The 7-changes-a-day rule: why our agent limits itself

When people hear that TinyRank makes at most 7 changes a day on a client's site, the first question is usually: why so few? The agent has already audited the whole site. It knows about the 60 problems. It could fix all 60 before lunch. Why make an agency wait weeks?

The answer is that fixing all 60 before lunch is the worst thing it could do, for the site, for the agency and for the client relationship. Let me go through it.

Nobody can read 60 changes

TinyRank's whole pitch is that an agency can see exactly what was done, with the old value and the new value, and undo anything it disagrees with. That promise only works if a person can actually read the log.

Seven changes is a short list. An agency owner can open a client's page, read seven lines, nod, and close it. Sixty changes is a wall. Nobody reads a wall. They skim the first few, assume the rest are fine, and the moment the tool is unsupervised in practice, the undo button is worthless because nobody knows what to undo.

I have watched this happen with other tools. People say they review the changes. They do for a week. Then they stop. Seven a day keeps the list short enough that the habit survives.

Google does not reward sudden change

If a site that has not changed in two years suddenly rewrites every title, every description and half its internal links overnight, the crawler notices. Not in a sinister way, just in the ordinary way that a sudden large change to a site is a signal, and Google's reaction to signals is to take time and work out what happened.

Spreading the same fixes across a few weeks looks like what it is: a site being gradually maintained. Each day's changes get crawled, settled and understood before the next batch lands. If something goes wrong, you also know which day's seven changes caused it.

A small blast radius

Every change TinyRank makes is reversible and checked on the live page afterwards. But I built the product assuming that one day a change will be wrong anyway. Maybe a page title has a nuance the agent missed. Maybe a client has a reason for a strange heading.

If that happens on a day with seven changes, the damage is one of seven, caught the next morning when the agency reads the log, undone in one click. If it happens on a day with sixty changes, nobody finds it for a month.

What the number is based on

Seven is not magic. It came from looking at the backlog on the first sites we ran it on and asking two questions.

First, how long should a typical backlog take to clear? The six sites in our first week had between about 10 and 40 open problems each after the first audit. At seven a day that is one to six days for the technical fixes, which feels right. The site is noticeably better within a week, and nobody is waiting months.

Second, how many items can an agency owner check in the time they actually give a client's SEO each day, which is about two minutes? Seven lines of old value and new value, at a glance, comfortably.

Five felt slow. Ten was more than people read. Seven it is.

The order matters more than the number

Because the agent only does seven a day, it has to choose. It works through the most valuable problems first.

Roughly, that means problems on pages that already get traffic come before problems on pages nobody visits. Problems that affect whether a page can be indexed at all come before cosmetic ones. A missing title on the homepage comes before a slightly long title on a 2019 blog post.

So the first week's seven-a-day are the changes that matter most, and the long tail of minor things trickles out over the following weeks. If you only ever read the first week's log, you have seen the important part.

Some days it does nothing

This is the part people find odd. If the queue is empty, the agent does nothing that day. It does not go looking for something to change to hit a number.

A tool that has to produce seven changes every day will eventually start making changes for the sake of it: rewriting a perfectly good description into a slightly different perfectly good description, so there is something to show. That is activity, not work, and it erodes trust in every real change in the log.

So some days the log says nothing happened on this site, and that is the correct answer. The site is re-audited every 30 days, and new problems show up as the client adds pages, so the queue refills on its own.

What does not count against the seven

The seven are changes to the site itself: page edits, titles, descriptions, alt text, links, plugin updates. Business directory listings, which TinyRank builds on weekdays, do not count, because they happen on other websites, not the client's. Backlinks, released every 30 days, do not count either. Recommendations that need agency approval, like rewriting a thin service page, sit in a separate queue and are not limited.

If you want it faster

You cannot turn the limit up, and I do not plan to add that option. If a site has a genuinely urgent problem, like the whole site set to noindex, the agent prioritises it on day one. Everything else can wait a few days, and the agency and the client are both better for being able to read what happened.