ScentBird Takes Flight: How We Used Facebook Automation to Increase our ad spend from $30k to $400k
I am a Head of Marketing at aitarget.com — AI-powered Facebook Ad Optimizer. Our goal is to provide Facebook Advertisers & Growth Hackers with technology to scale businesses significantly and educate them how to apply it.
We made an effort in creating something undoubtedly helpful for our audience. I’d like to share the recent success story of the aitarget.com customer below, so the story is not mine. Our customer is Scentbird known as “Netflix for Perfumes and Colognes.” This start-up participated Y Combinator program in 2015. I hope you enjoy their advice and strategy to scale sales and revenue described below.
My name is Oleg Popov, and I’m the Head of User Acquisition at ScentBird.
Basically, I’m responsible for growing our customer base.
Our business model is similar to the Dollar Shave Club; we sell subscriptions to get a 30-day supply of any luxury or designer perfume every month. For a monthly subscription of $14.95, our customers get a great 0.3 oz fragrance from 450 available products in stock.
The Start of Our Growth Journey
I joined the team January 2015. At that time, we had just started looking for scalable marketing channels. We were also fortunate enough to be part of Y Combinator Summer 2015, which meant the funding was available to invest and scale- but we needed to make wise investments.
Facebook was showing some potential, but we couldn’t spend our budget effectively, while also controlling our Cost Per Acquisition (CPA). We were able to spend around $30,000 USD for our Facebook ad media buy per month.
One of the first changes we made was to switch from banner to video advertisement. After two months, we doubled our sales while reducing the CPA. Then we discovered the slideshow format. Both slideshow and video become our main ad content types. By June 2016, we had increased our Facebook ad spend to $100,000 USD per month.
There were two of us in the company who ran campaigns on Facebook: a hired manager and me. We both experienced similar frustrations:
● Most of our actions were routine, and we thought we should be able to automate them by setting rules.
● Facebook works pretty unstably, which meant we couldn’t just set up a campaign and then create more new ads in there. You have to tweak it constantly.
● The process was quite time consuming. If you’re running dozens of campaigns, it quickly becomes a nightmare.
● The more we spent, the more data we collected during a work day, which meant we can make more decisions during the day. And it takes time.
Partnership with AITarget
There are many solutions on the market: Nanigans, AdEspresso, Automated, Smartly, AITarget, etc. We considered and tested automation systems of AdEspresso, Automated, Smartly and AITarget. We chose to work with AITarget, because of their solution flexibility: they have more possibilities to set up rules, and they were ready to create some custom tools for us. After a month of testing, we realized we found the perfect tool that manages our automation routine tasks. It allowed us to grow our Facebook Marketing Budget up to 400K per month by October 2016, without scaling the team.
A Guide to Our Process
I’m going to be totally transparent, and tell you about our rationale and process for optimization using the AITarget tool. I need to provide a disclaimer: our rule-set is not a universal to set up. We developed our routine through our own trial and error, constantly testing new ideas, and always analyzing the data.
Every business or advertiser will have to develop it’s own routine, but here I’m providing a basic guide as a strategic starting point.
Kill the Poor Performers Early
Regarding Facebook buying, an advertisement (ad), or group of ads (ad set), may not perform as well today as it did yesterday. One main objective with automation is to understand as early as possible when an ad isn’t performing, so you can switch it off, leaving the performing ads live. Apply this rationale to every ad you’re running, so the worst performing ads are all turned off.
Understand the History
Before you go in and change any settings for ad optimization, spend some time analyzing your historical data to understand conditions (Spent Amount / Cost per Click / CPM / CPA / CPI), if the information is available.
We downloaded and analyzed all statistics for the past six months to look for any correlations, so we could build rules around them.
Our Automation Rules
1) When to Shut Down Creative
We established a rule set to determine when to stop running creative (all numbers are for example only)
spend GREATER than $100 Today & Purchase LESS than $1 Today
Action -> Stop
We noticed that if a creative ad spend is greater than 50% of our goal CPA, and did not result in any sale, then there’s a high probability it won’t perform well today, so we shut it off.
Spend GREATER than $30 Today & Cost-per-Click (Link) GREATER than $2 Today
Action -> Stop
2) Keep Your Eyes on the CPC & CPM
We applied the same rationale to control the Cost per Click (CPC): the idea is to understand that particular ad will have a high CPC for today and switch it off before you understand it doesn’t perform on CPA basis.
It’s easier to catch a high CPC in the early stages, rather than the CPA. If you find a pattern, this also works for the Cost per 1000 Impressions (CPM). You can also set up optimization more accurately, if you estimate the price of middleway conversions in the funnel (for instance, authorisations).
spend GREATER than $200 Today & СРА GREATER than $75 Today
Action -> Stop
3) Set the CPA in the ‘Corridor’
Ensure that the CPA is included in the ‘corridor’ under your KPI, while allowing the ad spend to continue to grow. Establish the rule so that it will react on your spend spikes because not all conversions can be tracked uniformly.
Spend GREATER than $75 Today and CPA LESS than $75 Today
Action -> Start
4) Timebox Your Data
There’s a common challenge in any analytics system where conversions are not tracked correctly due to delays in delivery (in our case, we set up a 15-minute period to check the rule set). You’ll need to establish rules that could fix the situation when a system turns off an ad, especially moments before conversions have increased.
Spend LESS than $40 Today & CPC (Link) LESS than $2 Today
Action -> Start
The same rule set can be used in CPC control because clicks tracking can be delayed as well.
5) Turn Ads Back On
It’s extremely important to remember to turn ads on again the next day. There are two methods:
Turn on all ads (that you turned off the day prior) and then, in 3–5 days, exclude the worst of them from the automation.
Establish the rule set of turning on ads with affordable CPA from the last 3 or 5 or 7 days.
The second way is more sophisticated, because it may not need your involvement, but it is quite complicated due to a high chance of unintended consequences. We use a hybrid method to exclude the worst performing ads from automation.
If your target CPA is $75, then set up a higher barrier, and reduce the CPA.
Impressions LESS than 10 Today & CPA LESS than 85 Last 7 days & (Time GREATER than 1 a.m. & Time LESS than 2 a.m.)
Action -> Start
6) Adding a New Ad to a Set
When adding a new ad to an ad set, you’ll need to timecode it, mitigating the chances of it turning off in the middle of the day (that may end badly if you have big budgets).
We’re also working with AITarget on a solution, which could help analyze conditions to decide whether to activate the ad set or leave them until the next day, after the moment when all ads are turned off.
7) Play with Your Budget (Conservatively)
The second rule set that may be used for a budget control is to increase the budget up to 20% — 30% if an ad set has a good CPA at the beginning of the day.
Spend / Daily Budget GREATER than 15% Today & Spend / Daily Budget LESS than 25% Today & CPA LESS than $70
Action -> increase budget on 20%
This rule set tracks the moment when the amount spent is enough, but before it’s too late to increase the budget.
Our daily routine consists of these rule sets. Although our rules vary slightly, they are universal for us. We were able to eliminate the need to scale the team because of daily routine automation managed by one person.
Please, email any suggestions to email@example.com