Invest in the website first, or in advertising? How to spot the bottleneck

Should we redo the site or put money into advertising? It's framed as a choice, but it's actually a diagnostic question: where does the neck narrow. If the problem is that no one's showing up, a new site won't fix it. If the problem is that whoever shows up doesn't understand what you sell, buying more audience just shows the same misunderstanding to more people. The answer comes from reconstructing four numbers and seeing where the biggest share gets lost.

The chain: relevant visits, requests, opportunities, sales

You don't need a sophisticated measurement system. You need four numbers for the same period, defined before you go looking for them.

When the four numbers are laid out in a row, the step with the sharpest drop is the bottleneck. Few relevant visits: an audience problem. Lots of visits and very few requests: a path problem. Lots of requests and few opportunities: a problem with what you're attracting. Lots of opportunities and few sales: the problem isn't digital.

Three scenarios, and what they tell you

Few people show up, but the right ones

Visits are few, and requests, proportionally, are there: whoever shows up writes in. This is the scenario where advertising makes the most sense, because you're buying an audience for a path that works. First, though, a check: can your response capacity and production capacity handle an increase? If you're already taking three days to respond today, the first effect will be paying to lose more opportunities.

Lots of people show up, but the wrong ones

The visits are there, the requests aren't, and looking at which searches bring people in, you find they're about definitions, ballpark prices, DIY, or a similar but different product. The site isn't broken: it's answering questions that aren't yours. Buying advertising here is the worst choice, because it amplifies a mismatch. The work is on the content and the service pages, not on the look.

Requests come in and get lost

The number of requests is reasonable, but no one knows what happens to them: different people answer, quotes go out days apart, there's no list. This is the most common scenario at companies that are doing well and are busy: the bottleneck is downstream, and no change to the site touches it. Here, the highest return comes from a shared list of requests with a response date next to each one, not an investment.

Three checks about the company, not the site

Before choosing between the two paths, three checks that cost an hour and change the answer.

1. Can your offer be said in one sentence? Try writing it: what you do, for whom, what problem you solve. If three different sentences come out of the company, the site won't be able to be any clearer than that.
2. Who responds, and how fast? Take the last ten requests and measure the actual response time, not the one it feels like.
3. Do the contact channels actually work? Send a request through the form, from a phone and from a computer, and check that it arrives where it should. Then call the published number. These are trivial checks that every so often reveal a channel that's been silent for months.

Don't blame everything on design or traffic

The two most popular explanations are also the most convenient, because they point to an outside culprit. Before picking one, consider what's in between.

• Seasonality: many industries have dead months and busy months. Comparing two different periods measures the calendar, not the site.
• Price and terms: if they've changed, the path has nothing to do with it.
• The local market: a new competitor, a lost bid, a big customer that was absorbing capacity.
• Capacity: some companies aren't losing opportunities, they're turning down work. In that case, the right investment isn't digital.
• The type of purchase: if the decision takes months and involves several people, today's result depends on what was done a long time ago.

A bounded test, decided in advance

Whichever path you take, it's worth treating it as a trial with written boundaries, not as a change of direction.

1. One single question: what you want to know at the end. It's not working, but is it bringing in requests we can actually serve?
2. A budget you can lose entirely without argument, decided at the start and not adjusted midway.
3. A duration tied to volume: if you receive few requests a month, two weeks won't tell you anything. Better to admit it beforehand than afterward.
4. A reading criterion written down before you start: which number you're watching and at what level you'll count it as a result.
5. What you do for each outcome, including the outcome "it's unclear," which almost always means the period was too short or the signal was mixed with other changes.

One change at a time. If you redo the service pages and turn on advertising in the same week, you'll get a result and no explanation, and when next year's question comes up about where to put the money, you won't be able to answer it.

What this guide doesn't cover

This guide is about choosing which step to work on first. Diagnosing a single page — why that page isn't bringing in requests, and what's missing from the promise, the proof, or the form — is covered in the guide on a site that gets visits but few contacts. The link between search data, citations, and acquired customers has its own guide.