An increasing amount of attention is being directed towards influencer prospecting today. As Tomoson’s study claims, influencer marketing is now the fastest growing customer-acquisition channel. A lot of brands are interested in partnering with influencers because of their ability to inspire action in their network. Statistics have shown that close to 50 percent of today’s currently active Internet users rely on influencer recommendations.
But, different products and services, as well as different campaigns, will require different kinds of influencers. If you want to make the most out of your influencer collaborations, you should look beyond just the number of followers.
Instead of people who just have big networks, your attention should be also directed towards medium-sized influencers who are known and respected in specific industries (relevant to the products or services you are trying to promote).
Apart from gaining visibility and engagement from particular groups of people, a finely targeted micro-influencer campaign has the power to turn entire businesses around, depending on what they want to achieve. The only problem is – micro influencers are often quite difficult to find. Especially those who don’t speak your language.
Today, I’m going to share with you how my agency, Four Dots, resolved an urgent task to locate an impressive number of influencers. In this article, I’m going to take you step-by-step through the entire process, so you can learn how to easily replicate it on your own whenever you need to find new influencers:
The Task Objective
We needed to locate 2000 influencers who have a noticeable following in the Lifestyle and Auto industry, from German-speaking countries, specifically, Austria, Germany, and Switzerland.
Our goal was not only to locate and group the prospects in one place but also to collect as much data as possible on these influencers, which, additionally, meant organizing them under different criteria.
We gave ourselves a tight deadline. It was imperative for us to see if we could deliver the mentioned number of influencers from these specific niches and markets in record time, along with the additional information, referred to in the paragraph above.
Apart from that, the biggest challenge in collecting and sorting this many contacts and accompanying data originated from the fact that nobody in our team fluently speaks German. We also had no prior experience in these markets, and the DACH auto industry wasn’t a familiar niche to any of us in Four Dots. The task required a detailed analysis of both markets and an exhaustive data mining strategy.
The End Result
Using Dibz and the process we’re about to describe, we were able to locate and sort exactly 2138 influencers for both major targeted categories (Auto and Lifestyle).
From German, Austrian, and Swiss auto industry – our agency prospectors were able to collect all the required data on 339 bloggers. We organized the data into two groups, defined by the number of social media followers:
- Group A – bloggers with more than 50k followers on Instagram, Facebook, and Twitter
- Group B – bloggers who have up to 50k followers on Instagram, Facebook, and Twitter
At the end of our data mining, Group A had 47 bloggers, while Group B was a bit larger, counting precisely 292 bloggers.
The same principle was applied to DACH lifestyle industry, where we were able to find all the required data for exactly 1799 bloggers. Bloggers from this market were sorted into 3 groups, instead of 2 like with the auto industry, because the data allowed us to define an additional level of influencers:
- Group A – bloggers with more than 50k followers on Instagram, Facebook, and Twitter
- Group B – bloggers with 5k to 50k followers on Instagram, Facebook, and Twitter
- Group C – bloggers with less than 5k followers on Instagram, Facebook, and Twitter
Group A counted 188 bloggers, while B and C together contained data on 1,611 influencers.
For an agency that doesn’t have any prior experience in DACH markets, nor a person who is fluent in German – getting this much data on this many individuals was a huge win for us.
We started with manual data mining. The first step of our strategy to find 2000 DACH influencers was to look for adequate bloggers/social influencers that met our desired criteria.
The process has been systematically and thoroughly streamlined in order to help us find as many high-quality prospects as possible. After conducting a detailed keyword analysis, we began looking for bloggers.
First, we started manually searching the Web for influencer prospects, just to get the feel of the situation. It was a bit of an experiment to help us get a sense of how long it would take someone to manually find this amount of data. We quickly realized that this method was consuming a lot of our limited time. Then we decided to up our game and automate every possible part of this mining process.
From that point on, much of the work in this department was done using our prospecting tool Dibz, which helped us collect a lot of valuable data we needed for this project. Even though Dibz is primarily a link prospecting tool, it can also be used for different types of data mining, such as, in this particular case – influencer research.
This, of course, brings us back to the “the tools is only as good as its users” article we published on Four Dots’ blog a while ago. Like many tools out there, Dibz is a powerful piece of software, but its use and value is directly dependent to the user’s ability to put it into action. If you understand its power and functionality, Dibz can assist you with a wide variety of tasks.
The tool made it possible for our agency prospectors to quickly search the Web using specific predefined parameters. Our experiment showed that Dibz was extremely effective in the long run and really got to shine when our staffers were fresh out of opps they could find by hand.
Here’s how the entire process looked like:
The Step-by-Step Rundown
Let’s start with the requirements. The task was to extract the following data:
– URL of the blog
– The first and last name of the blogger
– Blogger’s email address
– Which CMS was used to create the blog
– Number of Facebook, Twitter and Instagram followers
Considering we didn’t know which blogs to analyze or what the blogging scene, etiquette and standards were in these regions, we started the search by manually entering queries into the search bar and taking notes. After taking a look at some of the best-ranked sites in the areas that were of interest to us, we compiled a list of relevant keywords, common URL footprints and in-text phrases and were fully equipped for the automation of the process.
Before we proceed to explain how we used Dibz to find more than 2000 influencers, let us first describe what doing this manually looked like.
After typing in a query, and selecting German in the search settings as the language for the results, we’d have to open a result, determine whether it qualifies (remember, people, doing this are by no means fluent in German), if it does, we’d copy the URL to a spreadsheet, and start looking for other info.
This involved opening each of the relevant social media accounts in a new tab, checking and finally recording the number of followers, for each of the three observed social networks.
Contact email and bloggers name would, ideally, be found on the ‘impressum’ page, however, looking for either could, with some websites, occasionally take much longer and sometimes ended without us ever finding complete or reliable info.
Even checking the country wasn’t always as simple as it may seem, as most sites that didn’t have a country-specific TLD, but instead had something like .com, and weren’t always too transparent about their geographical focus or origin.
Finally, determining which, if any, CMS the site being examined was using was done with the help of a Chrome extension, and along with finding the blogger’s name, is the only part that we didn’t manage to automate, but that was still much simpler to do in bulk, with the list that our link prospecting tool provided.
All in all, it took anywhere from 5 to 20 minutes to find all the info about a blog, or decide that it’s going to have to remain incomplete. Multiplying that by the required number of results, i.e. 2000, we get a projection of around 166 to 666 hours (we swear this is not our attempt to associate manual labor to the Dark One, or to contradict proverbs vilifying idle hands), and that’s only if we assume that direct Google search would keep yielding valid prospects at a consistent pace – which we know is never the case when you get into larger numbers.
Another reason why this is a conservative estimate is that it fails to account for all the sites that would need to be inspected only to be disregard. Even with several people on the task, this was unacceptable in the time frame we set for this task. That’s when we turned to Dibz.
If you also want to be able to use our software for influencer prospecting, even without being all too familiar with their language or niche, here’s how we did it:
Initial research, with manual prospecting, led us to a range of keywords, that we entered into the search term field, each in a separate row. For the purposes of this demonstration, we’ll just use three of them – “fahrad” (bicycle/bike), “auto kaufen” (buying a car) and “wagen” (vehicle).
Since we knew we had to classify the blogs according to the country, the next section, “Custom Search Parameters” was used to specify the type of websites we were after, as well as their TLD. We did this by entering 3 parameters, each of them to be combined by our tool with each of the three keywords listed above.
The parameters were:
This did mean that we’ll eliminate all the websites that would otherwise qualify, which have a different TLD (like .com or .net), but it also allowed us to sort the ones we did find easily, and to explore the three options with great depth.
To summarize, at this point, our prospecting software is instructed to find all the websites that could turn up for any of the following queries in a typical search:
inurl:blog site:.de fahrad
inurl:blog site:.at fahrad
inurl:blog site:.ch fahrad
inurl:blog site:.de auto kaufen
inurl:blog site:.at auto kaufen
inurl:blog site:.ch auto kaufen
inurl:blog site:.de wagen
inurl:blog site:.at wagen
inurl:blog site:.ch wagen
After selecting the client and campaign, we set the language to ‘German’ and TLD to ‘Global’. If the campaign had required data from just one country, for instance, Germany, we could have simply set the desired extension (.de) at this point, without having to use the custom search parameters from step 2. Date range was set to ‘Any time’ and depth to 100, since we wanted as broad a search as possible.
This completes the instruction that the software is given. It’s been told to: conduct all the listed searches; find up to 100 results in German for each of them – without filtering according to the date of publishing; combine them and remove duplicates; let the internal spam filter (which users with higher levels of authorization can finely tune) take a look at the results and remove those below a certain threshold; and finally, extract metrics for each of the results – on page and domain level alike.
Dibz data export provided us with an easy way to sort results according to country – by filtering the Search Operator column, or sorting it alphabetically; result URL; social networks followers, and, where possible, contact email for the site in question.
Before sorting the data, we created several more similar searches – using different keywords, and specifying different, previously skipped TLDs.
Once we had a sizeable list, we sorted the data and considered the gaps we were left with. We had an incomplete date in the ‘Country’ column since only the described TLD setup brought back results which can be easily sorted; didn’t have any of the bloggers’ names; had nothing on the CMS they were or weren’t using, and only had email addresses for some of the results.
On the other hand, we did have a huge list of likely prospects, their backlink data (just like spam value and Domain Rating, not required for the final list, but useful when assessing the value of a website), and all the social networks info, which would have been one the biggest time wasters if we had to get it manually.
Since it was obvious and understood from the start that, regardless of how it was found, each website would, at one point, have to be directly inspected, checking which CMS a site was using was done on the go, with one click on the extension and a copy/paste.
Since we already had the list prepared, we were able to speed the process up by opening sites in bulk, checking whether they’re still active and whether they meet the quality standards and getting to the contact page – when necessary (bloggers often had a little sitewide ‘about me’ section, so if software was able to find an email, it was obvious whether the email was the right one, and what the blogger’s name was).
The Bottom Line
So let’s add these hours up. Coming up with query combinations, exporting and sorting the results may have taken a total of 5 hours. With all the data already available; quality indicators like Spam Score and Domain Rating; and being able to open sites in bulk, it took about 35 hours for one of our team members to go through the list, evaluate the websites, and enter the missing data. Totalling 40 hours, and resulting in 2138 results, this comes down to 1.12 minutes per link.
Once again, this doesn’t take dismissed sites into account, but just to mention that, not only were there fewer websites to skip, as Dibz allowed for a more specific search, but it took less time to decide that a site was subpar or unrelated; so even in this regard, Dibz saved us quite a bit of time.
If we’ve had to conduct a more extensive research, time savings would be even greater proportionally – setup time might remain constant – in our case 5 hours, but for the actual part involving going through the sites themselves, the more of them there are, the more time you’ll be saving by using our prospecting tool.