High Value Customer Acquisition on Twitter: A Step-by-Step Guide

Customer acquisition has long been an engine of growth for e-commerce retail, especially with the expansion of low-cost digital marketing channels like affiliates and product listing ads. In the early days of e-commerce, it was a land grab — acquire as many customers as possible, for as cheap as possible. But we’ve reached a tipping point as e-commerce approaches maturity. Many brands have seen their cost per acquisition (CPA) going up as retailers saturate digital channels to reach new audiences. At the same time, the return on every new customer acquired is going down — as customers, more price-sensitive than ever, are lured away to low-cost competitors. The way for brands to break the cycle and drive sustainable long-term growth is to focus on acquiring higher-value customers.   

State of the world

The world of acquisition marketing today is one of highly fragmented channels and tools. Different channels rely on different types of technology and point solutions to optimize around their own metrics — which often differ depending on the role that the channel plays in the path in the conversion. Adding to the confusion, channels frequently don’t communicate with one another — making it difficult to assemble a complete view of the customer.

Acquiring higher-value customers is a noble goal — but a goal can be difficult to operationalize when each channel is run by a different channel manager who is measured against different success metrics, and when no single system of record links together user behavior across these different channels to identify high-value customers.

To understand this challenge, take a look at a few major channels — along with sample platforms, optimization technologies, and frequently used success-metrics for each channel.

 

Channel Sample Execution Platform Sample Optimization Technology Typical Success Metrics
SEM (search engine marketing) Google Adwords, Bing Kenshoo, Marin CPC (cost per click), ROAS
Paid social Facebook, Twitter Nanigans, Hootesuite CPM (cost per thousand impressions), Cost per Like, Cost per Conversion, ROAS
Affiliate Commission Junction (CJ) Commission Junction Traffic (site visitors), cost per conversion, ROAS

 

However, over the past few years, a solution has emerged: a number of acquisition marketing platforms — most notably Facebook, Google, and Twitter — have realized the tremendous potential locked away in CRM data. These data include insights about which users are most valuable, are interested in different products or price points, and more. And these ad platforms now make it possible for retailers to reach specific users in a 1-to-1 way, much like email has always enabled.

The purpose of this blog post is to explain not only how to create and target a specific group of users on Twitter, but to target predicted high value customers using Custora’s predictive insights with Twitter’s person-level targeting.

Introducing Tailored and Lookalike Audiences on Twitter

Retailers can upload their list of customers, subscribers, or users to Twitter. Twitter cross-references the emails provided with its database and creates a Tailored Audience based on matching email addresses, mobile advertising IDs, mobile phone numbers, or lists of Twitter IDs (user IDs or usernames). The marketer can then target promoted tweets, promoted accounts, and promoted trends to just this group.

Twitter Tailored Audiences

There are currently three main types of Tailored Audiences in Twitter:

  1. Tailored Audiences from lists: Created by uploading your own list of email addresses, mobile phone numbers, Twitter usernames and IDs, or mobile advertising IDs
  2. Tailored Audiences from web: Data about people who have visited your website collected by using Twitter’s website tag
  3. Tailored Audiences from mobile apps: Data about people who are using your mobile app collected with conversion tracking for mobile apps

Lists

This subset of Tailored Audiences is created by uploading a file containing your own data. Your records are then matched with people who are active on Twitter so that you can target them in your campaigns. We will be using the list capability when we do high value lookalike targeting.

Currently you can upload lists of:

  • Email addresses
  • Mobile phone numbers
  • Twitter usernames
  • Twitter user IDs
  • Mobile advertising IDs — iOS and Android

Supported file formats are .csv and .txt, and the maximum file size is 5 GB. Your list can be separated by new lines or commas. It’s worth noting that not every list you upload will be eligible for use in a campaign. In an effort to respect user privacy, the list must match more than 500 people on Twitter for it to be usable. To reach this minimum size requirement, most lists have several thousand or more entries. Lists can be uploaded through Twitter’s Audience Manager tool (more on that below).

Website Tags

Twitter defines a “web audience” as a group of active Twitter users who have visited your website. This group is collected by a website tag that you put on your site. This is a snippet of code that allows you to remarket to Twitter members based on visits to the sites where your tag is present.

Once placed, the tag begins to collect the cookie IDs of visitors and matches them to specific Twitter users. From there, you can set up Twitter campaigns targeted to these recent website visitors.

Mobile Apps

The third type of Tailored Audience focuses on mobile, allowing you to create targetable audience segments based on mobile app actions, such as installs, purchases, or sign-ups. This can help you drive the highest possible ROI in your app install and engagement campaigns.

Not only can you ensure that your ads aren’t displayed to individuals who have already installed your app, but you can augment tailored audiences from mobile apps with look-alike targeting to reach high-value users most similar to those who have your app installed.

Here is what it looks like in the Twitter App.

 

Select”Tools>Audience Manager” from the upper right side menu.

Then upload your csv or txt file with the customers that you want to target.

Once the file is loaded you can see it in Audience Manager.

Easy, right? Even better, marketers can generate something called a Lookalike Audience from a Tailored Audience. Twitter will look at all the profile data of the people in the custom audience and create a new list of Twitter users that share similar interests and demographics.

What is Lookalike Modeling?

As defined in Advertising Age, lookalike models are used to build larger audiences from smaller segments to create reach for advertisers. The larger audience reflects the benchmark characteristics of the original audience. More specifically, from a social network perspective, a lookalike audience is an algorithmically-assembled group of social network members who resemble, in some way, another group of members. Lookalike models enable marketers to target a larger audience of prospects that share important characteristics with specific customer segments—segments such as high lifetime value customers, or customers with an affinity for a certain product or promotion.

Why Use Lookalike Modeling?

Lookalike modeling is proven to significantly improve online advertising results. A study by Exelate found that:

Lookalike modeling “results in double or even triple the results of standard targeting, according to the 30 percent of advertisers and more than half of agencies who reported using the tactic.” — Exelate Study

Things to note when doing lookalike audience targeting:

  • The original, ‘seed’ group is a user list provided by the retailer. This is not interest or behavior targeting.
  • The seed group is identified by email address, phone number, website cookie, mobile phone number, or Twitter ID..
  • Marketers don’t need to provide any other information about the seed group—just their identifier.
  • The social network runs the algorithm and decides who of their members ‘looks like’ the seed group using what they know about the seed group members.
  • The marketer does not know why the lookalike group matches the original group.  The algorithm is proprietary and secret.

Marketers are not given access to the names on the list, but they can deliver ads to them as a group

Using Lookalike Targeting to Acquire High Value Customers

In our experience, one of the best segments for Twitter Lookalike Audience targeting are customers with the highest predicted lifetime value. In order to use Twitter to acquire more high value customers you need to do five things:

  1.     Calculate the CLV for each customer and identify your high value customer segment
  2.     Use that segment to create a “High CLV Custom Audience” in Twitter
  3.     Use the “High CLV Custom Audience” to create a lookalike audience in Twitter
  4.     Build and launch compelling tweet campaigns for your high CLV segment
  5.     Measure results

Calculating CLV

CLV (Customer Lifetime Value) is a prediction of all the value a business will derive from their entire relationship with a customer. Because we don’t know how long each relationship will be, we make a good estimate and state CLV as a periodic value — that is, we usually say “this customer’s 12-month (or 24-month, etc) CLV is $x”.

Many brands use historical sales performance to measure CLV. But CLV is a future-looking metric. It is the estimated future cash flows that you can expect from each customer. Historical CLV, since it is backward-looking, can produce misleading results when the company, the market, or both have changed. In addition, historical CLV is limited when trying to measure the CLV of customers using new channels or tactics. Luckily, there are several methods that can be used to predict CLV, including extrapolation, supervised learning algorithms, and probabilistic modeling.

If your company is lucky enough to have a data science team, that team has likely built CLV models that you will need to access for lookalike targeting. If not, you will want to purchase a solution like Custora which will accurately predicts CLV for each customer based on advanced statistical methods like Bayesian Inference and Pareto/NBD.

If you would like to learn more about predictive models for calculating CLV, please check out the very comprehensive overview in Custora U.

Creating a “High CLV Tailored Audience” in Twitter

Once you have created your “high CLV” customer segment in Custora, you will need to create a .csv file of all of the high CLV customers.

Within Custora you would set up a One-Time Campaign (or Recurring Campaign) focusing on Future High-Value Customers: Platinum, Gold, and Silver (i.e., the top 10%), with “send options” set to Extract. Because this is an acquisition campaign where results measurement is not incremental based on a holdout control group, there is no need to use a control group.

As I mentioned earlier, the audience lists are all managed and created within Twitter’s “Tools>Audience Manager” section located at the top navigation. After the list is uploaded and ready for use, simply choose it as your tailored audience source.

Unlike on Facebook which requires several steps to create a lookalike target audience, with Twitter you just need to select the “Expand reach by targeting similar users” button on the Audience Source screen.

Creating a Compelling Promoted Tweet Campaign

Deep customer insights are the secret to creating compelling advertising creative. In this case these insights might include information on high-value customers’ demographic attributes, product preferences, or other facets of their behavior (price sensitivity, return behavior, etc.). The goal is to discover who your high-value customers are and what makes them unique — with an eye to building ad creative that will appeal to them. For example, if a retailer knew that their highest-value customers are fashionistas from Los Angeles, they would design a message that would appeal to these shoppers.

Custora customers can use “Segment Insights” within the platform to glean actionable intel on your highest CLV customers that can inspire campaign ideas. You can learn, for example, that your highest value customers are disproportionately likely to be married women in their 40’s and 50’s, clustered in southern cities like Austin and Houston. And they gravitate towards denim pants — inspiring your team to create a “southern belle” themed campaign.

Finally, it is time for the retailer to launch a Twitter advertisement focused on this lookalike audience. If the protocol for creating and launching Twitter ads is not familiar to you try reading this great post by Hootsuite “How to Use Twitter Ads, the Complete Guide for Business”..

Measuring Results

Once the campaign has been launched, you should be able to view results within Twitter Campaign Tab.

Click on any ad campaign to see detailed results:

By default, you should be able to see campaign metrics like total reach and budget spent to date; if you have the Twitter conversion pixel installed, you will be able to see performance metrics like cost per conversion and, ultimately, ROAS.

Success for the High Value Customer Acquisition campaign is based on the ROAS (Return on Ad Spend) reporting tools within Twitter. You will typically evaluate the performance of lookalike segments against the performance of similar segments in past campaigns.

 

ROAS before lookalike (based on demographic targeting in Twitter) ROAS with lookalike (based on predictive high-value customers)
0.75x 3.90x

 

ROAS is defined based on the revenue spent by customers who either viewed the advertisement, or clicked on the advertisement — and the attribution window may vary. For example, one retailer may choose to give Twitter credit for the purchases of any users who saw an advertisement and then went on to purchase within the next week. Another retailer may choose to only give credit for customers who actually clicked on the ad, but may have a more lenient 28-day click-through attribution window.

These are matters of preference for individual retailers. The most important thing is that the retailer is comparing the ROAS from lookalike-powered campaigns with that of their next-best high-value customer acquisition alternative in an “apples-to-apples” way.

The execution lead on the retailer side should continue to monitor these results weekly within Twitter Analytics to ensure strong performance. A recommendation from leading performance marketers is to focus the campaign creative and call to action on key events onsite — e.g., strong promotions, new arrivals, etc. The retailer should also refresh the Twitter list regularly with the most up-to-date high CLV customers.

On a quarterly cadence, the entire team should reconvene to review results, refresh insights on high-value customers, and plan out budgeting for the coming quarter.

For Custora customers, the success of High Value Customer Acquisition is measured differently from the test-and-control approach for other Custora solutions (e.g., churn prevention). The difference stems from the fact that most Custora campaigns are focused on communicating with existing customers, where it’s easy to carve out a control group from a given segment to benchmark how that segment would have responded in the absence of a marketing message. In contrast, an acquisition campaign is focused on reaching customers who aren’t currently part of the customer base — meaning there’s not a ready way to measure the incremental impact above a controlled benchmark of known customers.

Taking a Systematic Approach to High Value Lookalike Targeting

Twitter is not the only social network that enables lookalike targeting. Google and Facebook also allow you to use your own customer data to create audiences and target across channels to acquire customers that look like your most valuable customers. And leading Data Management Platforms (DMPs) enable cookie matching and lookalike targeting with a CRM audience segment. Given the superior performance of lookalike targeting and the fast growing capabilities across digital channels, it makes sense to invest in a system that automates the identification and sharing of high value audiences. If you’d like to see how this is done in Custora, please reach out. We love to show off our platform.

About Custora

Custora is an advanced customer segmentation platform that puts the power of predictive analytics in the hands of marketers. We help retailers improve the ROI of their email, display, direct mail, and Facebook and Twitter campaigns. Custora surfaces smart (and predictive) customer segments, integrates with marketing channel technologies to execute more effective acquisition and retention campaigns, and measures the impact on customer lifetime value. Over 100 brands including Lucky Brand, Ann Inc., Tiffany & Co., Crocs, and Teleflora use Custora to improve new customer conversions, grow revenue from existing customers, and improve team efficiency.  

 

Leave a Reply

Your email address will not be published. Required fields are marked *