The ever-elusive psychographic/behavioral list. It's easy to build a demographically-defined list (location, job title, company type, etc). Not so easy to build a list that includes people with certain beliefs, or who have had certain experiences. And yet, I keep trying to figure out how to do the latter. I've had some success...
The demographically-defined list
I have always cared a lot about list quality. When I started building lists for clients, I started with the "tight & clean" approach:
- Use LinkedIn Sales Navigator (LISN) Advanced. Reason: LinkedIn has fresher data than Apollo et al, and The Advanced tier lets you upload CSV files, allowing much easier creation of account lists which let you scope people lists more easily.
- Process:
- In LISN build a search that results in a list of companies that employ the people you want on your list.
- Use Phantombuster to scrape that LISN search. Turn the Phantombuster scrape into CSV files of 1,000 companies each.
- Import each CSV into an Accounts list in LISN.
- Use those Accounts lists to scope a LISN people search such that your people search yields exactly the kind of people you want on your final list. Use LISN exclusion filters heavily to reduce noise.
- Use Evaboot to scrape the people search and get email addresses for those people.
This is a slow, difficult process. Over time, I came up with something better, the "refine a big, noisy list" approach:
- Use the cheaper LISN account. Reason: you don't need to upload CSVs to LISN using this approach.
- Process:
- In LISN, build a people search that results in the people you want on your list.
- Use vayne.io to scrape that LISN search. Vayne can multiplex a search across (I presume) a fleet of bot accounts they run, meaning you can scrape way more quickly and with less risk to your LISN account than Phantombuster poses.
- The scrape you get from Vayne will be noisy, but it will contain the subset of people you are looking for. It takes a bit of code to clean the list up, but either a pivot table or an LLM looping over the big noisy CSV or a combination of both can get you a squeaky clean list with considerably less effort, delay, and cost than the other approach.
- Use anymailfinder.com to look up email addresses for the cleaned Vayne list. Use Zerobounce to further validate the email addresses that anymailfinder.com gives you, and use scrubby.io to validate the catch-all email addresses.
The psychographic/behavioral list
I've got nothin shippable, just some experiments I've tried. :)
It's worth mentioning that there will be good proxies for certain behavioral characteristics. Ex: Most people who have been a manager for multiple years are likely to have had to fire one or more employees. Building a list of people with the behavioral characteristic of "has had to fire someone" isn't going to be that hard because having manager in the job title -- a demographic characteristic -- is a good proxy for the behavior we're interested in.
Phantombuster does let you scrape Facebook and LinkedIn groups. So does Apify, which I've moved to for most of my scraping work. I experimented with scraping a Facebook group once. IIRC the group was for people dealing with anxiety disorders -- a great example of a psychographic characteristic. The scrape went great. It yielded maybe around 10,000 people? Then I tried to convert that big beautiful list into email addresses and got asymptotically close to zero email addresses. If it had been a scrape of a LinkedIn group, I'm certain that closer to 50% of the list of people would have successfully converted to email addresses.
Last weekend I built a proof of concept for a prospective client. They wanted a list of companies that are likely to buy a certain kind of firm in the future. Since we don't live inside the world depicted in Minority Report, the next best alternative to find:
- Companies that in the recent past have bought that certain kind of firm. They may buy another one again.
- Companies that ordinarily buy a range of firms that include the certain kind of firm in question.
This resolves to a scraping job followed by an LLM filtering job.
So... the best I've been able to to come up with for building a psychographic/behavioral list:
- A demographic proxy
- Scraping a LinkedIn group that's likely to have members with the desired characteristic(s)
- Scraping the open web, or search results from the open web, and then filtering those results through an LLM.
Maybe I'm underplaying my ability to build a psychographic/behavioral list a bit. If so, it's because the inherent difficulty of this kind of list-building is greater than the difficulty of building a demographic list. But if the quality of the outreach that I'm on the receiving end of is any indicator, most people aren't even trying at all to get a relevant list. So what I consider a minimum bar for quality might be quite a bit beyond what most other outbound marketers are attempting?
Anyway, sorry for the lengthy writeup; I wanted to do a bit of a checkpoint of where I'm at with respect to list-building expertise. Maybe some of this will be helpful to others.