How Agent is changing the way Island Living does their everyday marketing work

One weekend, clicks on one of Island Living's campaigns jumped twentyfold. With no obvious cause for the spike, Denise Vrontas, who runs paid ads for the company, turned to Agent for answers. In less than 10 minutes, not only was the source of the spike identified, but a fix was pushed live to their account—for Denise and the team at Island Living, Agent has quickly become the starting point for every investigation.

The problem

Numbers move for many reasons

Furniture sales move with the season, sale events, stock levels and big orders that can land days or weeks after the first click. Many customers browse online and buy in the showroom, making conversion attribution a difficult task, whenever a number moves suddenly, it could be a real problem, a tracking gap or nothing at all, and each of these scenarios calls for a very different response from the team.

Every answer means a deep dive

Finding out which means digging through locations, placements, devices, product feeds and change history across Google and Meta. For an in-house team that also runs sales events, stock and creative, unexplained swings can cost hours. Reacting to early numbers is risky too, because cutting a campaign that is quietly doing well hurts more than the dip that prompted it.

A number moving isn't news. What matters is whether it's real—and what's driving it.

The solution

A click spike, traced to its source

The alert came in on a Monday: clicks on a campaign built to bring people into the showroom had jumped about twenty times over the weekend, with nothing to show for them. Agent broke the traffic down by location, network and device, and found that 98.6% of the clicks came from phones in Batam, Indonesia.

Most came through cheap YouTube placements. The campaign was set up to reach people interested in visiting Singapore, as well as people actually there, so one more exchange with Agent switched their targeting settings to people physically present in the location—less than 10 minutes after the first question and a fix was pushed live.

The whole account, at once

When online sales looked low on a single day, Agent first checked whether the drop was real or a tracking gap. It then followed shoppers step by step from product view to payment and found the fall was sharpest right at the point of making a payment, which made the checkout the first thing the team needed to test.

In the same thread, Agent checked all 1,000+ products in their product feed, finding 217 that could not be served—most because the products had sold out, and 71 more because of page errors, another easy fix. Later, when the team spotted a campaign set to target Chinese speakers only, Agent checked every other live campaign in the Island Living account for the same setting and offered fixes where appropriate.

I want the root cause before making any changes—Agent finds it, and I make the call.

A new way of working

A weekend clicks spike that went from alert to fix in under 10 minutes, checking more than 1,000 products in a conversation, updating dozens of campaigns with a query—Asking Agent the why behind a particular change is now a regular part of how Island Living runs Google and Meta Ads, and pushing changes couldn't be more simple.

Just as useful is knowing when not to act. For furniture, 29% of the sales value from a click can arrive more than a week later. When the reported return on one Shopping campaign fell sharply, Agent looked at sales by the date they actually happened and showed they were holding steady, so the advised course of action was to hold rather than cut a campaign on a premature conclusion—letting Agent handle the busywork leaves more time for the team to focus on the work and decisions that matter.

Furniture sales tend to arrive late—reading the numbers too early is how good campaigns get cut.

Takeaways

• Trace a spike before you react. Break unusual traffic down by location, network and device first. The cause usually sits in one of them.
• Check whether a drop is real. Compare reported results with sales by the date they happened, and allow for sales that arrive late.
• Check every campaign, not just one. A wrong setting in one campaign is a good reason to check the same setting everywhere.