Compared to last year

A client emailed the Monday after Black Friday to ask how the weekend had gone compared to last year.

I could answer the first half of that. The chart I was looking at started in March, because March is when we launched their new store and stood up a new analytics property to go with it, and everything before March sits in a property belonging to the agency who had the site before us.

So the reply I sent had this year’s numbers in it, a paragraph explaining why the comparison wasn’t available, and an offer to go ask the old agency whether that account still exists. They took it well. They had also run the same promotion two years running and had no way to find out whether it went better this time.

Nobody threw it away

There’s a version of this story where an agency holds a client’s data hostage, and that version is real, I’ve heard people tell it. This wasn’t that, and almost none of the ones I’ve seen up close have been.

March went like this. Launch was midweek, the tracking had to be on the store before the store took an order, and getting a Google account created on the client’s side needed a person in finance who was out that week. Somebody on our team already had one open. So the property got made in our account, the client got added as a user, and everyone moved on to the other 40 things a launch has in it.

That isn’t a decision to own a client’s data. It’s a decision not to spend an afternoon on paperwork during launch week, which is a much smaller and much more reasonable thing. I’d probably have made the same call with the same tab already open.

And it works, which is the part that gets you. The client logs in, the client sees their numbers, and the arrangement is indistinguishable from the correct one right up until somebody asks a question that needs the year before.

Account, property, view

Google Analytics has three levels and only the top one is ownership, which I don’t think is obvious from anywhere inside the interface. Roughly:

Account: Our Agency          <- the thing that gets owned
  Property: theirstore.com   <- the thing the history is in
    View: All Web Site Data  <- the thing the client was given

Text

Being a user on the view is the same experience as owning the account. Same login screen, same numbers, same everything, so nobody finds this out by looking. They find it out by asking something the setup can’t answer, and for a store that lands in November.

Properties can be moved between accounts, at least in the ordinary case where both companies still exist and somebody at the other one answers email. It takes a few days of back and forth, it is never the most urgent item on anybody’s list, and so mostly what happens is nothing. Womp womp.

Three Novembers

The number they asked me for is the least of it.

The shape of a year is actually gone. When traffic climbs and when it falls off a cliff, what a normal day in February looks like so you can tell whether this one is bad, whether the thing marketing has been doing since spring did anything or whether the business would have grown anyway. None of that is one number. All of it is a couple of years of the same number.

It’s also what you point at when you want to spend money. A store that can show you three Novembers can make an argument about the fourth, and a store with eight months of data can make an argument about nothing at all.

It’s the data I ask for before I write anything, too. The whole point of the support matrix I was going on about in January was to stop guessing which browsers matter from blog posts and go read the client’s own numbers instead. Eight months tells you what people used this week. It won’t tell you whether a share is going up or going down, and going up or down is the part anybody would make a decision on.

In their name, on their card

So this is the thing I now try to do in week one instead of at the end. The rule for every line is the same. The account is in the client’s name with the client’s billing details on it, and we get added to it as users.

The domain, registered to them, with contacts who still work there. Everything else on this list can be rebuilt from scratch. This one can’t be bought back at any price.
DNS, which is often somewhere else entirely, and which is the thing nobody can find on launch day.
Hosting.
The repository, in an organization they own, with us as collaborators. This is the one I find hardest to argue for, because a repo feels like ours in a way a domain never does.
The CMS admin, belonging to a named person who works there, instead of a shared login that three agencies have used.
Analytics, at the account level, for reasons covered at some length above.
Everything else the business runs through. The mailing list, the payment gateway, the shipping account, the reviews platform. Anything with a login and a bill attached to it.

I already had a line about part of this in the README I was so pleased with last month, under who to call, where it says to write down whoever has the DNS login. Writing down who has it turns out to be one step behind making sure it’s them.

The end of a project is the worst possible time to do any of it. The budget is spent, everybody’s attention has moved to the next thing, and creating seven accounts is a strange favor to ask of somebody as the closing act of a relationship they thought was finishing. It gets put off. Then it stays put off, and a couple of years later somebody like me is writing a paragraph in an email about why a chart starts in March.

Do it in week one and the deliverable at the end is one page listing what already exists. What the account is, whose name it’s in, which email address the login is attached to, and where the password lives.

They could leave tomorrow

Doing this properly makes an agency very easy to leave, and there’s no point pretending that’s free. A client who owns all of their own accounts can go get three quotes next week and we’d find out about it when the transfer request arrived.

The argument that got me here isn’t a noble one. A client who feels stuck isn’t a happy client, they’re a client waiting for a reason, and my sense is they take the first one that comes along and then they tell people why. A client who knows they could leave and stays is choosing us on purpose, once a year. That’s a much better room to be standing in.

But mostly it’s just their business. The store is a thing they paid for, years of their own numbers are not a thing for anybody else to be sitting on. I’d rather say that plainly than dress it up as strategy.

As for the client who emailed on Monday, I did ask. The old agency still exists, somebody answered. It looks like the property can be moved once two people find 20 minutes for it. They’ll have this Black Friday to compare against next November either way. It’s everything before March that depends on whether somebody at a company we don’t work for feels like finishing an email thread, which is a strange place for a business’s first four years to be living.