Decrypting the Silent Market with First-Party Signal

October 2024 | RE Owls Intelligence

This document introduces the RE Owls Domain-Specific Model (DSM) for Real Estate Liquidity.

While the industry relies on "Lead Generation" (finding people), we rely on "Intent Vectorization" (finding truth). The first capabilities pertain to identifying the "Silent Market," predicting incentive thresholds, and mathematically engineering the path to a signed contract.

The RE Owls intelligence team consists of behavioral data scientists leveraging information-poor environments and a specialized unit of Revenue Engineers dedicated to the asset class.

1. Methods

a. Technical Challenge

Recent advances in AI have been largely useless to Real Estate developers for one structural reason: Data Blindness. No sales team has adequate data for AI predictions.

To succeed in Real Estate sales, an intelligence system must be very good at:

• Leveraging first-party data,

to know who visits{what, how, where, with whom} and who needs {what, when, why} to {register, visit, buy} what.

• Leveraging small, and messy data about past wins/losses.

Go beyond the (“💩 Garbage in = 💩 Garbage out”) limitation.

• Leveraging users data to imitate the reps instinct.

b. RE Owls models

RE Owls models and algorithms are fruit of:

1. Fine-tuning of existing models on real estate specific datasets

2. Internal project knowledge regarding market, unit mix, pricing

3. Data collected via our proprietary sensor networks

c. Introducing RE Owls’ first model

RE Owls develops a pre-trained model to generate an abstract (vector) representation of every visitor—even the anonymous ones. 

This abstract representation is then used to perform a variety of qualification tasks: estimating the likelihood of a won / lost status, finding similar leads, recognizing lookalikes, and spotting changes in closing tendencies that are not observable to the naked eye, extrapolating missing data.

The intuition behind this approach is that a large model, pre-trained on a corpus of real estate behaviour, is representative of the "Collective Subconscious" of real estate purchases — i.e. patterns of expression of the general human. 

Our goal is to make sense of traffic data that is often limited to a name, number and email, and understand why particular leads buy while others don't. Given this information-poor form, the historical data lacks information.

We augment the "won/lost" history with granular behavioural telemetry found in the Digital Foyer (site visits, floor-plan revisits, pricing toggles, map hovers, conversations). That way, RE Owls is able to generate rich representations of leads that have the capacity to qualify the affinity of future potential buyers to the inventory. 

Leveraging the 'collective subconscious' to perform this task is similar to an experienced Sales Director who, after years of experience, can intuitively guess the likelihood of closing a client just by watching them walk through the door. RE Owls now helps them do it faster, hand-in-hand with AI. 

2.Performance metrics

a. Lead Status Prediction: Power of the lead vector

Here we are interested to see how our vectors perform when a basic classification approach is applied to them.

We first calculate vector centroids for the sets of lead vectors associated with status  won  (Sold Unit) and then the same for lost (Dead Lead). For each lead, we calculate the cosine similarity with each of the centroids. We take the argmax of those two similarities to be the predicted status.

Comparing the true statuses and the predicted statuses, we measure:

• General Accuracy: number of leads with correctly predicted statuses / total number of predictions.

• Precision: The number of correctly predicted closed-won statuses divided by the total number of leads predicted as of closed-won.

• Recall: The number of correctly predicted closed-won statuses divided by the total number of actual sales.

b. Anonymous customers - results

Pipeline data - prior classification by RE Owls

Pipeline - classified and augmented by RE Owls

RE Owls isolates leads that will close (close) / that will be lost (white). It is accurate at 96% based on CRM history data.

c. Anonymous customer - early analysis

From our CRM history dataset sample, we understand that:

• Sales Reps waste 70% of their time on lead that will never close.

• RE Owls model correctly predicts in 96% of the cases the right lead win/loss outcome

Let’s do some (brutal) maths

• If you take a team of 10 sales reps x $80,000/annual cost for the company x 70% = $560K is wasted by the company on leads that will never close.

• You can now x3 your team efficiency.

3. The Execution

Technology without execution is just a dashboard. We are an Agency, which means we deploy agents to act on the vectors. 

• The Sensor (Digital PC): The "eyes" that collect the First-Party Ground Truth. 

• Agent Oliver (The Analyst): The "brain" that calculates the Cosine Similarity and sets the incentive strategy. 

• Agent Hank (The Operator): The "hands" that execute the intercept when the vector aligns with the "Won" centroid. 

We don't guess. We calculate.

* Sales Reps waste 67% of their time on lead that will never close - is a figure calculated over several CRM historical datasets. This data has been confirmed by several studies including Steven Tulman, LeadMonk, Zippia