The Change That Made Things Worse

Why reverting is hard

Not technically, which is usually straightforward if a copy was kept.

The obstacles are that somebody chose the change, somebody was paid for it, and undoing it says something about both.

There is also a genuine uncertainty, since the figures are rarely conclusive at small volume and it is always possible the change needs more time.

Which combines into a site carrying a change everybody suspects is worse, indefinitely, because nobody wants to be the one to say so.

Decide before you look

The mechanism that removes the awkwardness entirely.

Before making the change, write down what result would mean putting it back and when you will check.

Something like: if enquiries are down more than a fifth over four weeks, we revert.

Then the revert is a rule being followed rather than a judgement about anybody, which is a completely different conversation.

It also stops the goalposts moving, since a threshold decided afterwards is always just below whatever the figures show.

Two sentences written before the change is made saves the whole argument.

Give it long enough, and not longer

• Four whole weeks for most changes.
• Longer at low volume, up to eight.
• Shorter if something is plainly broken.
• Not indefinitely, which is the default.

The third is worth separating. A form that stopped working, a page that will not load on a phone, or an obvious error is not a test result, and waiting four weeks to confirm it is a mistake.

Revert first, diagnose after

The order that limits the cost.

Where the figures are clearly worse, put the previous version back and then work out why the new one failed.

Diagnosing while the worse version is live means every day of the investigation costs enquiries.

The previous version is a known quantity and returning to it is the safe position, from which you can take your time.

That requires having kept the old version, which is the practical prerequisite for all of this and takes one copy before you start.

Where no copy exists, reverting becomes a rebuild, which is precisely why the change never gets undone.

A worked example

A business redesigned their site, which looked considerably better and cost a reasonable amount.

Enquiries fell by about a third over the following two months.

The response was to assume it needed time to settle, then to assume the season, then to make further changes to the new design, none of which recovered it.

After five months somebody restored the old site from a backup, and enquiries returned to the previous level within a fortnight.

Comparing the two, the old site had the phone number in the header and a price range on the service pages, both of which the redesign had removed for visual reasons.

Those two elements were added to the new design, which was then restored, and it performed slightly better than the original.

The five months of waiting had been the expensive part.

Partial reverts are usually right

Since a redesign is many changes at once and rarely all wrong.

Rather than choosing between two whole versions, identify what the new one removed or moved and put those specific things back.

Contact details, prices, proof, and the position of the enquiry prompt are the usual culprits, because they are the elements a visual redesign treats as clutter.

That gets the benefit of the new design without the loss, which is better than either version.

Compare the two versions side by side and list what is missing rather than debating which is better, since the list is factual and the debate is not.

Say it plainly

The conversation, since it has to happen with whoever made the change.

State the figures and the threshold that was agreed, without editorialising, which is why writing the threshold down matters.

Frame it as information about this business rather than about their work, which is usually accurate: the change was reasonable and the audience responded differently.

Ask what they would change rather than announcing a revert, since a competent supplier will frequently identify the cause faster than you will.

And where the relationship is ongoing, a supplier who handles this well is worth keeping, since every supplier will produce a change that does not work eventually.

Keep the copy

The prerequisite, stated separately because it is the thing that gets skipped.

Before any significant change, take a full copy of the site and store it somewhere you can find it in six months.

Note what the current figures are at the same time, since reverting is useless if you cannot tell whether it helped.

Keep it for at least a year, since the realisation that something is worse frequently arrives slowly.

That is ten minutes before a change and it is what makes the change reversible rather than permanent.

Redesigns fail in a predictable way

Worth naming, since the pattern repeats across businesses that have nothing else in common.

A visual redesign is judged on how it looks, by people looking at it on a large screen with no particular question in mind.

The elements that get removed are the ones that look untidy: a phone number in the header, a price beside a service, a block of proof, a plain enquiry prompt.

Those are exactly the elements a visitor with a question was using, and their absence is invisible to everybody reviewing the design.

Before signing off any redesign, list what the current site tells a visitor and check each item still appears somewhere, which catches most of this before it goes live.

The counter-case

Reverting is not always right.

Some changes genuinely take time, particularly anything affecting how a site is found, where several months is normal before the picture is clear.

Reverting also has its own costs: another round of disruption, another set of addresses changing, and a team that stops proposing anything.

And at low volume the figures frequently cannot distinguish a real decline from an ordinary quiet month.

Write the threshold first, keep a copy, look on the agreed date, and prefer putting back the specific elements over reverting everything.

What to do

1. Take a full copy before changing anything.
2. Record the current figures.
3. Write the threshold and the date to check.
4. Look on that date, not later.
5. Revert first, diagnose afterwards.
6. List what the new version removed.
7. Put those elements back rather than the whole thing.

Step three is what makes the rest possible, since a threshold agreed in advance turns an uncomfortable judgement into a rule that somebody simply applies.

Knowing what changed and when is covered in keeping a record of what you changed.

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.