Most messaging apps utilize the phone’s local address book to jump start a user’s network. Typically, these apps connect you with other people who are already on the network by comparing phone numbers. They then prompt you to invite the remaining contacts in your address book to try the app — typically via SMS or email.
Ideally, when suggesting individuals to invite, you want to highlight people that lie in the cross section of “likely to be invited” and “likely to install/use your app”. But how do you identify these people when you can access only the limited information in the address book. For example, it would be highly valuable if we could see the contacts that a user has flagged as “Favorites”, but the API for iOS does not allow this.
Below, we outline a few techniques that can be used to make better referral suggestions. It’s important to emphasize that the effectiveness of these techniques depends upon individual user behavior and device operating system. There is no “silver bullet”, so your best strategy is to employ a variety of methods.
The names of people in your address book actually reveal much information.
- One way to analyze names is to look for keywords that represent contacts with special relationships to the phone’s owner — exactly what types of relationships you care about will depend on the context of your app. For instance, if you want to suggest close family or friends, you might consider promoting contacts with names such as “hubby”, “wife”, “sis”, “bro”, “bff”, “bestie”, “dad”, “mom”, etc…
- Another way to determine close family and friends is to look for contacts with the same last name as the phone’s owner. Unfortunately iOS does not reveal the phone owner’s entry in the address book. However, if your signup process requires a full name you are all set. If not, you can leverage the user’s phone number or email address to locate the corresponding entry in the address book and then grab its last name.
- Names can also be used to infer gender and age. This is where census data can be especially useful because there exist databases of names correlated with gender types and years.
- Finally, you can infer close social relationships by looking for special formatting or symbols in names. If the majority of contacts in a user’s address book contain both first and last names, then we might guess that contacts which break this pattern are perhaps more important. Furthermore, if an entry contains emojis, you can interpret positive or negative emotions based on the specific symbols used (e.g. smiley faces vs sad faces). We have observed this trend particularly in the teenage demographic.
Area codes and mutual contacts are effective ways to dissect an address book for suggestion purposes.
- Area codes can serve as a proxy for geographic closeness. If you know the area code of your user, then you can identify other users in the address book that he/she may be close to.
It’s important to note that people often maintain their mobile phone number when they move to a new region, so your mileage may vary here.
- If your app has access to the address books of other users, then you can make invitation suggestions based on phone numbers that appear in multiple address books. It may be effective to display the names of the sources for social credibility. For instance, “You and Jeff are mutual friends of Sally — send her an invite.”
Another vector for ranking the quality of a potential invitation is the email domain of the contact.
- For social apps, you may want to prioritize contacts with email addresses from major consumer email services such as Google, Yahoo, Microsoft, and Apple. This is not a one-size-fits-all solution, but you can tune your algorithm to optimize for age and nationality.
- Enterprise apps can simply do the opposite and ignore the major consumer email domains — instead prioritizing email addresses from a database of companies (like the Fortune 500). If the phone owner has provided a work email, then you can find colleagues by looking for contacts with matching domains.
Address books contain a wealth of meta information beyond simply names and phone numbers. Some factors to consider include:
- Whether a contact entry contains a photo.
- The type assigned to a phone number — e.g. “home”, “work”, “mobile”.
- If tags are provided for the contact.
- The volume of data provided per contact.
However it’s important to be careful when evaluating these factors because the information may not have been manually entered by the phone’s owner. This is particularly the case when an address book is synced with another source such as GMail or Facebook.
Providing suggestions of who to invite can help in increasing the success of referrals for an app. Often times the address book is your only data resource for generating recommendations. By using names, phone numbers, emails, and other metadata, you can create heuristics to rank people based on the target user for your app. It’s a mix of both art and science — so it may not work perfectly, but it’s a start.