This question usually arrives in month two. Money has been spent, content has been made, sales look the same, and the natural question is whether any of it is working.
The problem is in the question rather than the answer. If sales are your only measure, you are watching the last link in a chain, and the last link is the slowest thing to move. Other things move before it, visibly, and those are what you should be looking at in the first two months.
Why sales move last
People do not buy the first time they see you. They see you, forget, see you again, register the name as familiar, then eventually need the thing and go looking for you by name. That path takes weeks or months depending on what you sell and what it costs.
Which means the work you do today does not produce its result today. And if month one is a death sentence for anything you try, you will kill the things that were working at exactly the same rate as the things that were not.
The early signals, in order
These move before sales do, ordered from earliest to closest to money:
- The question changes. From "what do you do" to "do you do this specific thing". That means the message landed.
- People repeat something specific back to you. A phrase or an idea you have been consistent about. This happens well before a purchase.
- Branded search goes up. People typing your name directly. This is the strongest early signal available.
- Referrals start. Somebody arrives saying a name you do not recognise sent them.
- The objection changes. From "I do not know who you are" to "it is expensive". A price objection is progress, not rejection.
That last one is routinely misread. When a buyer starts discussing price, they have finished with the question of why you and moved to the question of how much. That is forward motion even when the deal does not close.
The technical detail that makes the picture wrong
There is a documented reason owners feel their marketing is not working when it is, and almost nobody explains it.
Google Analytics documentation states plainly that all attribution models exclude direct visits from receiving credit for a conversion, unless the entire path consisted of direct visits.
Apply that to the sequence above. Somebody sees a post, remembers it, and two weeks later types your name into a browser and buys. That records as a direct visit, and the system is designed not to credit the post. The most effective thing you did is the least visible thing in the numbers.
There is a second detail. The lookback window. Google documents 90 days as the default for most key events, adjustable to 30 or 60. If your buying cycle is three months and the window is 30 days, the system is structurally unable to connect the first touch to the sale.
The conclusion is not that the numbers lie. It is that they describe what can be attributed, which is a smaller thing than what happened, and you need to hold that difference in mind while judging.
One question solves half of this
The cheapest measurement tool in existence: ask every new customer how they heard about you. Not in a survey. In conversation, while you are already talking to them.
The answers will be imprecise, because people forget and say the first thing that comes to mind. Even so, twenty answers give you a truer picture than any dashboard, because they capture the routes no system can see: somebody saw you in a group, somebody heard about you from a friend, somebody remembered you from last year.
Telling signal from noise
- The trend matters more than the number. One good week is not a signal. Three weeks in the same direction is.
- Compare against a period with comparable conditions. A month with a season in it against an ordinary month is not a comparison.
- If you changed several things at once, the result cannot tell you which one worked. That is not a measurement failure, it is a decision you made without noticing.
- A number that moves while nothing changed is usually noise.
The numbers that are actually worth it
If you are going to track numbers, keep them few and be able to defend each one. The four I use are set out in how to measure marketing results, which is the companion to this piece: that one is the numbers, this one is the signals that arrive before them.
And there is a trap that sits before all of it, which is measuring a marketing goal and believing it is a business goal. The difference is in marketing goals are not always business goals, and it explains half of all cases where the charts look good and the money is not there.
When to say it is not working
There has to be a threshold, or you are spending on hope. The one I use: after a period equal to twice your buying cycle, if not one of the five signals above has appeared, the thing is not working.
Notice the condition is not "no sales". It is "not one signal". Zero sales alongside sharper questions and a name people recognise means you are halfway. Zero on everything means you are in the wrong place or saying the wrong thing, and both are fixed by changing something rather than by waiting longer.
Honest limits
All of this assumes you are doing something consistently every week. If the work is intermittent, no measurement will help, because what you end up measuring is the intermittence.
Second, there is no month count here, because it varies entirely with what you sell. Somebody selling a low-priced product has a cycle of a week. Somebody selling a large service has a cycle of four months. Anyone who gives you a timeline without asking what you sell is guessing.
And if the real obstacle is that assembling the numbers eats a day every month, that is not a reason to stop measuring. Rivl has a practical treatment of that specific problem in reporting automation for a small business, written for small operations rather than large ones.