It’s surprising just how many of the data feeds currently available on the various affiliate networks are either partially or completely unusable. I suspect that the primary reason for this state of affairs stems from a simple misunderstanding about exactly what is required of a good data feed. In case my suspicions are correct, here is my attempt to explain the differences between a “good”, useable feed and a “bad”, unusable feed. Obviously the following are just my opinions, so all of the usual disclaimers apply.
Good Feeds Need Definitive, Consistent Categorisation
It might be helpful to think of a data feed as an extension of a merchants own website. A good feed, when processed correctly by an affiliate, can result in additional, highly-targeted web pages that advertise a merchant’s content to consumers that they would not have reached otherwise.
To allow this to happen, though, a merchant’s data feed – like their own website - must be well organised and categorised. No merchant would dream of organising their website by the same broad, meaningless categories that are all too often contained within their data feeds. An online clothes shop, for example, would never simply list page after page of “Clothes and Accessories”. Similarly, an online Camera shop would never simply list page after page of mixed up cameras, tripods, cases, and batteries. So why do so many of their data feeds do this? Broadly speaking, a merchant’s data feed should categorise its products in broadly the same manner as they are categorised on their own website.
Typically, affiliates take a data feed and automatically process it into a form suitable for advertising to consumers. The
automatically part means that data feeds are generally read by machines (computer programs) rather than people.
And there’s the rub. While a human being can read a description or look at a picture and easily tell what a particular product is, any automated (machine) process can’t. A “dumb” machine needs to be told exactly and unequivocally what each product is; it cannot easily discern it from a photograph or some prose in a description.
A product’s category fields, therefore, need to tell a “dumb” machine exactly what the product is. We don’t care whether this is encapsulated in a single category or by several hierarchical categories, just so long as by category alone we, and our dumb machines, can tell exactly what it is.
Sadly, in my experience, the vast majority of available data feeds fail, either completely or partially, to meet this basic requirement. I invite all merchants to take a look at their data feeds. If you can’t tell what your products are by simply looking at the category field/s, then your data feed is unusable. If you can tell what some of your products are, and not others, then your feed is partially unusable. It’s almost that simple.
E.g.
Category: Plasma Television (Very Good)
Category: Television (OK-ish)
Category: Televisions, Video, and Photography (Very Bad)
Good Feeds Need “Make” and “Model” Fields
With decent categorisation we can generally establish, for example, that a given product is a Plasma TV - so far so good. But most consumers want to know the brand of the Plasma TV they are about to buy, and a sizeable proportion of those want to know the model as well! I don’t know, maybe they read a review in a magazine or something. In any case, they are pretty stubborn about it. They don’t just want a Plasma telly; they want a Sony ZXHoojimaflip Plasma telly.
A good feed, therefore, needs both
make and
model fields. Again, it isn’t enough to bury this information within the prose of a description field. It also isn’t enough to combine them within a single field:
Make: Sony
Model: KXXXX7P (Very Good)
Name: Sony KXXXX7P (Bad)
Description: The amazing KXXXX7P from Sony is... (Very Bad)
Other Common Causes of Bad Data Feeds
• Mixing up Accessories for Products with the Products themselves. A wall-bracket for a Plasma TV is not a “Plasma TV”, it’s a “Plasma TV: Accessory”
• Other, more bizarre, “pollution” within categories – e.g. a Toaster within a “Plasma TV” Category
• ImageURLs that pull in images of hugely varying size – from tiny to absolutely enormous (these can wreak havoc on the layout of a Web page and are difficult to constrain)
• Missing out the additional contextual fields that are often esssential for the type of products being offered. Around 90% of Mobile Phone suppliers, for example, do not include the Monthly Tariff for their Pay-Monthly phones.
Turning a Good Feed into a Great Feed
In addition to the obligatory fields - such as category, make, model, price, description, image, and url - really good feeds also contain:
• Short Summary fields – e.g. “Plasma television with wall mount and speakers”
• “Availability” – is it in stock? How long to delivery? Etc.
• Delivery Cost
• In-store Price (if different from Web Price)
• Special Offer Text
And finally, the best data feeds are up to date data feeds. It should go without saying that a data feed needs to be as up to date as the Web data to which its URLs refer. Any stale, broken links or price discrepancies inevitably lead to lost sales for merchants and to lost commissions and credibility for affiliates. Moreover, prices tend to go down rather than up. If your data feed lags those of your competitors you are going to appear overpriced on the growing number of sites that offer comparisons.
Data Feed Problems Specific to Specific Affiliate Networks
Buy.at
Good Points: Probably the best feeds of the bunch. BAs feeds have an excellent field structure that actively encourages good categorisation as well as offering lots of scope for the inclusion of additional information.
Bad Points: There aren’t a lot of these feeds available. Also, despite their good field structure, BA has one or two really good examples of really badly categorised feeds. Successfully extracting more than handful of meaningful product details from their John Lewis feed, for example, remains one of the Grand Challenges to modern science.
Affiliate Window
Good Points: Huge number and variety of well maintained (up to date) feeds. Plus numerous other nice touches, such as the provision of a time stamp for all of their feeds, so that stale ones can be easily identified and avoided.
Bad Points: Over-simplistic feed structure that actively discourages adequate categorisation and the inclusion of essential information.
AW’s field structure doesn’t even have a natural placeholder for “model” information, for example. The end result is that
AW contains by far the largest proportion of “unusable” data feeds.
TradeDoubler
Good Points: Good scope for categorisation and a good provision of fields for the level of detail required (e.g. make, model, etc.)
Bad Points: Poor reliability. Many of
TD’s feeds update sporadically while many others don’t update at all. When they do update, many contain broken links, and many more are just empty files.
TD has no mechanism for monitoring or detecting broken feeds – they seem to rely entirely on affiliates reporting problems. Once reported, problems tend to either be solved very quickly, or else not at all.
Right. I’ve run out of steam. I hope that this overly looong posting is received in the manner in which it was intended: as a small step in helping to raise the data feed quality bar. From my experience with US data feeds; it seems to me that the UK is far ahead of the game. It would be a shame to squander this lead.
, we have been listening and improving things a great deal. For example we have a choice of different image sizes for every product over all Merchants. You will never get broken images on Paid On Results feeds, as the images come from the Network, and should a product not have an image you get a "no image" image in its place. As for the categories, we use the Merchants own categories as we feel they suit a product better than shoehorning them into categories we have created, however, we are looking at ways to create a network-wide category list as an option for affiliates to use instead of the merchants categories.