Field notes: what we changed this week

Every Friday we write down what actually changed in the work: not the plan, not the theory, the specific thing we did differently and why. Most weeks the changes are small. This week a few of them were worth writing up because they came from being wrong about something, which is the only kind of change worth reporting.

We moved a client's lead form above the fold, then moved it back

A home services client had a contact form sitting halfway down the page, below a block of trust badges and a short "how it works" section. Standard advice says get the form higher. We moved it up. Conversions dropped about a fifth over the next nine days.

The reason turned out to be obvious in hindsight. This is a considered purchase, several thousand dollars, scheduled weeks out. People do not fill out the form until they believe the company can do the job. The "how it works" section and the badges were doing the convincing. When we shoved the form up top, we asked for commitment before we had earned it.

So we put it back, and instead added a second form at the bottom for people who had already read everything. Two entry points, no argument about which one is right.

The lesson we keep relearning: "above the fold" is a rule for low-consideration actions. For anything expensive or slow, the page has to do persuasion work first, and the form should sit where the persuasion ends.

We killed a weekly report nobody read

We had been sending one growth-stage client a Monday morning summary email: traffic, top pages, channel breakdown, a few callouts. We built it early in the relationship because it felt responsible. Then we noticed the open timestamps clustered around our own check-ins, meaning the client mostly opened it right before a call, to look prepared.

That is a report doing zero work. It was not driving a decision, it was performing diligence.

We replaced it with two things:

• A short async note only when a number moved enough to matter, with the decision it implied ("CPL on the paid search brand campaign is up 30 percent week over week, recommend we pause the broad match ad group, need a yes by Thursday").
• A live dashboard the client can open whenever they want, with the vanity numbers left off.

The volume of communication went down. The number of actual decisions per month went up. That is the trade we want.

A report that only gets opened right before the meeting is not information, it is theater, and you are paying for the actors.

We changed how we scope the first 30 days

We used to open engagements with a discovery phase: interviews, an audit, a positioning workshop, then a plan. It reads well in a proposal. In practice it front-loads a month of activity that produces a document, and the client is paying (or, in our case, being measured) against outcomes that have not started yet.

This week we rewrote our default onboarding so that discovery and shipping happen at the same time.

What that looks like in practice

Instead of a clean sequence, the first 30 days now run two tracks in parallel:

• Track one, learning. The interviews, the analytics review, the competitive read. This still happens, it just does not block anything.
• Track two, shipping. We pick one thing we are already confident about (usually a broken conversion path or an obvious email that is not being sent) and we build it in week one.

The point is not speed for its own sake. The point is that the fastest way to learn how a business actually converts is to change something and watch. A workshop gives you the client's model of their customer. Shipping gives you the customer's actual behavior. The second one is more honest, and you can only get it by doing.

We are keeping the discovery work. We just stopped treating it as a gate.

We added a check to the site grader for a mistake we kept seeing

Our free grader scores a page on the things that move outcomes. Running it across a batch of local service sites this week, one pattern kept showing up that we were not flagging clearly enough: pages that load a phone number in an image or inside a script that fires late, so the number is not in the initial HTML.

Why it matters, in order of who it hurts:

• Click-to-call on mobile does not work if the number is not a real tel: link.
• Some accessibility tools and some AI crawlers reading the page will miss the number entirely.
• If the number is rendered by a slow script, a fast bounce means the visitor leaves before your primary contact method appears.

For a business where the phone is the money, this is not a nice-to-have. So we added an explicit check. The grader now looks for a real, clickable phone number in the delivered HTML and calls it out when it is missing or buried. Small thing, but it was one of those problems we had been fixing manually for clients while never surfacing it for everyone else.

What did not change, on purpose

Two things came up this week that we decided to leave alone, and leaving things alone is a decision too.

A client asked us to start posting daily on a platform where their buyers do not spend meaningful time. We said no. Adding a channel because it is available is how teams end up busy and flat. The bar for a new channel is evidence that the audience is there and buys, not that a competitor is posting.

We also got asked to add a chatbot to a site that gets low traffic and closes deals over the phone. A chatbot solves a volume problem. This client has a trust problem, not a volume problem. More automation would have made the site feel less like the small, responsive company it actually is.

The pattern in both: the tempting move was addition. The right move was restraint. Most weeks, the highest-leverage change is subtracting something that is quietly costing you.

If you want a second set of eyes on where your site is quietly leaking, run the free grader, it now catches the phone number problem above. And if you would rather have a partner whose pay depends on the same outcomes you care about, you can apply for partnership.