OP-013
Catalogue improvement often involves activity that is easy to count.
An Operator may review 240 products, rewrite 80 descriptions, standardise 600 attribute values or complete compatibility information across an entire range.
At the end, they can show exactly what was completed. The assigned products were reviewed, the missing values were added and the pages were updated.
That can be useful evidence of what was done.
But the reason for doing the activity was usually not to produce 600 edited attribute values.
It was to improve something that affects how the Shopper finds, understands, compares or chooses products.
Perhaps more products needed dependable compatibility information. Perhaps a filter needed better attribute coverage. Perhaps important specifications were inconsistent. Perhaps Shoppers were being left to resolve distinctions that should have been clear from the available product information.
The activity and the intended improvement are related, but they are not the same thing.
An Operator can know exactly how much was changed. They may still have a much weaker basis for saying whether the change actually made things better for the Shopper.
That matters whenever catalogue optimisation is treated as more than a sequence of completed tasks.
When the Operator changes a catalogue with the intention of making it better for the Shopper, what gives them a dependable basis for deciding whether it has actually improved?
Consider a range of replacement parts where compatibility information has become inconsistent.
On some product pages, the machines the parts fit are listed. Other pages contain only partial compatibility information. Some products have no dependable compatibility information at all.
The Operator reviews 240 products, adds 96 compatibility records and corrects another 18. Every assigned product is checked.
At that point, the review is complete.
But suppose the original reason for the project was that Shoppers could not reliably establish whether some parts suited their machines.
The number of products reviewed does not establish whether that problem has been resolved.
There may still be products with unresolved compatibility information. Some newly added values may still require confirmation. Other products may have been reviewed correctly and left unresolved because there was not enough dependable information available.
The completion figure remains accurate. It is an output measure: it tells the Operator what activity was completed. An effectiveness measure answers a different question: Did that activity actually produce the improvement it was intended to produce for the Shopper?
A measure records or checks something specific. It becomes evidence when the Operator can use it to support a judgement about whether the intended improvement occurred.
The distinction becomes clearer when different catalogue changes are compared.
Suppose an Operator is improving a category where voltage materially affects which products a Shopper can choose.
There are 120 applicable products.
Before the review, only 73 contain dependable voltage values that support consistent filtering.
The Operator then corrects missing and inconsistent values.
After the review, 117 products contain dependable voltage values and three remain unresolved.
The Operator now has more than evidence that products were edited.
They have a basis for comparing what the activity was intended to improve.
Dependable voltage information increased from 73 of the 120 applicable products to 117.
Dependable voltage values are now available for filtering on 117 of the 120 applicable products when the Shopper narrows the range.
That does not tell the Operator everything about the category.
It does tell them something specific about the improvement they were trying to make.
Now imagine that the same project also involved rewriting 40 descriptions.
That figure may be accurate, but it would be a poor measure of whether voltage filtering improved.
A measure can therefore be precise without being relevant.
What matters is not simply whether something can be counted. The measure needs to relate directly to what the Operator was trying to improve for the Shopper.
That does not mean every catalogue change needs a formal baseline, analytics system or before-and-after report.
Some improvements already have a dependable reference point.
The machine’s weight is shown as 31 kg on the product page.
The Operator checks the source information and confirms that the correct weight is 24 kg.
The page is corrected to 24 kg.
The Operator does not need to create a numerical quality score before deciding whether the product page now shows the correct weight.
The confirmed value already provides the basis for the judgement.
The Shopper now has access to the correct specification rather than the incorrect one.
The same applies elsewhere.
A missing attribute or incorrect product relationship may have a known requirement.
Compatibility information or classification may be checked against a dependable source.
Evidence does not always need to come from a numerical measure, dashboard, score or formal reporting system.
The Operator needs something dependable to compare against: what was there before, what should be there, or both.
Sometimes that basis is a count.
Sometimes it is a repeatable assessment.
Sometimes it is a known requirement or confirmed value.
The right evidence depends on what the Operator needs to establish.
Return to the voltage example.
Increasing dependable voltage coverage from 73 products to 117 is good evidence that the relevant part of the catalogue improved.
It also establishes that more of the applicable range now has dependable voltage values available for filtering.
But those figures do not establish that Shoppers used the filter more often.
They do not establish that Shoppers chose products faster.
They do not establish that conversion increased.
Those are different claims and would require different evidence.
The reverse is also true.
A rise in sales after a catalogue change would not, by itself, tell the Operator which part of the catalogue improved. Price changes, promotions, stock availability, traffic and other factors may also have changed.
The Operator therefore does not need one measure—or one kind of evidence—to prove everything.
The evidence needs to match the question they are trying to answer.
If the question is whether voltage filtering now covers more of the applicable range, check how many applicable products now have dependable voltage values.
If the question is whether a known specification was corrected, compare it with the dependable value that should apply.
If the question is whether Shopper behaviour changed, catalogue evidence alone is not enough.
This keeps the improvement judgement both useful and proportionate.
A completed catalogue change and a demonstrated catalogue improvement are not the same thing.
When a change is intended to make the catalogue better for the Shopper, completing the change is not enough. The Operator also needs evidence that the thing they set out to improve actually became better.
That evidence needs to relate to something that affects what the Shopper can find, understand, compare or choose.
It may show an improvement from the earlier position, or show that the catalogue now meets a dependable requirement.
The evidence does not have to take one form.
A before-and-after count may be appropriate for attribute coverage. A confirmed requirement may be enough to judge a corrected specification. Another improvement may need a repeatable assessment.
What matters is that the evidence relates to the improvement being claimed.
Completing the activity establishes that something was done.
It does not, on its own, establish what became better for the Shopper.
Before beginning a meaningful catalogue improvement, the Operator can make two things explicit:
What part of the Shopper’s ability to find, understand, compare or choose products is this change intended to improve?
and:
What needs to be checked to determine whether that improvement occurred?
Suppose a category contains 300 replacement parts and compatibility information is known to be inconsistent.
The practical question is not only:
Which products are we going to review?
It is also whether the relevant products have dependable compatibility information when the review is finished.
The Operator may establish how many applicable products currently have dependable compatibility information and compare that with the position after the review.
If an established requirement already exists, that requirement may provide the reference point instead.
The Operator can then assess whether the problem that justified the activity has actually been reduced or resolved.
That does not require measuring every correction or building a reporting system around ordinary maintenance.
If one incorrect specification is corrected against a dependable source, the basis for the judgement may already be sufficient.
Completing the planned activity should not be the only basis for saying the catalogue improved.
The evidence only needs to be strong enough for the claim the Operator is making.
This gives the Operator a firmer basis for deciding when an improvement is finished.
If the relevant part of the catalogue now meets the requirement, further change may not be necessary.
If an important gap remains, the improvement is not complete simply because the planned activity is.
If the available evidence is still insufficient, the Operator can recognise that they still do not have enough evidence to judge the result.
That makes repeated catalogue optimisation easier to judge against what each change was intended to improve.
The Operator does not have to treat every completed task as proof of progress.
Nor do they need sales or conversion evidence for every catalogue correction.
They can use evidence that matches what they are responsible for improving.
Some improvements directly affect the buying decision—for example, changes to compatibility information, filtering or specifications.
In those cases, the Shopper may have a better basis for finding, understanding, comparing or choosing the relevant products.
The Operator does not need to claim more than the evidence supports.
They need enough evidence to know whether the change actually improved the part of the buying decision it was meant to support.