TWO DAYS
AT BOULÉ

THE ONLY THING THAT CHANGED WAS THE CAMERA

Same women. Same product. Same ballroom. Same ad account, same audiences, same offer. Before the shoot the ads ran on phone footage. After, they ran on this.

WHAT IT DID

The 46 days after the shoot, measured against the 46 days before it. Store figures from Shopify, cancelled orders excluded.

REVENUE, DAILY

The convention ran 6–11 July and sold almost nothing on its own: $733 across the 6th, 7th and 8th. Revenue starts when the footage gets deployed and the rebuilt store goes live on the 12th.

PAID MEDIA

I was already running these ads before the shoot, on the client's phone footage. Same hands on the account, same audiences, same products. That makes this a clean read on creative: swap the raw material, hold everything else, watch what moves.

Click-through and watch time dipped slightly on far greater volume. Those are shown because they are real. Frequency is reported per ad, not per campaign: campaign-level figures aggregate overlapping audiences and read alarmingly high. The typical ad was seen less often after the shoot, because 55 creatives shared the load instead of 13.

THE WINNERS

Five creatives carried this account. Four of them are still underfunded — the highest-returning ad in the whole window ran on $455.

EVERY WINNER HAS A HALF-LIFE

Return on spend by week, for each ad's own lifetime. The shape is the same every time: open enormous, halve within a fortnight, settle near break-even. This is what running out of footage looks like.

Cost per purchase, weekly across the account, tells the same story: $9.50 → $13.28 → $18.67 → $28.03 as the library aged. Then “25 for 25” launched on 13 August and the account jumped back to 8.12x the same week — on only three posts. New creative reset it, not posting harder.

THE REBUILD WE REVERSED

When the engine started fatiguing, the obvious move got made: pause it, launch a cleaner rebuild. That decision was reversed the same day. Here is what the data actually said.

WHO BUYS, AND WHERE

Where the money actually worked. Every figure below is from the same 46 days.

THE AWARENESS BUY, MEASURED PROPERLY

A fourth campaign ran alongside these on OUTCOME_ENGAGEMENT — optimised for reach, never for purchases. Scoring it on return on ad spend would score it by a metric it was not built to produce, so it is measured on its own terms here.

It also logged 7,377 link clicks against only 962 landing page views — a 6,415-click blind spot, the population that in-app browsers and tracking prevention swallow. Some of them bought later and were booked as direct or organic. That leak cannot be sized from the data, but it is large enough to understate what this campaign contributed. Meta does not expose follower attribution per campaign, so its share of audience growth is unprovable in either direction.

THE ORGANIC ACCOUNT CAME WITH IT

47 posts published in 46 days, all cut from the same two days of footage. None of this is paid reach.

Profile visits rising nearly 12x is the clearest signal in the set: people saw the work and went looking for the brand. Follower history is capped at 30 days by Meta, so the 9 July baseline cannot be recovered.

WHAT THESE NUMBERS DON’T PROVE

Everything above comes from the client's own Shopify and Meta accounts. Here is what an honest reading has to concede.

The cleanest evidence is August on its own: no convention, storefront settled, and return on ad spend still holding at 5.81x against 2.68x before, on 2.6x the daily budget.

SO WE STOPPED GUESSING

Two days of shooting fed this account for about six weeks before the library thinned out and efficiency slid. That is not a failure of the footage — it is the shelf life of any creative. So the shoot stopped being a one-off and became a schedule.