Why Lookalikes?
Replicate your best accounts.
It can be challenging to describe your ICP in terms of data points, fields, and filters. It's easy to describe your ICP with examples. We can all say "these are my best accounts" or "I want to replicate customer A, B, C." Lookalikes and Similarity Score make the easy way possible.
Keyplay Approach
If you’ve used enterprise-y ABM platforms in the past, you may be familiar with the concept of lookalike models in “predictive” scoring. Keyplay’s approach is fundamentally different.
Instead of a highly complex implementation that depends on your first party data, Keyplay lookalikes focus on external market data. Our lookalikes find & score companies with highly similar business models relative to your best accounts. To do this we collect and pre-process tons of data about each company, use LLMs to create a similarity graph, and surface the results in your Keyplay account scores.
How to Use Lookalikes & Similarity Score
1.) Define a Lookalike List
Go to Account Scoring and you’ll see the Similarity section below:
Click Define Lookalike List to add accounts you want to replicate. We'd suggest adding anywhere between 10-50 lookalikes to start (the more the merrier!).
💡 This list should NOT be your entire customer list. Instead, focus on the best of the best — high NPS accounts, customers in hot segments, or the top opportunities you’d most like to clone.
2.) Prioritize with Similarity Score
Once you’ve added lookalikes, click “Refresh Data.”
Every account in your CRM and Recommended Accounts with get a Similarity Score based on how closely related it is to the accounts on your lookalike list. We’ll also show which lookalikes where the closest match so that you can build highly targeted plays around this data.
Use Similarity Scores as part of your overall fit score
There are 5 options for how to use similarity scores in conjunction with the signal score (pulled from the number of signals matched in your scoring model) to assign an account an overall fit score.
Depending on your use case, you'll want to choose a score mode that aligns to the factors you want to focus on most for your overall scoring.
Here is an overview of the different score modes and some example use cases to help you decide what makes the most sense for your overall scoring.
Take the Highest Score.
Definition: Overall score will be which score is highest between similarity score and signal scores.
Example Use Case: You know that accounts of any employee size which match the signals in your model are a good ICP fit, but you also know that accounts which look like your enterprise customers are always a good fit, even if they don't match all the signals in your model. Using the "take the highest score" mode with your enterprise customer list as lookalikes will ensure that both accounts that match the signals in your model and match your enterprise list will get high scores.
Similarity score + bonus points for signals.
Definition: This mode is mostly based on similarity, but a strong signal score (65+) will add 20 bonus points to the overall score.
Example Use Case: You have a few key signals that are somewhat strong indicators of an ICP fit, but the core of your ICP model strongly resembles your best fit accounts. Using the "similarity score + bonus points for signals" mode will make similarity scores the basis for your overall scoring, but give accounts that match your select signals a bump in overall scoring.
Signal score + bonus points for similarity.
Definition: This mode is mostly based on signals, but a strong similarity score (65+) will add 20 bonus points to the overall score.
Example Use Case: You have spent a lot of time refining, back-testing, and analyzing your signal scoring model and weightings, and feel like it's highly representative of your ICP. However, you still want accounts that match your best fit accounts to receive a bump in scoring. Using the "signal score + bonus points for similarity" mode will make the signal score the basis for your overall scoring, but give accounts that match your lookalike list a bump in the scores they receive.
Similarity score only.
Definition: You'll still see the signal score in the app and CRM sync, but it won't count toward overall score or tier.
Example Use Case: The most important element of your scoring is that you can replicate your best fit account list, and you want focus your time on those accounts only. You would rather have signals be a supporting data point rather than impact overall scoring.
Signal score only.
Definition: You'll still see the similarity score in the app and CRM sync, but it won't count toward overall score or tier.
Example Use Case: You know exactly what signals make up your ICP, and you don't want to cloud that information with another score. You would rather have similarity scores and lookalikes be a supporting data point rather than impact overall scoring.
3.) Filter & Tag New Lookalikes for Targeted Plays
Similarity Scores will help you find & tag accounts for a highly relevant account-based plays. For the purposes of the example below I’ll pretend that Allbirds is one of my highest NPS customers. I want to run a play to similar companies talking about the extraordinary value I’ve delivered to Allbirds. Here’s how I’d orchestrate this play:
Filter CRM and/or Recommend Accounts by “Similar To.”
Save & Tag those accounts with a play name.
That Tag will then sync to my CRM so that I can orchestrate an ABM or outbound play around this idea.
Recap
Lookalikes are a powerful way to build targeted plays model your ICP, and find high value accounts. Keyplay makes it easy to replicate your best accounts with an intuitive lookalike model and actionable similarity score based on company business models.
You can get started today by adding a list of lookalikes to your Keyplay account, deciding how to incorporate similarity scores into your overall scoring, and surfacing net new lookalike accounts from your Keyplay Discover Tab.