Show me Numbers

I had some discussion this morning on Twitter about what level of ownership the founders are likely to still have after the company’s latest round of VC funding. I thought I would do a bit of digging and see if I could estimate it.

Note: If anyone is aware of any funding rounds not included below, have specific information on any of the assumptions made or can spot flaws in my calculations please feel free to highlight them in the comments.

First round – July 2007

This is the trickiest element as I don’t think terms of this deal were ever disclosed. Techcrunch reported at the time an estimate of $1-$5m of funding raised. It was later reported that the deal size was net funding after costs of $4.8m. The unknown factor though is what level of equity Union Square Ventures (the first VC) received in return for this investment.

In the absence of any firm figure for this dilution we need to make an estimate. This was obviously a pretty early stage investment at a relatively significant ($5m) level so one could expect the dilution to be fairly significant. We also know from the Second Round (see below) that almost a year later Twitter was valued pre investment at $80m. So balancing these factors lets assume a pre investment valuation for the first round of $20m which would mean that ownership post First Round would have been:

Founders – 80 per cent
VC – 20 per cent

This estimate is highly material to the rest of the calculations as it sets the initial level of founder ownership that all other rounds will then dilute. In the conclusion below I indicate the impact of different assumptions for this round to the current level of ownership.

Second Round – May 2008

Investment size was reported this time at $15m with a pre investment valuation of $80m. Post investment this gives revised ownership of:

Founders – 67.4 per cent
First Round VC – 16.8 per cent
Second Round VC – 15.8 per cent

Third Round – February 2009

Investment reported at $35m with a valuation of $250m though it is not clear if this is pre or post investment. If we assume pre this gives the following ownership post investment:

Founders – 59.1 per cent
First Round VC – 14.8 per cent
Second Round VCs – 13.9 per cent
Third Round VCs – 12.3 per cent

Fourth Round – September 2009

Investment size reported at $100m with a valuation of $1bn. Again not stated whether pre or post so lets assume pre gives the following ownership post investment:

Founders – 53.7 per cent
First Round VC – 13.4 per cent
Second Round VCs – 12.6 per cent
Third Round VCs – 11.2 per cent
Fourth Round VCs – 9.1 per cent

Fifth and latest round – December 2010

Investment reported at $200m at a valuation of $3.7bn. Again not stated whether pre or post so lets assume pre gives the following current ownership estimate:

Founders – 51.0 per cent
First Round VC – 12.7 per cent
Second Round VCs – 11.9 per cent
Third Round VCs – 10.6 per cent
Fourth Round VCs – 8.6 per cent
Fifth Round VCs – 5.1 per cent

Conclusion

This analysis would estimate the Twitter Founders ownership at 51 per cent with a valuation approaching $2bn.

If you vary the First round dilution assumption you get the following alternative estimates for the current level:

NB It is also worth noting that if the valuations for rounds 3, 4 and 5 were all post investment valuations this would lead to an additional dilution in all the ownership percentage figures of approximately 3.25 per cent i.e. 51 per cent would fall to 49 per cent. The valuation figures would also all fall by approximately 8 per cent as the post investment current valuation would be $3.7bn not $3.9bn ($3.7bn valuation + $200m investment).

Andrew Lim – Editorial Director, Recombu and Founder of UKTJPR “PRfilter is a fantastic way to manage press releases and find interesting stories.”

James Holland – Editor, Electric Pig “Catering to the whimsy of fickle journalistic tastes is no easy task, but the intelligent tuning behind PR Filter shows great promise. A service that cuts the clutter, and brings me news I can actually use? Sign me up!”

Stuart Miles – Owner/Editor, Pocket Lint ““PRFilter looks to be the service that will help me get the news I want and filter out the press releases I don’t”

To date the use of technology to solve the issue of irrelevant or badly targeted PR content has been relatively limited. Database structures used for press release targeting are generally based around categorisation or perhaps keywords. Depending on the level of granularity this can often result in a poor match of a particular press release to individual journalists or bloggers.

Recently new language analysis based databases have started to be released that look at a journalist or blogger’s output in order to try and identify those who talk about a particular topic the most. This improves the intelligence of the approach for the sender if they use such tools effectively.

But even tools such as these do not address the issue from the individual journalist or blogger’s perspective. They don’t allow the recipients themselves to decide how relevant something must be to get their attention. Meanwhile spam filters or rules based inbox systems are often crude or time consuming to manage.

At RealWire we thought we would try and take a different approach. Having built a system to improve the targeting of our own distribution (which we will be applying in the coming weeks) we decided to go further. We asked ourselves – what if we could adapt the system to provide relevant releases to individual journalists and bloggers across thousands of releases a day from multiple sources?

So after months of development, in a bold experiment to both demonstrate our filtering technology and as a potential solution to the issue of irrelevant PR we have built PRFilter.

We believe PRFilter is something different:

Like the language analysis databases, PRFilter’s active interest technology™ builds a profile of a journalist or blogger’s interests from their own, or their publication’s, published articles. It then refines and updates this profile as new articles are published.

But then it flips things on their head and applies this profile to an inbound aggregated stream of press releases from multiple sources, presenting the individual journalist or blogger with the releases it thinks are most relevant to them – in a given time period, in selected geographies and even on a certain topic.

The user can then set their own personal relevance threshold and subscribe to alerts which pass this test (currently via RSS, other notification systems to follow). They can even train the system to improve its predictions by providing feedback on when it is right and wrong.

Making finding relevant stories a quicker and easier task and ensuring that senders of PR know that when their releases are indexed by PRFilter they will be seen by the most relevant media.

As the quotes above show we have already had some great feedback from initial beta testers, but like all beta applications we know it won’t be perfect and are keen to get feedback from all interested parties. Either contact me @AdParker, [email] adam@realwire.com, follow @PRFilter or register your interest in a beta account or updates here.

Following my post the other week regarding online readership, sparked by some aggressive sales tactics by one of our competitors, I got talking to Andrew Smith of Escherman and we agreed to jointly carry out a more extensive piece of analysis looking at 50 different online news sites.

We selected ten sites each from the following areas – UK Nationals, Business, Marketing, Technology and Consumer. There was no particular selection process, just an attempt to have a reasonably representative sample and we both hasten to state that this is a relatively limited exercise which should therefore be taken with at least a pinch of salt. Particularly since indexed urls have been used as a proxy for content as this can be impacted significantly by site structure (as stated in slide 21), with some sites having sub domains and/or a more complex content directory structure.

However at the same time with the data that is readily available we think it provides some (arguably) valuable food for thought. So after several hours of research, number crunching and graph generation here are the results (I suggest you view in full screen mode unless you have excellent eyesight):

Site data is sourced from Google for the number of indexed urls via the “site:domain” command and from AdPlanner for the traffic data.

At RealWire we have recently become aware that a major wire service is making a big deal out of their website’s high traffic numbers. In fact they have been specifically targeting the market trying to argue that their service is hugely better where visibility is concerned.

However they don’t mention the following three crucial issues about the traffic to their site.

1. That volume of traffic is clearly going to be affected by quantity of content.

2. The time visitors actually spend reading their content.

3. The relevance of those visitors to the content.

At RealWire we are always keen to make sure that discussions are based around the facts so let’s look at each of these in turn to see how our readership, engagement and relevance are all in fact apparently superior to Big Wire Corp’s.

Quantity of content

Q: Which is the more “popular” of the following two sites?

Site A – 1 piece of content and 1,000 page views in a month
Site B – 1 million pieces of content and 1 million page views in a month

Well according to their literature Big Wire Corp would apparently see Site B as a more popular destination because it has 1,000 times more page views. Makes sense, bigger is better right? Wrong.

Site A’s one piece of content has been viewed 1,000 times, whereas each of Site B’s stories has only be viewed once on average. Now which site is more popular? Site A of course.

Now let’s apply this concept to Big Wire Corp’s website.

First of all let’s get an idea of volume of content. The Google “Site:[url]” command gives you the number of unique pages indexed by Google on a particular site – a good proxy for the amount of content.

In this case the answer is 406,000.

Next we need an idea of traffic to the site. Google AdPlanner provides estimates of monthly page views.

In this case the answer is 7.5 million

If we then divide page views by content, we get an estimate of the number of views per article per month. Answer 18.5.

RealWire’s equivalent data from the same sources is
Content – 5,500
Page views – 200,000* (less than 3% of Big Wire Corp’s figure)

This gives 36.4 page views per article per month.

Twice the Big Wire Corp figure suggesting RealWire has higher readership for each article.

*I happen to know that the page view figure is too high in RealWire’s case (we do have analytics of course) but it could equally also be so for Big Wire Corp and so until I can get a hold of actual numbers for them I am being consistent.

Engagement

Q: Which of these two sites is engaging its readership the most?

Site A – average time spent on each page 2 minutes

Site B – average time spent on each page 5 seconds

Site A obviously. Each of the readers are spending 24 times longer reading an article on average than on Site B.

So let’s apply this to Big Wire Corp again.

Again Google AdPlanner can help. It tells us how many visits the site receives and how long each one lasts. From this we can get Time spent per Page as follows:

Big Wire Corp numbers

Time spent per visit = 3 minutes 50 seconds (230 seconds)
Total page views = 7.5 million
Total visits = 3.5 million

Time spent per page = 107.5 seconds or 1 minute 47.5 seconds

RealWire numbers

Time spent per visit = 8 minutes (480 seconds)
Total page views = 200,000
Total visits = 64,000

Time spent per page = 153.6 seconds or 2 minutes 33.6 seconds

43% more than Big Wire Corp suggesting readers of RealWire content are more engaged.

Relevance

Q: Which of these two sites is most likely to have the most relevant readership to a UK relevant story?

Site A – 100% of visits from the UK

Site B - 1% of visits from the UK

A: Site A – Yes I know these are getting ridiculously easy now!

Big Wire Corp’s market report focusses on US usage of their site when comparing themselves to others such as RealWire. However given that the vast majority of their content is from US companies it will come as no surprise that the vast majority of their traffic does as well. Google Ad Planner again helps us out.

US traffic – 76% of total

But the majority of RealWire’s clients and therefore content are from the UK. So what’s Big Wire Corp’s UK traffic like?

UK traffic – 3% of total.

And RealWire’s UK traffic? Well AdPlanner estimates around 75% but the real figure is nearer 45% or 15 times the Big Wire Corp figure.

Suggesting that RealWire’s traffic is around 15 times more likely to be relevant.

Conclusion

When evaluating traffic between sites it is imperative that you don’t get drawn in by the size of headline traffic numbers and that you consider:

a) normalising traffic for levels of content

b) how engaged the readers are

c) how relevant the readers are

Or you could find yourself reaching some very misleading conclusions. Just ask Big Wire Corp

* Hattip to Andrew B Smith for highlighting the value of Google Adplanner for such analysis

PR Week published its 2010 league table of the Top 150 PR agencies in the UK last week. The main headline was that overall the agency market was estimated to have grown by 0.75 per cent during 2009.

As comforting as this figure might be to all of us working in, or with, the PR industry my own take on the figures suggests that the picture may not have been quite so rosy.  I estimate that a reduction of around 5-10 per cent is probably a more realistic range and is more consistent with David Brain’s analysis yesterday from a global perspective.

This conclusion is based on an analysis of the changing make up of the table and other supporting evidence.

For clarity PR Week’s 2010 league table is based on income generated in calendar year 2009. Similarly the 2009 league table is based on income generated in calendar year 2008. 

Analysis of the league table positions

The table below shows the income that an agency had to achieve in each of the last two years in order to be ranked at the particular positions shown in the league table;

The table shows that to achieve a particular position in 2010 requires significantly less income at all levels compared to the 2009 league table. For instance to be ranked 25^th in the table this year required an income of £7.52m, but to achieve the same position last year required £8.35m, a 9.9 per cent reduction.

Excluding the change at the 150^th position the reductions are between 8.0 and 14.4 per cent with an average of 10.9 per cent.

In producing this table I have made adjustments for new entrants and mergers to make it more accurate:

2010 New entrants

Of the agencies that were new entrants into the Top 150 this year six of them would have been included in last years list had they submitted their figures. I have added these into 2009’s list for consistency.

Mergers

The numbers are also affected by the mergers of Grayling/Trimedia, Ketchum/Pleon and Tonic/Huntsworth Health. In these cases I have also combined the income for the three combinations in 2009’s list as well in order to compare like with like.

Other evidence that doesn’t fit with the headline estimate

Agencies joining and leaving the list – is it a fully representative sample?

It would appear that Stephen Waddington’s challenge to the industry to submit their numbers for the benefit of all has gone unanswered by some. 30 agencies (excluding those that have merged) that were in the top 150 last year do not appear in the 2010 league table.

Between them these agencies had total income in 2008 of £67.2m and varied in size between £13.80m and £1.13m. Given that the 150^th position agency in the 2010 league table has income of £0.44m these agencies would have needed to have suffered a reduction in income of between 61 per cent and 97 per cent to have not made the cut.

This seems highly unlikely and what I suspect is more likely is that these agencies just didn’t submit figures this year. Obviously individual agencies could have any number of reasons why they chose not to, or were unable to, submit figures. However the absence of 20 per cent of last year’s list begs the question as to whether relying on just those who have submitted represents a valid sample. Have some agencies chosen not to submit numbers for fear of how they might look? Is the table therefore more likely to be biased towards those that performed better?

It is impossible to calculate a sensible estimate of the impact of these absences, but it does seem reasonable to question whether the sample, that the 0.75 per cent growth figure was based on, is truly representative.

Performance of the major Marcoms groups’ PR brands

The major global Marcoms groups that publish figures for PR specifically performed as follows in 2009:

WPP is organic change from 2009 analyst report
Interpublic from PR Week article 17 March 2010
Omnicom is organic change from 2009 analyst report

It is obviously possible that the UK elements of these businesses performed better than the rest of the world. However given that the UK recession was amongst the longest and deepest and in the absence of any specific data to the contrary, I would take these figures as they stand.

Between them these three groups account for approximately £2bn in PR revenues worldwide and these figures would also indicate a reduction of approximately 5-10 per cent.

Conclusion

It was clearly going to be a challenge for PR Week to pull together a robust analysis this year due to the optional submission nature of the list. However I think there is significant evidence to suggest that the headline performance of slight growth is more than a little misleading.

But what do agency heads think? Is my estimate of minus 5-10 per cent nearer the mark?