Author Topic: Keywording Gold from StockPerformer (Read 4033 times)

leaf · « **on:** November 16, 2013, 08:42 »

StockPerformer just released some data at their mexpo presentation that is also on their blog. There is some pretty super data put together about which keywords are worth using and which you may as well leave out.

Female and Caucasian (in a pic of a woman) is essentially useless for example
https://www.stockperformer.com/blog/keywording-for-success/

Sean Locke Photography · « **Reply #1 on:** November 16, 2013, 09:38 »

One question is, which 250,000/100,000 images were analyzed?

The collection is thick with students, laptops, education, food, family. In a random selection of 250k images, I can't imagine these would be low in supply, although I am sure they do sell well in the aggregate as popular topics. Are people not using the proper keywords? Is that the problem?

Likewise, dog, baby, eating, do not seem to be particularly niche themes.

I can understand male and female not being highly used, as they are really modifiers for "adult" or "child", which you can achieve in one word with "man" or "girl".

It's also hard to really know how the system works. I would think a buyer might search on "jogging, health" initially, but would try to narrow things down later with "woman" and "shoes". Does the latter affect the order?

Interesting read though.

luissantos84 · « **Reply #2 on:** November 16, 2013, 09:47 »

very interesting indeed!

SS does provide info regarding what a buyer have searched for and to every file, they could collect that if contributors are willing to share it

face 35%
men 28%
portrait 17%
isolated 4%
happy 4%
smile 4%
white 2%
male 2%
background 1%
casual <1%
young <1%
cool <1%
guy <1%
people <1%

people do come at the bottom but portrait isn't doing poorly

Ron · « **Reply #3 on:** November 16, 2013, 10:16 »

I didnt know this, Chris McBurney, a.k.a “Lobo” iStock admin. Thats very interesting to know.

gostwyck · « **Reply #4 on:** November 16, 2013, 12:34 »

It's an interesting read but I believe their conclusions are flawed for several reasons.

Firstly, they are using Istock's own keyword weighting system as their sole source of data. Whaaaat? Does anyone take Istock's supposed 'data' in any way seriously? They can't even tell us what money we made 3 weeks ago! Is it possible that Istock really have a 'keyword weighting system' that actually works properly?

Istock's system, whatever it is worth, is based on their own precious CV system which limits the options available to both the keywording contributors and the buyers. For example SP have concluded that 'Isolated on White' is one of the least used keywords by buyers. From my own data at SS I know that buyers use the word 'isolated' quite often (up to 23% on my popular isolated subjects). Maybe Istock's 'Isolated on White' is just too long to be practical.

SP state that 'Food' is a very popular keyword used by buyers. Not on SS it isn't. About 95% of my sales are of food subjects, all have 'food' as a keyword, and yet according to SS the buyers almost never use that word when searching other than for specific ethnic subjects (such as 'Indian food').

I'm not sure what we can learn from all this anyway. For example, just because SP tell us that buyers never use or 'convert' with the term 'female' is anyone going to stop using it? I don't buy images much myself but when I am doing searches then 'female' would be my normal word to use for an adult rather than woman, lady, etc.

pancaketom · « **Reply #5 on:** November 16, 2013, 14:09 »

It would be very interesting to see something similar comparing different agencies. I think IS is messed up because of how their CV works compared to most computer searching. Also if it is anything like the keywords that DT lists for the sales about 20% are completely off.

I also find it hard to believe that something like "beach" is really a niche keyword.

Unfortunately this is sort of mixing up keywords and photo selection - Yes, it is important to use the correct keywords, but it should be much more important to use keywords that are appropriate for the image rather than keywords that buyers are searching for.

I also don't trust the source of data. We all know how likely it is that something that has been stated in the IS forums is incorrect - "we have no plans..." and various other statements. Also when you modify a search, how does that factor into the data?

Still, it is interesting, and good to know if buyers search on "female" or "woman".

edit: In looking at their methodology I think it might not be very valid - at least not for telling us what I think they are saying it is telling us. Take for example "christmas" not a subject I would think is a niche subject in microstock. What I think they are saying is that when people bought an image with "christmas" in the keywords, there is a good chance that they searched on "christmas". It certainly doesn't seem like there is any shortage of christmas images. Also on SS it is the #2 search term of all time - hardly a niche. (but for sure still worth putting in your keywords if it is appropriate)

Still, there is some interesting information - like putting "face" in the keywords might make more sense than "portrait". It would be interesting to know if putting portrait in there actually hurt your placement.

heywoody · « **Reply #6 on:** November 17, 2013, 06:10 »

Something like "female", in itself, probably wouldn't be used in a generic search for woman but, in combination with something else, it might be used more often than "woman", e.g. "female mechanic" as opposed to "woman mechanic".

ShadySue · « **Reply #7 on:** November 17, 2013, 07:33 »

I didn't think that report was helpful, because the ONLY thing that actually matters, we don't know; which is whether using keywords which are relevant but not searched on are punished. It may well be that they are, but I have no idea. It's been suggested by contributors here and over there but never confirmed or refuted. Based on the bizarre positioning of one particular file of mine (extremely low on it's most relevant keyword, but much higher on its second most relevant) it may be, but that's the only file I try to work out the 'punishment' thing on. The penalty of getting 1-9 (10?) downloads compared to 0dl seems much worse.
If unused relevant keywords aren't actually penalised, why wouldn't you incude them, for the 1% of buyers who might use the term?

djpadavona · « **Reply #8 on:** November 17, 2013, 07:40 »

I agree with Gostwyck. Because the keywords are pigeon-holed through the CV system, the results of the study are useless. You are better off using Google's keyword tool to delineate estimated search numbers, and just use common sense when considering how a searcher is likely to refine search words after the results of the initial search are returned.

grapegeek · « **Reply #9 on:** December 10, 2013, 14:51 »

Sorry a little late to this thread. We all know that IS has some type of ranking system behind the scenes. I proved this years ago about keywording a bunch of files exactly the same as to the top best match at the time. So, it's useless to try and keyword exactly the same as top results. We know the other factors that come into play with IS are exclusivity, age of file, amount of downloads, description and several other things, but there is surely some ranking system that IS uses internally to make some files float to the top in best match searches.

I also think it's safe to say that the top search terms in SS are going to be pretty close to the top search terms with IS. Flower, Christmas, Background, Vector, Medical, Family, Tattoo(?), logo, music, woman

DF_Studios · « **Reply #10 on:** December 11, 2013, 22:36 »

Useless data analysis from number crunchers who peed their pants with their amazing discoveries.

I used to work in market research and sat through endless presentations on the computer industry that had similar "earth shattering" trend analysis. Companies paid high subscription fees for access to this "data" and in the end it was meaningless.