LookStat – Microstock Analytics

Matt’s series on the Long Tail and microstock got me thinking and I decided to do some quick searches on human emotions on Dreamstime to see if anything interesting showed up in the data.

I did searches for 5 keywords (angry, bored, happy, sad, scared), ordered the results by download and captured views and downloads for the top 10 images. I also recorded the number of results returned. While this is completely non-rigorous analysis, there are some interesting observations.

If anyone has thoughts on how to better interpret this data, please don’t be shy. Also, if anyone would like a spreadsheet of the data, let me know and I’ll send it to you.

The things that jumped out at me from this exercise (in no particular order)

• There are way more images returned from a search for ‘happy’ than for all of the other terms I looked at. While this does make sense from a marketing perspective, the skew seems out of whack when compared to the relative download counts.

• Conversion rates for ‘bored’ and ‘scared’ are very high and image counts are very low. To only have 3,000 images for a generic emotion like ‘bored’ in a 4 million plus image universe is surprising to me.

• Don’t ignore emotional variations when conducting shoots. Collectively, they account for a significant number of downloads and there are fewer results in the pool, so in theory standing out should be easier. Long live the Long Tail!

• Search terms that have high conversion & relatively low result counts may indicate areas of opportunity but (much) more testing is needed before you can draw any conclusions.

• PIcNiche is a really cool site and they basically automate what I’ve done here using Fotolia data. I think it’s well worth a visit as you plan your next shoot. (Thanks for the pointer to it Lee.)

Image Count (number of results returned when searching for a term)

The number of images that are returned when you search for ‘happy’ versus the other keywords is staggering. There are over 15 times as many images returned for ‘happy’ versus ‘sad’. While this lines up with intuition (happy people are better for marketing) the spread took me by surprise.

Average Downloads & Views for Top 10 Images

To get this data, I just sorted the results for each search by download count, recorded the download and view numbers for the top 10 results and then calculated the average. There are also some interesting things to note in the individual image data which I discuss in the next section.

One thing to note here is that even though there are over 15 times as many images results for ‘happy’ as there are for ‘sad’, the difference between average DLs for the top 10 images is only 3x. While ‘happy’ clearly drives more downloads than the other emotions, if you incorporate them into your shoots, there’s an opportunity to capture incremental downloads.

The results for ‘sad’ were skewed by two images that had an extremely high number of views. When I removed those two data points, the average number of views for sad was: 1,900 which is more in line with the other results. Those two images were in numerous collections and my guess is they were featured on the home page or as editors’ picks at some point.

Average Conversion Rate

For this post, I defined conversion rate at the ratio of Downloads to Views. To some degree, this metric is a measure of quality/keyword relevance. If something is viewed a lot, but not downloaded that much, it either means that it is a low quality image (people viewed it but decided not to buy it) or it could mean that the image is not relevant to a high volume search keyword.

Again, the results for ‘sad’ are skewed by the two high view images. If I remove them, the average conversion for ‘sad’ rises to 9.6%. It’s interesting to look at the PicNiche data for the same search terms. PicNiche rocks in my opinion and I’m looking forward to seeing how it evolves. I think they are using Fotolia data and have access to the total downloads and views for a given term as opposed to the sample approach that I took.

If you superimpose Average Downloads per Image (Left axis) and Average Conversion Rate for the search term (Right axis) you can see that even though all terms apart ‘happy’ have similar download numbers, ‘bored’ and ‘scared’ have significantly higher conversion rates than the others. There are also far fewer of these images than in the other two cases. Clearly more analysis is required, but the hypothesis that high conversion & low image count = opportunity is probably worth exploring further.

Per Image Detail for the Top 10 Images (For the Data Geeks out there…)

Perhaps the most interesting observation here is that there isn’t a ton of variation between the top ranked image and the 10th ranked image in terms of download counts. Another thing that jumped out was how close to each other the download counts for the searches other than ‘happy’ were.

We’ve gotten a small flood of new signups recently and the system is behind in terms of account initialization. We’re working as hard as we can to rectify this situation and will continue to keep you updated via the blog and email. Thanks to everyone for giving us a chance and for your patience.

Matt Antonino has a great post on NiltoMil about the Long Tail and it’s applicability to microstock photography. I think the core of Matt’s thesis is spot on and the long tail framework is a useful way to think about your microstock portfolio.

A concrete example of this thinking as it applies to microstock is when shooting people. While the vast majority of microstock images sold are of shiny, happy, smiling people, there is still demand for other emotions and moods. If you’re on a shoot and can work in the emotional variations easily, it’s probably worth doing.

Just wanted to wish everyone a Happy Thanksgiving. I hope everyone celebrating it has a wonderful time with family, friends & good food. We’re working as fast as possible get a rush of new accounts activated but will likely need until early next week to catch up. Thank you, as always, to all our users for their support.

Lee Torrens has a post on his blog called Microstock Mash which looks at some of the emerging services in the microstock sector and we’re delighted that LookStat made the list. If you’re interested in microstock and you haven’t come across Lee’s blog, Microstock Diaries, yet – go there right now and subscribe. Lee’s a great guy and I learn something every time I visit his blog.

I also wanted to say a public thank you on the blog to Casey Zednick. He’s the CTO and also a founder of LookStat and the service wouldn’t exist without him.

Since we launched the LookStat beta, we’ve gotten some great early feedback from the microstock community. A couple of our early users (thanks Lee & Andres) have given me permission to publish their notes here.

Lee Torrens via Twitter

Loving LookStat so far http://twurl.nl/xhcito This this has AMAZING potential. Develop, LookStat team! Develop!!!

Andres Rodriguez in the Microstock Group Forum

Rahul it was great to meet you in person in NY at photoplus expo. I have had a look at the site and opened an account and it is unbelievable. The potential you have there for development is amazing, honestly. I can not wait until you have the full version running, any ideas on when will this happen?

We have a lot of work to do before we can realize the full potential of LookStat but early support like this is incredibly motivating.

Lee, Andres and all our other beta users have been very patient with the service’s rough edges and have given us a lot of helpful ideas and suggestions (and have gently pointed out areas where we are ahem doing a less than stellar job.) Thanks a ton to all of you. We’re working hard to keep things moving forward and every piece of input helps.

The system is catching up quickly and existing users should see their accounts being updated regularly. Users who signed up in the last couple of days will be seeing their accounts come online in short order. We’re grateful for the support and please don’t hold back with your feedback and ideas.

We’ve had a good day today in terms of interest in trying out LookStat. We’re working hard to get all the new accounts initialized and as a result, site performance has suffered somewhat. We’re working on solving in the issue in the immediate term and are taking steps to make sure we don’t have this sort of problem going forward. Still, these are the kind of issues we love to have.

I’ve put up a page on the blog which lists some of the bigger issues we are focused on at any point in time. Over time we’ll come up with a better way to communicate with our users about this but we wanted to put something up right away.

We’ve gotten a ton of helpful feedback already – please keep it coming.

One of the questions my co-founder Casey likes to ask is, “Are we winning?”

This is a big question, but in the context of microstock earnings, we’re trying to see what our earnings trend is and how our month is likely to shape up. To do this in LookStat, the easiest way is to view the last 30 days and compare that number to your last month’s earnings. This way you can get a quick read on how the month is shaping up. If the total for the trailing 30 days is greater than your earnings in the prior month, then you’re on track for an upswing, and vice versa.

Microstock Earnings - 30 day snapshot

This is imprecise due to seasonality and other factors, but it’s a great gut check.

Over time, we will be adding in more metrics to help answer the ‘are you winning’ question. Having data is great, but giving that data meaning is where things really start to get interesting.

I’ll be attending UGCX (User Generated Content) Conference in San Jose (Feb 9 & 10) and will be on a panel called “Servicing the Crowd: Business Opportunities around UGC.” I’m really excited to be there and I’d love to meet any and all microstockers who will be attending.