How I bootstrapped a 59x ROAS e-commerce cold audience in 8 weeks on Instagram
Bootstrapping an e-commerce audience on social media is hard, especially a niche one. Facebook self-serve advertising platform is a breakthrough for any business looking to bootstrap a paid audience, however, the targeting options to start with are very limited:
- You can target your competitors’ audience if they’re big enough. Chances are if your competitors big enough for Facebook to let you target them, then you’re already far behind in sales.
- You can use your own sales data and email to create Custom Audiences and Lookalikes. While this may perform well, it requires you to have sufficient historical data, which we all lack at the bootstrap phase.
- You can guess your Ideal Customer Profile (ICP) using Facebook demographics, interests and behaviors targeting. Such classifications however generally lack granularity and can easily be irrelevant (just check out your own Ad Preference to see for yourself).
Interest-based targeting is even more depreciated as Facebook Pages aren’t Facebook priority anymore, limiting its organic reach and therefore brands do not invest on it anymore. Brand communication is concentrated on Instagram and key targeting data is there, with thousands of highly targeted niche accounts matching any e-commerce segments. Most marketers would kill to be able to target an accounts Instagram’s followers if only Facebook would allow it in its Ads Manager.
There is a workaround to making that dream come true. Facebook lets you retarget people that interacted with your own Instagram Account. Using automation scripts, it is possible to generate account traffic from highly targeted accounts to your Instagram account which, you guessed it, we then retarget through paid advertising.
The following method will allow you to bootstrap your own Instagram based ICP leveraging Instagram scraped data. After a few weeks, Facebook will have enough engagement data for you to create a lucrative Lookalike of this audience, leading to a much broader reach corresponding to your niche’s interests.
Embark in our 8-weeks journey to creating a 59x ROAS e-commerce cold audience. We’ll use the example of an auction marketplace doing 150k€ GMV/month seeking to expand into a new niche: limited-inventory Urban Art auctions.
Step 1: Selecting your Source Accounts
The first step to creating your audience is to list all the Instagram accounts you believe have the right audience representing your niche (Source Accounts). In our example, we’ve targeted 77 accounts across different Urban Art artists’ personal accounts featured in the auction as well as well-known collectors, key opinion leaders, dealers, and galleries.
Tip: to make sure you are curating the right Source Accounts, pay attention to the level of engagement (average number of likes/number of followers) of their audience. Start with accounts that have above 5–10% engagement rate.
Step 2: Gathering data
Next, we collect every like on each identified Source Accounts into a single aggregated list. Using a custom script (note: reach out if interested in knowing more about the script) ran from your computer using Node.js or on the cloud using Phantombuster — thanks Guillaume and Martin for letting me use your great platform — we’re going to extract every single like in each post.
In our case, we’ve collected a little less than 21 million likes over 49,606 posts across the 77 targeted accounts.
Aggregating all the likes across the Source Accounts will result in a highly precise list of your desired audience as shown below.
You’re half-way through! You now have the golden list of highly targeted accounts that love your niche. Let’s work in getting in touch with them!
Step 3: Generating traffic to your Instagram Account
Remember how Facebook lets you retarget people that interacted with your own Instagram Account? Let’s have these highly targeted accounts interact with you, by having them either consuming your content or visiting your profile. The minute they do one of the two, they’re hooked, and you’ll be able to retarget them.
To do so, we used another automation script run on Phantombuster that automatically follows/unfollows and likes some content of these accounts. People can’t help being curious about who wants to follow them. Them receiving several notifications from you will results in 60% them visiting your profile and 15%-30% following back (these metrics will depend on your account’s name, your content and successful targeting).
Tip: people with a high number of followings and/or followers are probably professionals and may be less receptive to your ads. If you’re looking to target individuals, try sorting for the following accounts:
- No email in the description
- Number of followers between 80–1200
- Number followings between 80–800
Eliminating accounts below 80 followers/followings will single out fake bot accounts, while accounts above 1200 followers or 800 followings will probably be professional accounts.
If you are also aiming at growing your followers base, you will have a higher follow-back rate with this set-up.
Warning: Instagram is against automated actions on their platform. However, it is a widespread practice and Instagram seems to tolerate it (so far!) if you’re not pushing it too hard. On an aged account, a reasonable setting is to reproduce a human behavior. This means doing about 60 follow/unfollow per hour, with 3–5 likes per account, only during work hours (9am-7pm) and work-days (Monday to Friday), preferably from the same IP. Going above may risk your account.
If you do not feel comfortable using your official account, you can create a dummy account to retarget the traffic to, protecting your brand, but not leveraging the follow-back potential to increase your follower base and free brand awareness.
Step 4: Creating a Retargeting Audience
One week of automated interactions should result in 1500–2000 accounts in your audience that you can start retargeting to but I suggest you wait at least three weeks to cast a wider net before creating your retargeting audience and especially your Lookalike audience
To set-up your Retargeting Custom Audience, go to your Facebook Ads Manager > Audience:
- Click on “Create Audience” => “Custom Audience”.
- Select “Engagement”, last option in the list => “Instagram business profile”.
- Select your account and set up to “Everyone who engaged with your business”. Adjust the number of days to the timeframe you started automated interactions.
To set-up your Lookalike Niche Audience, go to your Facebook Ads Manager > Audience:
- Click on “Create Audience” => “Lookalike Audience”.
- Source: “Other Sources” => Select your Retargeting Custom Audience.
- Select Locations and 1% Audience Size.
Facebook will populate this Lookalike audience based on their social graph understanding of your source Retargeting Custom Audience. Remember, this is an audience we have identified as highly receptive to Urban Art, so the Lookalikes should be receptive to it too.
Step 5: Retargeting
To validate our method, we set out to create one creative, placed on Facebook Feed placement, duplicated across different test audiences:
- Manual Artist Audiences: geolocalized audiences (France & United-Kingdom) created based on the interests of well-known Urban Artists. While this technique may sound good, Facebook only allows you to target artists with a big audience, increasing the possibility of noise in your targeting. Many of the artists represented during the auction weren’t available as interests.
- Instagram 60d Audience: Retargeting Custom Audience created in the previous steps.
- Lookalike Audiences: Audience created based on the Retargeting Custom Audience.
As mentioned before, our tests were run on an auction website selling limited art editions. Inventory for the first test was constrained to 84 art pieces and each art piece is sold only to the highest bidder.
During our test, our goal was to maximize 1) ROAS and 2) Website Payment Info form fills.
The ROAS results speak for themselves:
When a member adds a payment info, it means they are participating in the auction (we require them to have their credit card set before auctioning as a precaution). While an audience may not end up buying because of our scarce inventory, them participating in the auction drives up the final price of the item. As a result, our marketing KPIs for us are as follows:
- Leads &Website registrations completed: catching their email is key to future auctions
- Website Payment info adds: this means the lead is participating in the auction, driving up prices.
- Website Purchases Conversion: the sum of the total price paid by the highest bidders of items
Given our KPIs, we can see that Retargeted Audience and Lookalike Audience outperform the interests-based targeting (Manuel Artists). We started our test with a 15€ budget per audience and doubled down on the Retargeting Custom Audience after seeing it performed so well.
Was this a one-off? One month later, we ran a second test, this time dividing equally the budget for a more accurate A/B test results:
The Retargeted Instagram Audience outperformed even better than the rest as we had a wider audience of 15.000 people compared to the 6000 people audience for the first auctions, thanks to one more month worth of interactions with the account.
Other than pure ROAS focus, based on Website registrations and people participating (i.e. Website Payment Info) on the auction and driving up the final price, Lookalikes from our Retargeted Instagram Audience performed better than traditional manual interests-based audiences.
Kudos to Facebook’s social graph for understanding well the mutual interest in our Retargeted Instagram Audience.
Conclusion and afterthoughts
Early results show that through this technique, you’re able to generate audiences that are 4x more valuable than using the manual interests based targeting and generate more leads. Lookalikes audiences based on this technique also performs well as registration and lead capturing shows.
An additional benefit is the brand awareness and Instagram Followers growth as you follow/unfollow 600–800 people a day. While harder to track, your referral traffic from social media and direct traffic will increase. During our test period, we saw a 300% increase in Instagram referral traffic.
Bear in mind, this is free traffic.
This technique is also an excellent supply acquisition source paired with Facebook Lead Generation campaigns. The Instagram Retargeted Custom Audience above generated in 4 weeks 52% of our inbound inventory deal value at a 99x Inventory Deal Value On Ad Spend. (note: this may deserve a separate Medium Story. Let me know if interested).
Website Traffic (people that visited the website) may have generated a better ROI but the results are flawed: the Retargeted Instagram Audience should also have visited the website. As such, because Facebook applies a last touch attribution model, the website traffic audience can include people from the Retargeted Instagram Audience.
There are however several limits to this technique and our tests:
It’s a niche approach. To stay below Instagram’s bot detection, you can’t grow your Instagram Audience more than 1500–2000 people per week. You could accelerate interactions by duplicating dummy accounts. Note also that this audience is highly qualified and the Lookalike Audiences you can create based off of it can be lucrative. We have yet to test it out for a bigger revenue goal and ROAS.
It still requires great content. We generated a lot of Instagram profile visits because our accounts name was obviously related to the niche(WEBSITEurbanart) that organically converted to traffic because our content was good. We had a very good ROAS because our inventory was great and our ad creative was well-crafted.
It isn’t free of risk. As previously mentioned, Instagram’s ToS doesn’t allow automation but Instagram seems to tolerate it. Many businesses have automated their growth on Instagram and the downsides of botting are exaggerated. As long as you don’t abuse it (below 600 interactions per day), Instagram won’t ban you nor limit your reach.
It requires an automation platform. Scrapping all contents can take a long time if you’re using only one computer (and one IP address). You can do it without Phantombuster’s cloud-based automation platform (or any other similar tool for that matter), but to achieve results fast, Phantombuster or other similar tools are almost necessary.
This technique was tested on an auction business model, with limited inventory, which limits the number of transactions. This technique can work even better with niche e-commerce free of these burdens.
Interested in testing it out for yourself? Reach out to me. I’d love to see it perform on other models :)
PS: growth hacking is great but do not forget the basics too — website traffic retargeting ❤ 221x ROAS