I can't hear you actually. Are you talking right now? Sorry, still can I hear you?
About now.
I got you. You may hear I'm in the shot so sorry. A little bit in the background. So, I noticed that when I When I actually end up pulling people, and I want to do the import of this table into HubSpot, it won't necessarily, so you guys use full name, but then HubSpot uses first name, last name. Is there a way to automatically parse?
Yep, totally. So two things, one, if you've imported from LinkedIn, they're a little bit hidden, but if you click on any of these columns, you'll see we do have first name and last name already parsed out for you, and you can add these as columns. If you did not find these people on LinkedIn or for some reason it's not there, it's another place where you could use a formula column. If you go to formula and we could say pull out just the first name And then we give full name. And then this would be your first name column. You would create another column to pull out just the last name. But you'll see, yeah, it's pretty solid. Cool?
Yep, awesome. That's good. Cool.
Yeah, Of course.
Is there a way to create your own waterfall suite?
Not currently. We are kind of working and playing with the idea of this. What I will say is if you click Enrich Data, and let's say like, Okay, so we have this thing called recipes here. If there's ever something that you use a lot and you want us to turn it into a recipe, you can do it yourself right now, but you can always reach out to us in Slack and just be like, Hey, I use this all the time. Can you make this recipe? And we can totally add that as something that you can use.
Right on. Is BARD something that you can also use as a generative AI?
Not right now. Yeah, unfortunately, not right now. We are looking into other LLMs that we can potentially integrate with. Right now, it's mainly open AI. Some reasons that I think are positive for that. One, we have a good relationship with them. So the rate limits are really high, which allows you to do this at scale much faster. With that being said, though, We've definitely heard a lot of people requested, so we're testing out some things.
Yeah, they just released a report where like, and I don't know to what effect it would have for prospecting, but essentially like GPT-4 versus like their Gemini version is like better. But honestly, like for what we're doing in terms of looking for information on a website or things like that, I don't know if it makes any difference.
Yeah. It's something, honestly, it'd be interesting, like if you ever seen bad results and you test it manually with Bart, let us know and we can definitely look into it. I am unsure if they have something we can even connect to right now, but Colin mentioned in the chat too, you can use Anthropx Cloud AI. You have to set up the integration yourself. If you're interested, just ping us in Slack and we can help with this. But it is another option you can totally go for.
Yeah, just like add a bit of color there. You can like use any arbitrary, you know, service as long as they have an HTTP API, but we've already built out sort of presets for using Anthropic. So yeah, but Matthew's doing right here, HTTP API, browse templates, if you go up to the top, you'll see the template right there for using Cloud AI. So you just click that top one. Yeah, and you can select like the prompt you want to put in, and you have to input your API key, but it'll work.
Super rad. It more than anything, it's just for experimentation.
Thank you, Colin. Yeah. Any other questions or like thoughts? Are you wondering like, is this even possible to do in Clay? Happy to answer anything there.
I had a real simple, quick question. I put it on Slack, but just maybe real simple answer might work here. Is there a way to, I use a lot of views. Is there a way to sort the views on a table in the desired order? I couldn't figure out how to do that.
It's a good question. Unfortunately, not right now. It's something that we have heard a lot and for anyone not familiar, views are here. Yeah, it's something we've totally heard a lot And we're working a lot on the table UI, especially over the next month or so. So yeah, we'll definitely keep that in mind. I'll make sure that the team also hears that feedback. Nice.
Appreciate it. Yep.
I got one a little bit more off the wall, but I've also been experimenting with the Google extension. And I was wondering, when I go to G2 and I look for example, sales enablement companies, it'll scrape like the first page, but in order for it to do a combination of all of the list, I have to go to each individual page and kinda like sift through it. Am I doing it incorrectly?
No, you're doing it right. So it's just a limitation of our scraper right now. We don't have any auto pagination, which is helpful because like, you'll never get blocked as a bot. It is unhelpful because it means you have to do it manually. It's something that also is on our list in terms of things for us to improve. Like we honestly, we haven't touched it too much in the last year, but we have some big plans hopefully in the next few months that you'll see in terms of.
And then like when it does actually, it does pull like at least the name of companies, which is great, right? Cause it like, it's super targeted as a list that I want to get after. But what it doesn't do is it doesn't pull in like, like thermographic data. So I don't know what to do after I pull the list.
Yeah. Is the data on the page that you're trying to scrape it from?
Can I give you an example?
Like yeah, that's perfect.
So actually right there, G2 Sales Enablement, like that, right there, like you click on that.
Okay.
It gives you 126, right? And that's great. But this first page will show you maybe things like 10 of them or 30.
Okay.
Okay, that's great, right? When you hit the, on the actual extension, when you hit the see all rows section.
Yep.
That That's the columns that are to pull. And all I, I think all I need from there would be like either a LinkedIn page or something. I don't know what to do from there.
Yeah, that's a good question. So one thing you could do, if you just use this prebuilt recipe within Clay, you can run the company domain enrichment, which pop in a company name, we'll find a domain. Once you find the domain that unlocks pretty much everything else, right? You can emerge the company based off the company domain. You can find revenue, fundraise, et cetera. So one option is to use this domain enrichment.
The other option- I missed that. Sorry, I was running out.
No, you're good. Yeah, it's just this domain waterfall.
Oh, OK. Yep. Got it.
The other option, if they actually link to, I'm not sure if g2 actually links out to the website here. Yeah.
They will on the company page, but they won't on the list.
Got it. Yeah, then in this case, the easiest thing to do is just use the domain model to find the company domain from the company website. Or company domain from the company name, yeah.
Okay. Does that one use credits? I don't know exactly which ones use credits.
Yeah, so there is, this uses two. The first is free. It's with Clearbit. It's, I'd say, a little bit less accurate than Google, which is why I put Google afterwards as well. But you could just remove Google if you wanna really conserve credits and run only with Clearbit.
That helps by itself, that's awesome.
Yeah, cool. And yeah, like Colin mentioned too, you could use scrape website on the page. The only thing there is you will have to run another integration. So it would be a little bit more costly, but you can scrape website based off of the G2 page that we pull out, which would be which we could add in. It's not currently being pulled. And then yeah, you can pull out any information from there as well.
Yes, because I'm thinking of using this, not just not to use but I like maybe like why combinator or like, yep, Gardner pure insights.
Yeah. People use it for like any of these specific like VC directory pages. Y Combinator we have pre-mapped luckily. If there's any that are not working, let's see, out of the detected list. Yeah, so for this one, same thing. They don't have the domain here. So you just need to do company domain data point or waterfall again. Cool. All right. Any other questions from, yeah, open to anyone. All right, Adam, let's see. Is it possible to share tables? And to answer this first one, currently it's not, but we have a product that's like in the final stages that will allow you to do just this. So this should be coming out super soon. It'll be really nice. You'll be able to see tables that people in the community have built, share it with other people that are not in your workspace. So, yeah, stay tuned for this one. It is currently in progress. As far as finding the work emails, what are the best providers? I haven't seen mixed results from recommended. Also the recommended set seems to use a large amount of credits, so order. Cool, that's a good question. So whenever you're using this work email Waterfall, first thing, we already have it ordered in terms of what we've seen the highest quality results with. We've actually done thousands, we've put thousands of emails through each of these providers to figure out the accuracy and order this. I will say it can be dependent on the nature list that you're putting in. So if you're seeing bad results with one specific industry, you may wanna switch this up. But overall, this is already ordered in terms of what works best to least. The other thing to note here, definitely in terms of it being expensive, if an email is not found, we do refund it. So you're not being charged for emails that are not found. And so like, let's say an email wasn't found for four out of the five providers, you will not be spending eight credits on the first four plus two on the last. If an email is found and it's invalid, we currently do still charge because we're still technically spending money finding that email. We have plans in the future to refund in that situation, but it's gonna take some technical work. But let me see if I fully answered your question.
Yeah, hey, sorry, there's some noise back here. Sorry, that's why I typed it. But yeah, no, thank you for addressing my questions. The I guess I'm just curious. I mean, the fact that you had mentioned that you guys have done iterations now and that you found that this is, I guess, quote unquote, the best way of going about this. And you said that it can be industry specific. So if we're like, we've noticed that we work in education. So I have noticed some spotty emails as far as definitely not the correct emails. And I'm just, I'm trying to figure out the best iterations here. If I should, you know, remove Prospeo and just move directly to Hunter, for example, and then have the validation provider, right. And something like that, something more simple direct, but you guys have obviously, you know, battle tested, you know, this extensively. Right. So, I guess I'm just, I'm just curious around, like, have you, have you come across other people using Clay in different ways? Is it common to change this up and in an effort to have better results and save credits at the same time?
Totally. Yeah. It's a really good question. We absolutely have seen people change this up. The way in which I've seen people find their own custom orders is essentially to run all of the providers with no waterfall. So just for, let's say like 20 or so rows where you know what the email is already, run all of the providers we have so you can fully test them yourself. And then you can change things up as needed. So like for example, we have some companies who, some companies who are going after like small SMBs and you know, things like nail salons or like these like IV, liquid IV businesses and for them like Nimbler and ICPS work specifically well. It's not like Nimbler, I haven't seen work amazing for all the cases. But for some use cases, they just have better coverage. So yeah, like Bob was saying, I would probably test them all one by one for a data set that you know, the answer to, and just do like your own experiment there.
Interesting. Yeah, no, that's, that's good to know. I've I've basically playing around with this whole bunch and trying to figure out what, what sticks. But, the fact that other people have been able to find success by changing this up and testing it out, I think makes sense. But yeah, I appreciate the clarification.
Yeah, yeah, absolutely. And also let us know what you find too. Because another thing that we can, we've, we've thought of before, but we haven't found like enough. We haven't tested enough of these different issues to actually do is right now, like this is our overall recommendation. If we get to a point where there's such a clear distinction between different categories that it makes sense to have other recommendations, we could have separate default options, right? Like if you're in the education specifically, education industry, and you just know that, let's say Prospero doesn't work, but Snove does, we could potentially make presets for those different use cases. Let us know what happens.
Absolutely. Makes sense. Thanks, Matthew.
Yeah, absolutely. All right. Yeah. Any other questions? Anyone's welcome to pop in.
The language in the formula generator, what is it that it's using?
It's outputting, I believe, JavaScript. Colin, correct me if I'm super off there. But it's like a version of JavaScript. I don't think it has access to everything that you could write yourself.
Yeah, it's like almost JavaScript. Like I'd say like 90% of what you write in JavaScript will work, but it's not exact. And like, we know it would be so much easier if it just did JavaScript and it's like a challenging technical problem. So like, mostly, but I wouldn't, yeah, I wouldn't, I wouldn't pay something in from like a JavaScript online generator and expect it to work 100% of the time. Yeah, it happened.
I worked in iPass. So like, we had like a flavor at meal soft, for example, it has a flavor of JavaScript, right, it's called data weave. So I know exactly what you're saying. Well, well, Hey, Matthew, Colin guys over at clay. Thank you so much for the help. I'm starting to dive in and learn. I'm gonna kill you, but thank you so much for the time.
Yeah, absolutely. And if anyone ever has topics that they want us to cover in these like live webinar type sessions, just ping us in Slack. Always happy to do so. But yeah, everyone have a good rest of your day.
Setting the stage for your recording...