Composite · F&B · East Singapore
A three-outlet group living on delivery platforms
The situation: three kitchens in the east, a following that knew the food, and a P&L that had quietly become a platform business. Delivery apps brought the orders. They also brought the commission, the discounting, and a customer who never had to learn the restaurant’s own number. Paid social was a habit of boosting weekend posts. The website was a menu photograph and a link that bounced to the app.
What I found: branded search was cheap and neglected. Regulars were still Googling the name and then tapping the app result sitting above the restaurant’s own site. Direct booking did not exist in any form a tired person would finish. Reviews lived on the platforms, so even reputation was rented. Spend on Meta was reaching people who had never eaten there and sending them to a page that could not take an order.
The decision: stop treating the apps as the shopfront. Build a short order path on the restaurant’s own site, location by location. Move a slice of spend to branded search so that people already looking for the name landed on that path. Collect emails and numbers at the till and from the new order flow, then talk to those people about weekday sets and events. Keep the platforms, because they still fill the slack hours, but stop paying them to reacquire the already-converted.
What changed, in the pattern I see on this kind of account: direct and walk-in mix rose over a quarter into a range where the platform line was a choice. Cost of acquiring a returning guest fell because the CRM actually had names. The work is not magic. It depends on the floor agreeing to mention the direct path, and on dishes that still make sense off-platform. What I would do differently: I would insist on a simple SMS note the same week as the site, rather than waiting for a polished email design. Speed of the first owned conversation mattered more than the look of it.
Anonymised composite, assembled from the kind of F&B account I usually take on. Figures describe a pattern, not a named client.
Composite · B2B · Jurong
A B2B industrial supplier with a good sales team and an invisible website
The situation: a firm in Jurong selling into plants and contractors across the region. The salespeople were strong. They knew the products, they knew the procurement habits, and they closed. Marketing, such as it was, was a brochure site, a directory listing, and a handful of ads pointing at the homepage. Leadership felt that “digital” should be doing more, without a picture of what more would look like.
What I found: almost nobody who mattered used the site to choose a supplier. They used it, if at all, to check that the company existed after a referral. The ads were buying clicks from people who would never be a fit. There was no way to tell which enquiries were a genuine specification and which were students downloading a PDF. Sales ignored the web form because it wasted their afternoons. That feedback loop had been running for years in spirit, even if the practice is new.
The decision: stop buying generic traffic. Build a short set of pages around the products that actually close, written so a procurement engineer can scan them. Put a form that asks two qualifying questions and routes the rest to a phone number with a named person. Give sales a weekly list they trust, even if it is short. Use LinkedIn and a small search budget only on the terms that match those products, and accept that volume will look modest next to the old vanity report.
What changed, in pattern: sales started answering the web again because the junk fell away. A typical range on this shape of account is fewer leads and a much higher share worth a conversation, with the sales cycle still measured in months. What I would do differently: I would have sat in on sales calls in week one, not week three. The language on the pages should have come from those calls sooner. I also delayed a simple “reply within a day” rule while I fussed over tracking. The rule would have paid for itself first.
Anonymised composite, assembled from the kind of B2B account I usually take on. Figures describe a pattern, not a named client.
Composite · Clinic · Singapore
A specialist clinic paying for branded search it already owned
The situation: a specialist clinic, appointment-led, with a reputation that already brought people in by name. The media buyer, whoever had last held the account, had left branded search running at a healthy cost per click because the dashboard made it look like a high-converting campaign. Non-branded terms were a jumble. The site’s appointment path required a PDF and a phone call during hours that working adults do not have.
What I found: a large share of “conversions” were people who would have booked anyway, clicking an ad that sat above the organic listing the clinic already ranked for. The budget was buying its own patients and calling it performance. Meanwhile, people searching for the condition, not the clinic’s name, landed on a thin page and left. The phone number on mobile was not tap-to-call. After-hours enquiries went into an inbox that was checked when someone remembered.
The decision: cut branded search to the minimum needed to protect the name from competitors, and say so in the report so nobody panics at the drop in “conversions”. Move money into a small set of condition terms with pages written for those searches. Make booking possible without a PDF. Measure calls and completed forms as separate things, and look at appointment show-rate with the clinic manager rather than celebrating the click.
What changed, in pattern: reported conversions fell, which is the honest part, and the cost of a new patient from search became a number you could defend. After-hours forms started turning into morning callbacks instead of silence. What I would do differently: I would have walked through the booking path on a real phone in the first meeting, in front of the owner. Screenshots hide the tap that does not work. I would also have written the “why branded spend is shrinking” paragraph before the first monthly call, not during it.
Anonymised composite, assembled from the kind of clinic account I usually take on. Figures describe a pattern, not a named client.
Composite · Professional services · Singapore
A firm with no way to tell which enquiries were worth answering
The situation: a professional services firm, fee-earners already stretched, receiving a mix of serious mandates and time-wasters through a generic “contact us” box. Everything received a polite reply because nobody wanted to look rude, and the polite replies were eating the week. Paid search had been switched on in a burst of optimism and then ignored. The partners disagreed about whether marketing was working, because nobody could describe what “working” meant in hours and fees.
What I found: the form asked for a name, an email and a message. The message field attracted essays, job-seekers, and people who wanted a free version of the advice the firm sells. There was no qualification, no routing, and no record of which partner had answered what. Search ads pointed at the homepage. The homepage pointed at the form. The circle was complete and empty.
The decision: rewrite the form around the kind of mandate the firm actually wants, including a field for timeline and a field for the problem in one sentence. Publish two or three pages that pre-educate so that the wrong-fit reader self-selects out. Route the rest to a coordinator with a script for the first reply, and a weekly half-hour where a partner only sees the shortlist. Pause search until those pieces exist, then restart on a tight set of terms.
What changed, in pattern: total enquiries dropped. Hours spent on dead-end replies dropped further. The partners started seeing a list they were willing to call, which is the only marketing metric a firm like this should care about in the first quarter. What I would do differently: I would have asked for two real (redacted) enquiry emails in the first week and used them as the brief. Invented personas waste time. The inbox already had the research. I would also have been sterner about pausing ads on day one. Spending through a broken form is a way to feel busy.
Anonymised composite, assembled from the kind of professional services account I usually take on. Figures describe a pattern, not a named client.