So far we have established why product matching accuracy is crucial for getting your pricing right. On the other hand, we all know that achieving 100% product matching accuracy is utopia. You may think to yourself ‘90% (or even 95%) accuracy that some providers offer seems quite good to me’. Let me explain why that’s definitely not the case.
You may think – only 5% of matches are wrong – I’ll work on these 5%, I’ll find correct matches and replace them. The image below illustrates this logic – all I have to do is find oranges within bunch of apples, and replace them with other apples.
Seems simple enough, doesn’t it?
I’m afraid you’re terribly wrong – because of one major misconception. Unlike the image above, you have no easy way of distinguishing wrong matches from good ones. It’s much more like a multitude of apples, where you know that 5% are rotten – but you simply do not know which 5% those are. And in order to find out which apples are rotten – you will need to examine them all. I will repeat this – in order to find wrong 5%, you will have to check not only 5% of apples – but you will have to check them all. This is a lot of work, and quite often defeats the purpose of the product matching.
The image below illustrates the above point – it’s a multitude of apples, and you simply cannot know which ones are rotten without checking them all.
It’s not that simple after all, is it?
Now, after bringing such bad news, I will say something that should make you feel better. Here at Product Matching AI we are very much aware of the above percentages, and that is why we strive for 99% (not 90, not 95 – but 99%) – this way our product matching becomes so reliable that you do not need to perform the quality of matches yourself – because that has been done by our team already. Remember – PM.AI is not only about AI – but also about humans, who perform the QA on product matches found by AI.