Manila · SM North EDSA & TriNoma

No single post moved the crowd enough for us to see it.

We lined up anonymised movement across two of Manila's biggest malls against every social post we could find. No individual post produced a mall-wide footfall response we could distinguish from normal variation — yet the same data clearly registers a mall-scale change when one happens. (Smaller or single-tenant effects can sit below what a location panel this size can resolve.) Here's the evidence.

—

See what the data can resolve ↓

Now look at the posts, day by day

—

Observed daily footfall (solid line) against what a normal same-weekday would predict (dashed line, with a shaded normal-variation band). Every captioned post sits on its date, sized by engagement. Pick a mall, then click a post to see the actual post and whether traffic responded beyond the band.

Mall

observed expected ± normal event promo organic biggest payday / holiday

Select a post dot to inspect it.

Every campaign, ranked

—

Posts grouped into campaigns (a burst of posts about the same thing), so repeated posts don't count as separate evidence. Best day (σ) is the campaign's single strongest day versus a normal same-weekday — if even that stays under +2σ, no day cleared the noise band. Sort any column; click a campaign to expand its posts. Filter by mall or type.

Mall

Type

Does engagement predict a footfall response?

—

Each post plotted by engagement (how far it travelled online) against how far that day's footfall deviated from normal (in standard deviations). If posts drove traffic, high-engagement dots would sit high. The shaded band is the ±2σ zone where a day is indistinguishable from noise.

The ten biggest posts, up close

The posts a marketer would bet on — highest engagement — each shown as footfall from 7 days before to 7 days after (day 0 marked). If posts drove footfall, you'd see a step at day 0.

What can this data actually resolve?

—

Before asking whether posts worked, know the instrument's resolution. A location panel is a sample: day-to-day it wobbles by hundreds of shoppers on its own. An effect has to clear that wobble to be visible.

Pooled across every event — the most powerful test

…and the averaged footfall path, day −7 to +7

Minimum detectable effect — 7-day campaign, cross-mall share of dwell ≥10 min visitors, 80% power

This is a campaign-scale detection floor. It is different from the per-day ±2σ band used later to flag one unusual day: a single day is far noisier than a 7-day window, so it needs a much bigger move to stand out.

—

The honest headline

—

Proof the instrument works: the renovation

—

While TriNoma was being renovated (Feb–May 2026), SM North's share of the two malls rose from 54.6% to 56.5% — i.e. TriNoma lost about 2 points of share while it was disrupted — then moved back toward baseline (55.0%) in the short 23-day post-period. The same system that finds no signal from individual posts clearly registers this mall-scale change.

Read shares carefully. The two malls are physically linked across EDSA, so this share partly reflects visitors substituting between them. A campaign that grew the two malls' combined traffic would not show up here at all — share only sees one mall gaining relative to the other. We use share because raw counts aren't comparable across the year (the device panel shrinks over time).

Who changed their behaviour?

—

During TriNoma's renovation, did its crowd change? Compositional shares (a percentage, so panel drift doesn't distort it) across before / during / after. Then: did the crowd on big event nights look any different? (It didn't.)

Reach by home distance

Cross-mall behaviour

Methods & caveats

What's behind every number above, in plain terms — so you can judge the conclusions rather than take them on trust.