What is an Algorithm?

In plain English: An algorithm is simply a set of step-by-step instructions for solving a problem or completing a task. It’s a recipe that tells a computer what to do. When vendors boast about their “proprietary algorithm,” they’re just saying they have their own set of instructions—which may or may not be better than anyone else’s.

Why It Matters to Estate Agents

You’ll hear “algorithm” constantly in PropTech sales pitches:

• “Our proprietary algorithm scores leads with 95% accuracy”
• “Advanced algorithms power our property matching”
• “Our pricing algorithm analyzes market trends in real-time”

Here’s what you need to know: The algorithm itself matters far less than the data it works with and how well it’s designed for your specific problem.

A sophisticated algorithm working with poor data will fail. A simple algorithm working with excellent data will succeed. When a vendor obsesses over their “proprietary algorithm” but won’t discuss their data quality, training methodology, or accuracy metrics—that’s a red flag.

Understanding algorithms helps you:

Cut through marketing nonsense. “Proprietary algorithm” sounds impressive but means nothing. Every piece of software has algorithms—they’re just instructions. The question is whether those instructions are effective for your specific use case.

Ask better questions. Instead of being impressed by algorithm complexity, you can ask: “What data does your algorithm use? How often is it updated? What’s its accuracy rate on properties like mine?”

Recognize when “algorithm” is a synonym for “we won’t tell you how this works.” Sometimes “proprietary algorithm” means “our secret sauce that actually works brilliantly.” More often it means “we can’t explain this clearly” or “it’s not that special, but we want you to think it is.”

Real-World Example: Property Matching

Let’s look at three different algorithms solving the same problem: matching buyers with suitable properties.

Algorithm 1: Simple Rule-Based

IF buyer budget ≥ property price
AND buyer bedrooms wanted = property bedrooms
AND buyer location preference = property postcode
THEN show property to buyer

This is an algorithm—a clear set of instructions. It’s simple, predictable, and transparently easy to understand. It also misses lots of good matches (buyer wanted 3 beds but would consider a spacious 2-bed).

Algorithm 2: Weighted Scoring

Score = 0
IF budget matches: add 40 points
IF bedrooms match exactly: add 30 points
IF within 2 miles of preferred location: add 20 points
IF property type matches: add 10 points
IF price is 10-20% below budget: add 5 bonus points
IF property has garden (and buyer wants one): add 5 points

IF score ≥ 70: Strong Match
IF score 50-69: Possible Match
IF score < 50: Poor Match

This is a more sophisticated algorithm—same concept (step-by-step instructions), but with nuance. It recognizes that not all criteria are equally important. Still transparent and explainable.

Algorithm 3: Machine Learning

Analyze the last 10,000 property viewings
Learn which property characteristics predicted:
  - Viewer attended
  - Viewer made offer  
  - Viewer completed purchase

For each new buyer-property pairing:
  Calculate probability of positive outcome based on learned patterns
  Rank all properties by probability
  Show highest-probability matches first

This is also an algorithm, but one that learns from data rather than following fixed rules. It might discover patterns you never explicitly programmed: buyers who viewed properties on Tuesdays converted 15% more often; properties with specific phrases in descriptions attracted serious offers.

All three are “algorithms.” The third isn’t necessarily better—it’s more complex and requires good data, but for a small agency with limited historical data, Algorithm 2 might perform better.

Common Misconceptions

“Proprietary algorithms are always better than standard ones.” Sometimes proprietary just means “different,” not “better.” Google’s search algorithm is genuinely superior. Your local CRM’s “proprietary lead scoring algorithm” might be worse than industry-standard approaches.

“Complex algorithms are better than simple ones.” Often, no. Simple algorithms are easier to understand, debug, and trust. A complex algorithm that’s 2% more accurate but impossible to explain or adjust might be worse than a simple one you can actually work with.

“AI doesn’t use algorithms.” Wrong. AI systems use algorithms—they’re just algorithms that learn and adapt rather than following fixed rules. Machine learning is an algorithmic approach to finding patterns in data.

“You can’t question proprietary algorithms.” You absolutely can and should. You don’t need to see the source code, but you should demand: What data does it use? What’s its accuracy? How does it handle edge cases? Can you explain a specific decision it made?

Questions to Ask Vendors

When a vendor emphasizes their “algorithm”:

1. “What specific problem does your algorithm solve?” If they can’t articulate this clearly, they’re hiding behind jargon. “Our algorithm optimizes lead conversion” is vague. “Our algorithm predicts which leads will schedule viewings within 48 hours with 82% accuracy” is specific.
2. “How does your algorithm differ from standard approaches?” This reveals whether “proprietary” means “genuinely innovative” or just “we did it ourselves.” If they can’t explain the difference, it’s probably not special.
3. “What data does your algorithm require, and what happens with poor or incomplete data?” Algorithms are only as good as their inputs. An algorithm requiring 5 years of historical data won’t work for a new agency.
4. “Can you walk me through an example of how your algorithm made a specific decision?” Good algorithms can be explained, even if the exact code is proprietary. If they can’t explain their reasoning, that’s concerning.
5. “What’s your algorithm’s accuracy rate, and how did you measure it?” Vague claims of “high accuracy” are meaningless. “85% accuracy on a test set of 5,000 properties, with performance dropping to 72% for properties over £1M” is useful information.

When Algorithm Choice Actually Matters

Algorithm matters most when:

• The problem is genuinely complex - Predicting property prices requires sophisticated algorithms because countless variables interact in non-obvious ways
• You’re processing huge amounts of data - Analyzing 100,000 property transactions needs efficient algorithms that can handle scale
• Precision is critical - Mortgage payment calculations need algorithms that handle edge cases and rounding correctly
• You’re competing on performance - If rivals use standard algorithms, a better one gives competitive advantage

Algorithm matters less when:

• Data quality is the real bottleneck - A brilliant algorithm on garbage data produces garbage results
• The problem is straightforward - Sending appointment reminders doesn’t need algorithmic sophistication
• Explainability trumps accuracy - For fair housing compliance, simple algorithms you can explain beat complex black boxes
• Human judgment is the final step - When AI suggestions are just inputs to human decisions, algorithm perfection is less critical

The Algorithm vs. Data Trade-off

Here’s an uncomfortable truth: in most estate agency applications, data quality matters more than algorithmic sophistication.

Scenario 1: Brilliant algorithm + Poor data = Poor results
Example: Sophisticated ML algorithm trained on 200 random property sales, all from 2019, none in your specific area.

Scenario 2: Simple algorithm + Excellent data = Good results
Example: Basic weighted scoring using 5,000 recent sales from your exact patch, updated monthly.

Scenario 3: Brilliant algorithm + Excellent data = Excellent results
Example: Sophisticated ML trained on 50,000 recent comparable properties, fine-tuned to your specific market, updated weekly.

Most vendors offer Scenario 1 but market it as Scenario 3. Smart estate agents demand proof they’re getting Scenario 2 at minimum.

The Bottom Line

An algorithm is just a recipe—a set of instructions for solving a problem. Some recipes are better than others, but the quality of ingredients (data) often matters more than the sophistication of the recipe.

When vendors tout their “proprietary algorithm,” don’t be impressed by the jargon. Ask whether their approach is actually better for your specific needs, and demand evidence beyond marketing claims.

The best algorithm for your agency is one that:

1. Solves your actual problem (not someone else’s)
2. Works with the data you actually have
3. Delivers results you can measure
4. Can be explained when necessary
5. Improves as your data improves

Complexity isn’t a feature. Results are.

Related Terms

• Machine Learning - Algorithms that learn from data rather than following fixed rules
• Training Data - The information algorithms learn from
• Model - The result of applying learning algorithms to data