Created by AI, Curated by Experts: The Loculyze Way of Analysing Mobile Signal Data

Why We Use Mobile Signal Data

Mobile signal data has been around for more than a decade, and it promises something retail property has always wanted: a consistent way to understand how people move, where they spend time, and how places compete.

But in many solutions, the outputs don’t pass the pub test. You’ll often see impressive visualisations and very precise-sounding numbers that fall apart the moment a property professional asks the obvious follow-up: “Does this reflect reality?”

Our founders pioneered the use of mobile signal data for retail property, and learned—often the hard way—where it breaks, where it misleads, and what you have to do to make it reliable. That’s why we’ve invested thousands of hours into building algorithms that take a dataset originally created for advertising, and transform it into investment-grade analytics.

When it’s analysed properly, the upside is enormous: you can quantify attraction, engagement, and competitive dynamics for any location, not just assets with a proprietary banking feed—and not just the neat, assumption-heavy story inside a traditional consultant report.

That’s what we mean by created by AI, curated by experts: models do the heavy lifting, and our retail + analytics team makes sure the outputs are conservative, explainable, and aligned with how real decisions get made.

Beware The Red-Flags and Pitfalls

Mobile signal data is powerful—but it’s a tough dataset. Providers that optimise for speed, scale, and “dashboard theatre” usually skip the hard parts. Traditional consultants trying to incorporate the dataset are often blissfully unaware of its dangers. If you’re assessing a solution (or sanity-checking a result), these are the common warning signs:

1. They won’t disclose the sample: “We cover ~70% of the population” can be technically true and still meaningless. What matters is how many unique customers and visits are actually observed for the specific place and time window you’re analysing—and how biased that sample might be.

   
   

2. Tiny time windows presented as certainty: “Last Thursday’s lunch surge” looks sharp, but often it’s just noise. If a tool happily compares “this week vs last week” everywhere, it may not be testing whether there’s enough sample to make that comparison statistically defensible.

   
   

3. Overconfident claims at very small footprints: The smaller the space (and the more complex the built form), the longer it takes to accumulate stable signal. If someone is confidently reporting performance for a single specialty shop or brand—or a specific level inside a large multi-level centre—treat it as a hypothesis, not a measurement.

   
   

4. Instant answers that ignore context: Many self-serve products work by pre-aggregating the world into small cells, applying a simple weighting, and summing whatever polygon the user draws. It feels precise, but it’s usually measuring raw detection, not customers.

   
   

5. “AI” as the explanation: If “how does it work?” is answered with “AI”, ask for the method. What was the model trained on? What assumptions are baked in? What checks exist against ground truth? If they can’t explain it clearly, be cautious.

   
   

So here’s how we do it—and why we’re comfortable being the opposite of “instant”.

1. We start with detection likelihood

Mobile-signal data doesn’t see all customers equally.

• Enclosed centres with long dwell times and high visit frequency get rich samples.

• Open-air main streets with fast flows and short stops get thin samples.

Loculyze trains a dedicated machine learning model for each destination, refreshed monthly, to correct for structural and behavioural differences in detection likelihood.

We don’t assume every ping is created equal. We model how likely it was to appear in the first place.

2. Time windows need statistical footing

Short time windows look sharp on a dashboard. “Last Thursday’s lunch surge” sounds precise. Often it’s just noise.

Instead of starting with “this week vs last week” and hoping it’s meaningful, we flip the process:

1. Start with a Moving Annual Total (MAT) to get long-run stability.

2. Then drill down to monthly, weekly, daily and hourly views, with the MAT as the anchor.

This prevents false confidence (or unnecessary panic) driven by sampling error dressed up as a bar chart.

3. We model journeys, not isolated pings

Seeing that someone visited a destination is one thing. Understanding where else they went—and what that implies about cannibalisation or synergy—is much harder.

The fast approach:

• Count devices in polygon A

• Count devices in polygon B

• Intersect the lists

• Call that “shared customers”

That systematically understates real cross-visitation because the chance of detecting the same device again (especially nearby, within a short time) can be much lower than people assume.

We apply probabilistic cross-visitation modelling to estimate the likelihood of observing devices along a journey path. The result is a more realistic view of trade areas and movement patterns—less coincidence map, more behavioural signal.

4. Built for confidence, not for speed

You can’t do any of the above properly in a few seconds or without a deep knowledge of the data and the industry.

• Modelling detection likelihoods per destination

• Reconstructing journeys

• Normalising time windows

• Validating model outputs against ground truth

That takes compute, iteration, and review.

So we don’t claim to be “instant”. A typical run takes a business day or two, because we’d rather deliver something you can take to an investment committee than something that merely renders fast in a browser.

5. Sample transparency, always

You’ve probably heard: “mobile data captures about 70% of the population.”

That level of coverage can happen for very large places over long periods, when enough signals accumulate to smooth out the unevenness. Most of the time, coverage is lower—and it varies heavily by location and time.

That’s why we put the underlying sample next to the modelled outputs in every module:

• Unique customers sampled

• Total visits sampled

If you can’t see the sample, you can’t judge reliability. And if we didn’t show it, we wouldn’t expect you to trust the numbers.

6. Prescriptive, not descriptive: the LoculRetail score

We’re not trying to drown you in charts and hope you find something interesting.

Our focus is squarely on the dollar value of a customer and visit :

• We combine visitation, engagement and customer value.

• We turn that into a single dollar-based index of expected sales.

• You use that index to compare and prioritise locations on one scale.

Well-managed centres should outperform the index. Under-managed or poorly positioned assets should lag it. Either way, you’re working from one clear, consistently applied valuation signal, not juggling six different metrics.

This is “created by AI, curated by experts” in practice: the models do the heavy lifting; our retail and analytics team make sure the outputs are simple, usable and aligned with how real decisions are made.

Proof: why aggregation isn’t enough

Here’s the uncomfortable truth for the simple aggregation/weighting of raw ping data used by many providers:

If mobile signal data was reliable enough to use raw, you wouldn’t see huge swings in coverage by location, time and behaviour.

But you do. That’s why it has to be modelled, not just counted and up-weighted.

To build our model we train mobile signal data against published traffic and turnover data across hundreds of centres and landlords. The question was simple:

Can a properly modelled view of visits and engagement match what the real world is reporting?

Loculyze Visits vs Reported Traffic

Across that portfolio:

• Correlation (Pearson r): 0.94, with the 95% confidence interval between 0.90 and 0.96.

  • Real-world traffic swings are mirrored almost perfectly by our model.

• R²: 0.95.

  • 95% of the variation in reported traffic is explained by our visits model.

• Loculyze / reported: 0.985.

  • On average we under-predict traffic by about 1.5%.

That under-prediction is deliberate. We deduplicate visitors where traditional traffic counters can’t. Counters see entries. We see people.

You do not get numbers like that by just aggregating raw-pings and multiplying by a constant “coverage factor”. The sample is too uneven. Without modelling, the errors compound.

Visits as a sales predictor

Even before we bring tenancy mix or segment-level spend into the model, a single deduplicated visit:

• Explains around 60% of the swing in annual turnover.

That’s a strong base signal for understanding value—again, only possible when you model detection properly instead of treating raw device counts as ground truth.

Loculyze Estimated Sales vs Reported Turnover

When we take the final step and estimate sales using our full model:

• Correlation (Pearson r): 0.96.

  • Estimated sales and reported turnover move almost in lockstep.

• R²: 0.95.

  • 95% of the variation in reported sales is explained by our model.

• Loculyze / reported: 0.97.

  • We under-predict turnover by about 3% on average.

Our sales index tracks real-world performance closely but remains conservative. Well-run centres beat it. Under-performers don’t.

Why this matters when accuracy matters

Unsophisticated analysis of mobile signal data—such as that delivered by self-serve platforms and traditional consultants trying their hand at data science—maybe fine for marketing teams looking for broad patterns or for answering “what’s happening roughly over here?”.

It’s not fine when the question is:

• Should we buy this centre?

• Where do we put the next store?

• Which assets deserve capex and which don’t?

Location performance is driven by:

• Detection likelihood

• Visit frequency and dwell

• Journey structure

• Time-window stability

• Sample size and bias

If the system isn’t explicitly modelling those, you’re getting fast answers based on incomplete context. At best it’s convenience analytics—good for a quick look. At worst its a dangerous and misleading decision engine.

Loculyze is built on the opposite philosophy:

Take the time to model reality properly, then give the client a clear, prescriptive answer—created by AI, curated by experts.

In exchange, you get:

• A per-destination model of detection likelihood.

• Noise-resistant time series anchored on MAT.

• Probabilistic journeys rather than coincidental pings.

• Full sample transparency.

• A single, comparative sales index you can actually act on.

If you’re choosing where to put your next store, which centre to buy, or how to reposition a portfolio, that trade—confidence over instant—is the only one that makes sense.