Twitter isn't claiming of all its accounts less than 5% are bots. It is claiming less than 5% of their mDAU are. Without exposure to their methodology we are just speculating.

It is entirely feasible that at any point in time 20% of accounts are bots. It is whether they have counted them in their mDAU that matters.

Fully GDPR compliant, we're looking at behavior rather than personal data. We've been on the front foot there - also why we are advocates of content - win/win without the need for intrusive data usage. More on our site..

And that is the messy part, finding the good from the bad - and is worth solving that issue. That's the wider point, this type of, or framework of problem is going to be more and more prominent.

Thanks for the reply! Isn't enough data about behavior (scrolling speed, typing speed, mouse movements etc) enough to identify an individual, especially when it's combined with something like Browser version? I'd assume those don't change much across sites, i.e. Alice behaves similarly when reading reviews for a new phone or reading the local news.

Height, weight, hair color and year of birth aren't PII, but if you combine them, you're getting pretty close to identifying individuals, it's a behavioral finger print - and though I have no idea if it holds true in general, I'd assume that there's even a strong transfer across devices (a power user scrolls faster on both mobile and desktop).

Is there a way to really solve it? It seems to me that all you actually can do is up the ante, make it harder to do, but with full browser control, it'll be hard to lock them out. And if I'm looking at the behavior of users on my site, I can probably build pretty good replay users that visit the site you're protecting. I remember the "attack" on IRC channels back in the day where you'd join a bunch of bot users that would then have a conversation that is replayed from another channel (possibly in a different language), so timing, interaction etc looked very real (though they may seem a bit rude for not reacting to other people).

In some rare cases it might be, i.e. if you had an obscure browser version or browser. But scroll across sessions you wouldn't be able to so in general I'd say no. We've also architected the system such that you can't do that. Clients get end processed data showing how people have consumed content. Further, utilizing things like first party cookies tied to the domain.

We put it through the lens of, what helps clients understand the value of content and if people are actually enjoying it.

In terms of solutions, there is a few - profiling real users, using benchmarks of the outputs to real actions (i.e. people reading content tend to do this after) help a lot. As well as measuring in the first place. A surprising number of people still don't measure adequately.

A soft solution is simplifying the supply chain, keeping it transparent. Accountability drops the further you are from the end customer.

Something I maybe missed in this post, is the cost of this. This is real tangible cost, and we have seen cases over time in the millions of dollars. That's money robbed from creating a free and open internet. Advertising provides a significant subsidy for the eco-system.

It's a bit more than that if you give it a full read...

Any unintended consequences of remote working that you've found?

Good or bad.

It certainly is a leap of faith initially but you can make small leaps to get started.

Yeah you're right, Tremor video is trading about dollar for dollar, revenue to market cap at the moment.

We have launched a native ad diagnostic that lets you see how a piece of content (typically paid) has performed.

What we do that is different in this is estimate the earned media impressions based on our own network wide dataset. It's a great way to compare the relative performance of a couple of pieces of content. Ideally yours versus a competitors :)

Whack in a few pieces of content to see how it performed - we have a couple of examples there.

Over time we'll roll out more data as we can.

Same as supermarkets - once you own more of the wallet you'll get the margins. They're just shifting it from the supermarket to them.. quietly.

This is a great thread, there is an idea in matching these with marketers as vice versa they're usually looking for great ideas to sink their teeth into.

There's a great book called Little Bets by Peter Sims: http://www.amazon.com/gp/aw/d/1439170436?pc_redir=1397720852...

He demonstrates how big ideas come from making lots of little bets, which can be imprecise but begin to give shape to a solution.

Ie you start wide to get your operating field then hone in as you get more feedback or data.

So there's huge benefits in being imprecise... at the right time.

language stringing together a series of concepts. Concepts are formed by creating a conceptual idea. These ideas are created by choosing concepts which are similar (not the same in some way). The way they are not the same is by leaving out the measurements. Which measurements are left off depends upon the concept. Ex. the concept of wavelengths leaves out the measurements of the wavelengths individually.

Therefore the language is already very imprecise to begin with. Other new concepts are created and defined by previous concepts which will be used to define the new concept.

So language is conceptual by definition. The actual definitions of concepts are the referrents of the concepts. Ex. cat means all cats which have ever lived and all cats which will ever exist. This language then is a translation from existence to a mental re-presentation of existence then is a translation from existential to a mental model. This process is VERY conservative (a compression if you wish). One 3 letter concept can re-present a memory image we have all learned and can therefore convey (? Some meaning) to another person who has learned to duplicate the process. Notice however that we each live separate lives and will have unique representations in our unique memories of which cat we think the concept represents at this moment in time.

If you think of the memory as an operating system for each one of us, then we each have to write that operating system as we live out our lives. They (OS's/minds)then are all different to begin with in an extraordinary number of ways.

So to rephrase your question, there is an illusion (only) of understanding. The question is .... is this useful and if so in what way. In my opinion it is only by a careful and slow dialogue that any real meaning or understanding is EVER transmitted between any two individuals. Maybe this imprecise off the top of my head answer will be a good example of this concept.

This brief does not even consider the slippage between the of the messages between the reader and the writer and the reader.

The concept of context alone took me 5 years to sort out before I came up with how it is established, so that I could begin to understand the problem of how and why languages work at all.

edit

reader and the writer and the reader. .... should read:

the writer and the reader.