The Year the Tools Changed

The year in six events

In March a considerably more capable general-purpose model arrived, followed within a week by a competing one from the largest search company.

In May, generated answers began appearing above the results for some searches, initially in one country and only for people who opted in.

On the first of July, the old analytics stopped collecting data.

In July a new social network launched, tied to an existing one, and a fortnight later a large platform renamed itself.

What each one required of you

• The models: nothing, unless you chose.
• Generated answers: nothing yet.
• The analytics: a forced migration.
• The new network: nothing.
• The rename: an afternoon of edits.

One of the five was compulsory and had a date attached. The other four were optional, and three of them are still optional, which is a different thing from unimportant.

The analytics cutover was the only deadline

Which is worth stating plainly against a year of louder news.

It had a date, it was announced well in advance, and on that date the old system stopped collecting.

Businesses that had migrated early lost nothing, and businesses that had not lost the ability to compare to any prior year.

That is the shape of a change that actually matters: a fixed date, a concrete loss, and no way to opt out.

Almost nothing else that happened this year had any of those three properties.

The volume of discussion about a change turns out to be a poor guide to whether it required anything at all.

If anything the relationship runs the other way, since a change with a date needs no persuasion.

A worked example

A small business kept a list of things they were told they needed to respond to this year.

By December it had eleven items, gathered from suppliers, newsletters, and a conference.

Reviewing it, one had a deadline they had met, two had produced small improvements they were glad of, and eight had changed nothing about the business.

Of the eight, six had generated more anxiety in March than the analytics deadline had in June.

The list took ten minutes to review and changed how they read the next year's announcements.

Keeping the list all year was the whole of the method.

The models are a tool question, not an event

Which is the honest framing after nine months.

They are genuinely useful for drafting, summarising, and getting started on something, and they are unreliable on facts about your own trade.

Nothing about them required a small business to act in March, and using one well is a small ongoing improvement rather than a transformation.

The businesses that got value from them this year did so by trying one on a specific recurring task and keeping it if it helped.

That is how any tool gets adopted, and the volume of discussion this year obscured how ordinary the adoption question is.

Nothing about that will have changed by next December, whatever the models themselves do.

The adoption question is the same one every tool has ever posed.

Generated answers are the one to watch

Which is different from the one to act on.

Answers appearing above the results change how many people click through to any site, and that is a real effect on a real channel.

What makes it a watching matter rather than a doing matter is that it was limited, optional, and not available here for most of the year.

The sensible response is to know what share of your enquiries come from search, so that if the effect arrives you can measure it rather than guess.

That measurement is worth having regardless and takes an afternoon.

Everything else being recommended in response to it is speculation.

What did not change at all

Which is the larger part of the year.

Customers still want to know what you do, where, at roughly what price, and how to reach you.

Sites still load slowly, forms still ask for too much, and telephone numbers are still images on some sites.

The businesses that improved this year mostly did so by fixing those, and none of it appeared in anybody's year in review.

That has been true of every year covered in this series and it was true of this one.

Four years of writing these has not once produced a different conclusion.

Keep the list yourself

Which is the practice worth taking into next year.

Every time somebody tells you a change requires a response, write it down with the date.

In December, read the list and mark each one: had a deadline, produced a benefit, or changed nothing.

After two years of doing that you will read announcements differently, because you will have your own evidence about how often they matter.

That is a better filter than any advice about which things to take seriously.

The counter-case

Some businesses were genuinely affected.

Anybody selling writing, illustration, or basic research had a materially harder year, and treating the models as a tool question understates that.

A business heavily dependent on the renamed platform also had real work to do and real uncertainty to sit with.

And dismissing everything as noise is its own error, since the analytics deadline was real and some businesses did miss it.

Judge each change by whether it has a date, a concrete loss, and no way to opt out, and keep your own list to check that judgement against.

Three properties, applied consistently, will sort a year of announcements in about ten minutes.

The list is what makes the judgement improve rather than repeat.

What to do

1. List what you were told to respond to.
2. Mark each: deadline, benefit, or nothing.
3. Check the analytics migrated properly.
4. Know your search share.
5. Try one model on one task.
6. Fix the ordinary things.
7. Start next year's list.

Step two is the whole of the method, since a year reviewed honestly usually shows one item with a deadline and eight that changed nothing at all.

Last year's version is at three things that shifted in 2022.

Frequently asked questions

West Coast Media Solutions Inc. provides web design, web development, hosting, digital marketing, and business consulting to organisations across Canada, drawing on more than twenty-five years in the field.