Paid User Acquisition for Early Stage Startups – Part 3: Facebook Advertising Basics

This is the third post in a 6 post series I am writing about how early stage startups should approach paid user acquisition. My last post outlined some of the basic rules of landing page optimization. This post will cover how to approach Facebook advertising. These rules are based on my first-hand experience running millions of dollars in paid ad spend over the last few years while consulting. They are also informed by the millions of dollars a month in spend we track and analyze on behalf of our customers at my current company, Interstate Analytics.

Facebook is an excellent channel at scale for many startups. I’ve personally worked with multiple companies spending 150-250k/mo profitably (positive return on investment within 30 days) to acquire new customers on Facebook. However, it takes a while to scale up and many of the tactics that work for people spending millions a year on Facebook will not work for you at an early stage. First I’ll break down tactics that work at scale and then I’ll talk about how they work or don’t work at smaller scale.

Facebook Tactics at Scale

At every company I’ve seen who spends millions a year on Facebook profitably, the tactics for running Facebook ads are very similar. I’ve been involved mostly with e-commerce so this may not apply if you’re working in gaming, B2B, mobile-only, etc; however, I hear from my friends who have worked in those sectors that strategies are similar:

  1. Generate a list of the email addresses of either all users who have purchased, or some smaller subset of ‘good’ users. It’s ideal that this list be large. At least 10,000 emails at minimum, even better if it’s 100k emails.
  2. Upload this list to Facebook to create a Custom Audience
  3. Create a Facebook Lookalike Audience of users in this demographic. Select options for smaller reach and higher accuracy.
  4. Create News Feed ads (not right hand side as as those only seem to work well for retargeting, not for targeting new customers) mostly on web (conversion to purchase on mobile is poor).
  5. Profit

There are a number of variations on this method including running different ad types (e.g. page post boosting, mobile ads, etc), creating larger custom audiences with less accuracy and then filtering them down with various types of demographic targeting, etc. Most companies spend some portion of their Facebook budget retargeting existing users instead of prospecting for new ones, but this lookalike audience strategy is still the way that 70-90% of Facebook budgets are spent in e-commerce.

Facebook Tactics at Not so Much Scale

The tactics above are great for larger companies with bigger budgets, but what do you do if you’re a small guy without a huge email list to build a custom audience from?

  1. Aggressively Collect Email Addresses
    We covered this in the previous post on landing pages, but it’s important to remember that email addresses are your friend. If you collect lots of emails early from PR launches and other ‘free’ promotional activities you can build a lookalike audience of people who are interested in your product. It’s not as good as a huge list of purchasers, but it’s is better than nothing.
  2. Test Interest Targeting
    Facebook has an option to target based on interest data, which is harvested from their users’ Facebook data. Not many companies are doing this at scale; however, if you don’t have a large enough list of relevant emails to generate a lookalike audience, it can be a good option at a smaller size. Try a couple different combinations – as we discussed in the first post in this series, don’t try more than a couple combinations because you won’t have the spend volume for it – and iterate. Partner audiences generally perform worse than Facebook interest targeting, so avoid those.

Facebook Tactics at All Sizes

  1. Roll Your Creative
    If you run the same creative to the same audience for too long on Facebook, you will eventually see performance start to fall off. The time this takes varies, but generally I’ve noticed it takes somewhere between 2-4 weeks. This seems to apply only to the visual creative, not the copy itself. You can run the same copy for long periods of time as long as you constantly refresh the creative.
  2. Leverage Facebook Auto-Optimization
    If you put multiple ads in a Facebook ad set, it will auto-optimize them for you and automatically display the ones that it believes to perform the best. This is not based on statistical testing, but rather a Facebook machine learning model. I’ve run tests in the past to A/B test statistical results against Facebook’s machine learning model and found that Facebook usually chooses the same result, with the added benefit of choosing it much quicker. This can allow you to test creative and copy changes against each other faster by running two or more ads in the same ad set and seeing which one(s) Facebook optimizes out. You’ll know when this happens because one of your ads will get lots of impressions and the other(s) will get close to none. This is Facebook telling you that your other ads probably suck compared to the one it’s giving impressions to.
  3. Bidding Strategy
    Bidding for conversions (also known as oCPM) is the best Facebook bidding strategy unless you have a specific reason not to bid that way (e.g. you’re doing brand marketing and are optimizing for views/reach). When bidding oCPM it can also be helpful to pick a goal further up the funnel (e.g. ‘Add to Cart’ instead of ‘Purchase’) if you have a low volume of conversions (e.g. less than 10 conversions a day). This gives Facebook’s auto-optimizer more data to work with. Once you’ve picked your goal you can either let Facebook auto-bid, or bid yourself. If you’re just starting out, it can be a good idea to let Facebook auto-bid in order to get a ballpark range of how much you can expect to pay for a conversion. Once you have a better idea of expected cost per conversion (CPA), and in particular CPAs for different user demographics (e.g. Gender, Age Group, etc) or placements (e.g. News Feed, Right Hand Side, Audience Network, Mobile Newsfeed, etc) you can worry about setting your own bids.
  4. Segmentation Strategy
    If your Lookalike audience is sufficiently large it can be helpful to segment it so you can get an idea of how different groups of users are performing. You do this by creating separate ad sets within a campaign that run all the same ads and just target different demographics. This shouldn’t affect performance, but it will give you a better idea of how CPAs differ by demographic, which can help with optimization. As a rule of thumb I wouldn’t recommend splitting audiences in a way that leads to audiences smaller than 20-30k people in each audience. Another thing to ensure is that if you’re targeting multiple placements (e.g. News Feed, Right Hand Side, Audience Network, Mobile Newsfeed, etc) that you always segment placements into separate ad sets or campaigns. You want to be able to bid differently on different placements as well as understanding performance differences between the same ads on different placements, so you must segment them.
  5. Attribution
    If you’re measuring performance directly in Facebook make sure you check the attribution settings before reporting on conversion data. Facebook’s default attribution model gives Facebook credit for any conversions that happen within one day of a user viewing any ad, or within 28 days of a user clicking on an ad. This means that if you don’t change the default attribution, then a user could scroll by your ad and barely notice it, then search for your site directly on Google and purchase and Facebook would take 100% of the credit for that conversion. If you’re spending across more channels than just Facebook, you can use a tool like Interstate Analytics for more accurate attribution.
  6. URL Tagging
    Make sure you are tagging all your ads with UTM parameters. You can do this in the URL Tags field in each ad. This will help you measure how each campaign and ad are driving results back to your site. For a hypothetical campaign called ‘redbull_sale’ targeting 20-30 year old males with newsfeed ads with a creative you call ‘creative_a’ and copy you call ‘copy_b’, a good structure for your URL would be something like: http://yourwebsite.com/?utm_source=fbads&utm_campaign=redbull_sale&utm_medium=newsfeed&utm_content=cr_a-co_b-20_30m With this setup, if a user clicks through on that ad you’ll know the source, campaign, copy, creative, and placement for the ad click in all of your analytics tools. If you don’t tag URLs at all (e.g. just sending users to http://yoursite.com), then Facebook is the only place you can check stats on conversion. Make sure you set up URL tags so you can sanity check your performance vs Facebook’s dashboard in other tools!

In the next post we’ll cover the basics of Google AdWords advertising. Sign up here to get notified when the next post goes live:



Paid User Acquisition for Early Stage Startups – Part 2: The Four Rules of Landing Page Optimization

This is the second post in a 6 post series I am writing about how early stage startups should approach paid user acquisition. My last post outlined some of the basic rules of customer acquisition. This post will cover how to approach landing page optimization coming from paid channels. These rules are based on my first-hand experience running millions of dollars in paid ad spend over the last few years while consulting. They are also informed by the millions of dollars a month in spend we track and analyze on behalf of our customers at my current company, Interstate Analytics.

The Four Rules of Landing Page Optimization

1. Have a Single Call to Action
Your landing page should only have one action that can be taken. Pull users in one direction, not ten different ones. Don’t make users click to a features page or a solutions page or a list-of-products page before deciding to sign up or give you their email. Users should not need to wade through tons of copy that describes every single possible use case. You should have enough copy and supporting material to get them interested and no more. Get users interested, then collect email. Save further education for onboarding.

2. No Distractions
You shouldn’t have navigational links to other parts of your site on landing pages. They’re effectively secondary calls to action that distract from your primary call to action per #1. For example, compare the landing experiences for Everlane between organic and paid traffic…

Everlane’s Landing Page for Organic Traffic
Screenshot 2016-02-29 09.06.32

Everlane’s Landing Page for Paid Traffic
Screenshot 2016-02-29 09.06.48

Or check out how Amazon’s navigation disappears the second you get into the checkout flow…
Screenshot 2016-02-22 09.32.40

Keep it simple and remove as many distractions as possible.

3. Collect Email First
95% of the time email collection should be the call to action your landing page focuses on. Put the field right on the page and at the bottom. Don’t make users click through to another page with more information before they submit their email address. If you do this you should see between 20-40% conversion rate from users hitting your landing page to submitting their email address. If you’re worried about the additional drop-off this will cause, you should not be, the ability to email a large percentage of your landing page visitors far outweighs the marginal drop-off that forcing email signup early will cause. One caveat: while this does always seem to be true for pages a user lands on via a paid source in my experience, it is not always true for visitors landing on your site via unpaid channels (e.g. Google Search, etc) who you may sometimes want to drop directly into different content.

4. More Steps Are Often Better Than Fewer Steps
This is counter-intuitive, but you’re often better served by creating a multi-step onboarding process rather than putting everything into a single step. This allows you to have a simple ask up front which is easier for the user to complete, then once the user is invested (by means of completing a couple steps) you can ask them to complete more complicated tasks without triggering any drop-off. I explore this in detail with a real-world example in another post I wrote previously.

In the next post we’ll cover the basics of Facebook advertising. Sign up here to get notified when the next post goes live:



Paid User Acquisition for Early Stage Startups – Part 1: The Five Rules of Paid User Acquisition

This is the first post in a 6 post series I am writing about how early stage startups should approach paid user acquisition. This first post will outline some of the basic rules of customer acquisition. These rules are based on my first-hand experience running millions of dollars in paid ad spend over the last few years while consulting. They are also informed by the millions of dollars a month in spend we track and analyze on behalf of our customers at my current company, Interstate Analytics.

In later posts we’ll dig into landing page best practices, as well as the top 4 channels startups should be considering early on in their lifetime: Retargeting, Facebook, AdWords, and paid content marketing (Taboola/Outbrain). Let’s get started.

The Five Rules of Paid User Acquisition
  1. Make Something People Want
    Paid User Acquisition is not a replacement for a good product. It can sustainably boost growth for your good product but it can’t save a bad product.
  2. Listen to Statistics, Not Your Gut.
    Optimizing paid user acquisition simply a matter of running lots of controlled tests, only unlike other A/B tests these cost real money to run. You should be familiar with how to run basic statistical tests on your data, and how to avoid multiple significance testing errors. Here are some good resources to read and use:
    Avoiding Multiple Significance Testing Errors
    Sample Size Calculator
    Chi-Squared Test Calculator
  3. Run a Small Number of Big Tests
    This rule derives from Rule #2. As a small startup, you have a limited marketing budget. This means that if you want to know anything with certainty, you can’t run hundreds of tests. For any given test, the smaller the baseline conversion rate and the smaller the difference you are trying to measure, the more samples you need to test. e.g. If your baseline conversion rate is 2%, in order to measure a 1% improvement with 95% confidence you need 2615 samples, but to measure a 4% improvement you only need 478 samples. (check the sample-size calculator above). Experiment with concepts that are very different. Minor changes in copy or creative lead to minor changes in conversion rate. Look for major changes.
  4. Learn Baselines
    In order to know which major changes are feasible, you first need to understand what ‘good’ means. There are established metrics of what ‘good’ means for everything you’d want to measure, whether its a click-through rate for Facebook newsfeed ads, a cost-per-click (CPC) for Outbrain ads, a conversion to email signup on your landing page, or a conversion to purchase from a Retargeting ad. Knowing these baselines will help you make sensible decisions about how you are doing and where you should be focusing your efforts. For example, if 3% of users clicking to your landing page from an ad convert to entering their email, you can probably improve that 10-12x. If you’re already at 30%, you can only improve 3.3x theoretically and likely only 0.6-0.7x realistically. I’ll try to include baselines in upcoming channel-specific posts wherever I can, but they aren’t secrets. If you Google or ask around you can usually get a good idea of how you should be doing.
  5. Focus On the Right Metrics
    Many companies focus on the wrong metrics for paid user acquisition. Cost Per Click (CPC), Cost Per 1000 Impressions (CPM), Cost Per Conversion (CPA), etc. The only number that matters is Return on Investment (ROI) measured over time. Return on Investment is calculated by dividing the amount of net revenue a campaign drove by the amount spent on the campaign. As a result, it can change over time as more people purchase. Tracking this is important, especially if you are a subscription business or any other business where you don’t expect to break even on a user’s first purchase. If you’re doing a great job you’ll hit 100% ROI in the first week, if you’re doing a good job you’ll hit it in the first month, if you’re doing an ok job you’ll hit it within 3-6 months. Metrics like CPA, CPC, etc can be leading indicators for ROI in some cases, but if you are relying on those metrics you have to recognize them for what they are, proxies to the real metric you’re trying to measure.

In the next post we’ll cover landing page optimization for paid user acquisition. Sign up here to get notified when the next post goes live: