The Leaky Bucket

In early 2024, Social Habit started as a content and website agency. We shot the videos, ran the social accounts and built the sites. Clients liked the work, and many of them left anyway. They hired us for content and a website, but they evaluated our work on the calls, bookings and sales that came after. When those did not show up as a number they could see, our work looked like a nice-to-have that did not translate into sales, and the contract ended at renewal.

We called it the leaky bucket: a retention problem. We signed new clients at the top and lost them out the bottom, and each lost client took with them what we spent to win them. Selling harder filled a leaky bucket faster. It did not fix the leak.

When clients are asked why they leave agencies, the answers are consistent. Setup, an Atlanta firm that matches brands with marketing agencies, surveyed more than 400 brand and agency marketers in 2024 and asked brands why they ended agency relationships. The top answer, from 48%, was dissatisfaction with delivery. Dissatisfaction with value tied for second. A client does not end a contract over the quality of a video. A client ends it when the agency cannot show what the money bought.

The first task was finding out what our clients’ businesses needed, which was not always what they asked us for. A dental practice asks for a new website. The practice needs new patients. A website is where a patient decides to book, and it does nothing for a patient who does not arrive. Search and organic content could bring those patients in, but that took months, and many clients did not stay for months.

Across the web, a new page starts with almost no visitors from search. Ahrefs, a company that makes search-engine research tools, checked about 14 billion pages in its index in 2023 and found that 96.55% got no traffic from Google at all.

The missing piece was traffic we could buy, aim and count. Paid ads send a chosen audience to a chosen page on a chosen day, and the platforms report what each visitor did next. Content builds attention slowly and pays off over months. Ads rent attention now and show what it was worth. A client who gets both sees the content working and sees the leads it produced.

We weren’t alone in the struggle, and the pattern of the problem was historically evident. In 1990 Frederick Reichheld of Bain & Company and Earl Sasser of Harvard Business School studied customer defection in service businesses for the Harvard Business Review. Cutting defections by 5% raised profits 85% in one bank’s branch system, 50% in an insurance brokerage and 30% in an auto-service chain. A client who stays costs nothing to win again. An agency spends weeks of calls, proposals and onboarding to win each client, so the same math applies.

My side of the marketing funnel was the bottom: the website, where a visitor decides to call or book. I designed and built those pages, and I knew what makes a page convert. The business demanded the top of the funnel too, the ads that bring strangers to that page, and I did not know ads. Learning a second trade that late felt like a step backward. It let me see the whole funnel, from a stranger scrolling past an ad to a client’s phone ringing. Before we sold ads to anyone, we had to name exactly what we needed and price it honestly.

The price came in three parts. The first was my time. We started with the point of highest leverage: Meta, the Facebook and Instagram ad platform where our content converted. I learned it myself, on our own campaigns and after hours, because I could not hire someone for work I could not judge. The second was client money. Ad budgets belong to the client, so every mistake made while learning spends someone else’s dollars. The third was the price of a lead. WordStream, an ad-software company, put the 2025 average cost per lead on Meta at $27.66, with $16.61 for real estate and $76.71 for dentists. A client spending $1,500 a month buys about 54 leads at the average price. Whether 54 leads is a good month depends on what one customer is worth to that client, and we needed that number before we promised anything.

My website skills turned out to be the advantage. The same WordStream data put the average conversion rate for Meta lead campaigns at 7.72%. A landing page that converts at 10% instead cuts the cost of every lead by almost a quarter. Most ad buyers send traffic to whatever page the client already has. We built the page. Ads and web together sold one number a client could hold us to: cost per lead.

The tracking had to work first, so we started where we controlled the most. The first campaigns went to clients whose websites we built. We owned the landing page, the forms and the tracking, so a bad result pointed at the ads and not at a page we could not change. Tracking went in before any money went out. Every campaign launched with conversions tracked on the forms and the phone buttons, so the first dollar produced a number.

The content side of the agency fed the ads: the videos we already shot for our clients every month became their ad creative. An ads-only shop has to buy or beg for creative. We had a camera crew on the account. The two services stopped being separate line items and became one system: content made the ads, the ads brought traffic, and the site turned traffic into leads.

Every day we checked for a broken funnel: tracking that stopped firing, spend running ahead of pace, ads the platform rejected. We split watching from changing, because Meta’s delivery system needs about 50 conversions in seven days to leave its learning phase, and large edits send the ad set back to the start. A manager who edits every morning keeps a campaign stuck in learning and pays the higher price of learning the whole time. Once a week we changed strategy, based on cost per lead. Once a month we reported to the client in the client’s own numbers: leads, cost per lead and the customers those leads became.

The monthly report plugged the leak. A client who reads a line like “42 leads at $61 each, 9 became patients” knows what the agency did. The content and the website finally had a number attached to them.

I had to get myself out of the work. I could not stay the ads department and run operations at the same time. Andy Grove, who ran Intel, wrote in High Output Management that the output of a manager is the output of the teams under that manager’s influence. An hour I spent in one client’s campaign improved one client’s results. An hour I spent writing a playbook or training a hire improved every account that person touched.

Delegation came last on purpose. Learning ads myself is what let me write the playbooks, hire the first people and tell good work from bad. The assessment routine made the handoff safe. The new team ran the same daily, weekly and monthly checks, and the numbers showed within days if something slipped. Grove warned against handing off work and walking away, which he called abdication. We handed it off and kept watching the numbers. Paid ads is now a standing department at Social Habit, running Meta and Google campaigns for clients without me in the accounts.

The bucket leaked because we sold effort and clients bought results. Paid ads did not replace the content or the websites. They showed each client our contribution to the bottom line: the leads, what each lead cost and the customers those leads became. A service business keeps a client when the client can count what the service produced. Find the number the client cares about, learn the skill that produces it well enough to judge it, measure it every week, then hand it to people who do it every day.