DeepMeta - Daily Stats

DeepMeta Blog

- Last update:

Feature

DeepMeta - Daily Stats

We're pleased to announce our latest DeepMeta 4 feature: Daily Stats. A timeline of your Asset downloads by customers over the last days and weeks

Franky De Meyer

By Franky De Meyer

Update, October 2026: Predictinator v2 released

The "$" estimate on each download is now made by Predictinator v2, built and tested on millions of downloads matched with the royalties that were actually paid for them, and adjusted to your own sales. It scores 5% better than v1, and has 18% fewer estimates that are two or more "$" off. How it works, what data it uses and how accurate it is per sales category: see the Predictinator v2 section below.

The Daily Stats section shows a timeline of your asset downloads by customers over the last days and weeks. Getty publishes new Daily Stats once a day, at midnight, Seattle time. DeepMeta picks them up automatically, so there's nothing to sync: if you've looked at your Daily Stats in the last 30 days, they're checked automatically shortly after midnight. Otherwise the update starts as soon as you open the page, and a spinning icon at the top shows while it runs.

We show up to 3 months of history. The first time you open the page, you'll see the last few days; older days are added in the background over the next hours. Daily Stats are only available for Creative uploads.

A few screenshots:

Day View

Daily Stats - Day View

Downloads that came in with the latest daily update (since last midnight, Seattle time) are marked with a yellow "NEW" label.

Select the types of assets you wish to see in the list: Images, Video and/or Illustrations. Within each day, week or month, sort by DL (Download Count), DL date (latest download) or Upload date, ascending or descending.

Week View

Daily Stats - Week View

Month View

Daily Stats - Month View

Chart

Above the list, a bar chart shows your downloads per day, week or month. Use the Asset type / Sales type switch to split the bars by Images, Video and Illustrations, or by download category. The chart can be collapsed if you'd rather see more of the list.

First Time Download

Thumbnails marked with a sparkles icon have been downloaded for the very first time (counting from your oldest uploaded Royalty Statements, so you'll only see it once you've imported them). A rocket icon means it's the first download of an asset you uploaded less than a week ago.

Daily Stats - First Time Download

Predictinator v2

Since October 2026, the 1 to 5 "$" icons on each thumbnail come from Predictinator v2. Getty doesn't tell you in the Daily Stats what a download earns, but a month or so later the Royalty Statements do. That gives us something v1 never had at this scale: millions of downloads of which we know the actual outcome. v2 is built on them, tested on them, and we're happy to explain exactly how it works.

What data it uses

  • The Daily Stats of DeepMeta users from October 2025 to August 2026 (Getty keeps them for 12 months), matched with the sales in the Royalty Statements they uploaded. A download is matched with the sale of the same asset on the same day. Almost all of them match on the exact day; iStock credits are sometimes dated a day earlier in the statements.
  • Only the Regular statements count. Downloads with refunds, and statements that need a re-upload, are left out.
  • The model itself only stores typical values per category, collection and contributor type, not who sold what. Your own statements are used for one thing only: adjusting your own estimates.

How it works

  1. A typical download. For each of the seven categories, it starts from what one download of the same kind typically earns: the same asset type (photo, illustration or video), the same usage (creative or editorial), the same contributor type (iStock exclusive, iStock non-exclusive or Getty) and the same collection. When a group is too small, it uses the next broader one. It takes the median, the middle value, rather than the average: royalties are so skewed that a handful of big sales would push the average one "$" too high for most downloads.
  2. Your own sales. It then compares what you earned in the last 12 statement months with what was typical for the same kinds of sales, per category. If your royalty rate or your usual prices are higher, your estimates go up; if they're lower, they go down. The more sales you have, the more weight your own history gets. Without uploaded Royalty Statements, only the typical values are used.
  3. Several downloads. The "$" stays the value of one typical download: for an asset downloaded in several categories, it's the average over its downloads. Three downloads of the same kind show the same "$" as one, and the Day, Week and Month View agree. The DL count next to it tells you how many there were.

What the "$" levels mean

The levels are the same as in v1. The last column shows how often each level actually occurs (downloads of June to August 2026):

RatingRoyalty per downloadHow often
$less than $0.5027%
$$$0.50 to $122%
$$$$1 to $540%
$$$$$5 to $2510%
$$$$$$25 or more2%

The seven categories and the sales behind them

The Daily Stats sort every download into one of seven categories. Here's which sales they turn out to be in your Royalty Statements, what they typically earn per download (the median, with the middle half of all sales between brackets), and how often v1 and v2 got the "$" exactly right or within one "$":

CategoryIn your Royalty StatementsShare of downloadsTypical royaltyExact: v1 to v2Within one "$": v1 to v2
GI Premium AccessGetty, Premium Access Time Limited51%$0.85 ($0.43 to $1.85)43% to 42%87% to 89%
iS SubscriptionsiStock, iStock Subscription38%$1.25 ($0.30 to $1.66)67% to 65%89% to 92%
GI Single AssetPartner Portal or Getty, RF Image and RF Clip6%see below72% to 76%82% to 83%
iS CreditsiStock, Credit Pack5%$7.71 ($1.96 to $9.63)92% to 91%100% to 100%
GI UltraPackGetty, Premium Pack Stills0.2%$53 ($36 to $67)74% to 87%85% to 99%
OtherCustom Content0.2%$93 ($76 to $115)92% to 100%92% to 100%
GI SubscriptionsGetty, Otherless than 0.1%too few sales to tell––

GI Single Asset is really two categories in one. Most of its downloads are Partner Portal sales that earn a few cents (a median of $0.03 for stills, $0.50 for video), while Getty's own RF sales earn tens of dollars ($46 for stills, $60 for video). v1 used one value for both, which is why it was so often far off. v2 learns from your own sales which kind you usually get, and its score for this category went from 28 to 48.

What's still hard

  • GI Premium Access is half of all downloads, and the hardest to call. What it earns depends on the customer's Premium Access agreement, which the Daily Stats don't show: 28% of these downloads earn less than $0.50, 29% between $0.50 and $1 and 36% between $1 and $5. Neither version gets more than about 4 in 10 exactly right. The good news is that v2 is more often within one "$" (89%).
  • iS Subscriptions come in two groups that depend on the customer's subscription plan: about a third earn less than $0.50, about half between $1 and $5. About 2 in 3 are exactly right.
  • iStock Unreleased is the one collection where v2 does worse: it rates nearly all of these downloads $, while a quarter of them earn $1 to $5. Within one "$" drops from 85% to 73% for this collection.
  • GI Subscriptions are still too rare to measure: 66 downloads in 11 months.
  • Your adjustment only knows what you've uploaded: keep your Royalty Statements up to date in DeepMeta, and the estimates follow your latest sales.

It's still an estimate: a single sale can differ a lot from what's typical for its kind.

Predictinator v1 (how the first version worked)

Due to the many different ways customers can license your assets, it's often difficult to predict how much you'll earn from a download. This is why Getty only provides the download count for Daily Stats, not the actual royalty. The royalty is published at a later stage, in the Royalty Statements.

Fortunately, the Daily Stats do specify one of seven categories for a download. After some research, we've been able to come up with a very rough estimate of the royalty you might earn from a download.

Meet "Predictinator v1":

Predictinator v1

As you may be able to tell from the looks of the machine ... accuracy is not its strongest point. But it's a start, and definitely better than nothing at all.

Each thumbnail gets a "score" of 1 to 5 "$" icons. Once again: this is a very experimental feature; "$" icons give an indication whether it looks like a low or high earning sale, or somewhere in-between. This is based on multiple matrices calculated from historical sales data, matching the seven categories (GI Premium Access, GI UltraPack, GI Single Asset, GI Subscriptions, iStock Subscriptions, iStock Credits and Other), the asset type, your exclusivity status for this asset type, the collection of the asset and the usage type.

Please use with large doses of reservation and scepticism, since the actual royalty depends on some other factors that cannot be accounted for at this stage.

We'll keep evaluating the accuracy (or non-accuracy) of this label and are already working on v2, with v3 making its way to the whiteboard.

Daily Stats - Rough Value Estimate

Detail View

Clicking any of the thumbnails in Day, Week or Month View brings up the Detail Panel, showing the DL counts for each of the seven categories: GI Premium Access, GI UltraPack, GI Single Asset, GI Subscriptions, iS Subscriptions, iS Credits, Other.

Daily Stats - Detail View

For most assets, the five top keywords are shown as well. These are the keywords that customers have used the most since your asset was published, for views and interactions. This is valuable information you can use for your future shoots and keywording.

Try it out

DeepMeta - Daily Stats

Info on the Getty Images website