The COI (Central Office of Information) released the annual report for government websites at the end of June 2010, giving us an insight into expenditure for all sections of government informational websites, and analytical data such as visitor numbers.

This data has allowed us to carry out our own CatN analysis and apply our billing metrics and pricing packages to central government sites data. It is worth noting that these sites are purely static content with some search functions and other small scripts.

This COI report concerns itself with purely informational web sites, not large applications like the HMRC (Her Majesty’s Revenue and Customs) online self assessment form, and any hosting claims we make are not based on hosting these applications. Similarly, considering that CatN vCluster is purely a php hosting platform, this analysis is based on the value that the government should migrate to using open-source software and php script for their informational and static web sites, and will be addressed further in this report.

Using the data provided by the COI we want to discover if the government spend on hosting, up to £4,661,000.00 for one site, is giving the government value for money. We are judging value for money as low cost per web site hit, high availability and performance levels in relation to a sites traffic and hosting expenditure. Please note: performance value for money will be addressed in a follow up article.


Selected sites data

We have taken a sample of 10 sites data that reflect trends across the 45 sites in the COI report, and fairly represent government web site sizes, judged by traffic and hosting/build expenditure. However, in order to sight trends across all government departments, in some cases all the provided data has been used. This is then grouped into department.


Site specific cost per request

From our sampled websites, we have calculated cost per request using annual requests and annual hosting expenditure to see if their current hosting provides value for money. The hosting expenditure to hit ratio reveals whether government funding is being used effectively, and can highlight websites that are being over/under funded based upon the traffic demand of the site.

Due to the static nature of these websites the cost of a hit should be low when compared with complicated applications like the HMRC online self assessment form.

Table of sample sites cost per 1000 page views

Government Site Annual hosting and infrastructure spend Annual page requests Cost per 1000 page views
businesslink.gov.uk £4,661,000 93,502,545 £49
communities.gov.uk £372,000 29,767,248 £12
eatwell.gov.uk £85,000 13,813,065 £6
hmrc.gov.uk £1,071,000 447,994,614 £2
mod.uk £127,000 29,111,900 £4
nhs.uk £2,610,000 395,352,875 £6
dwp.gov.uk £197,000 72,820,907 £2
number10.gov.uk £86,000 34,108,349 £2
dft.gov.uk £810,000 17,497,004 £46
fco.gov.uk £645,000 53,650,286 £12

Bar graph comparing cost per request of sample government web sites


Clearly the least efficient sites from our sample in terms of cost per hit are dft.gov.uk and businesslink.gov.uk. These two static sites are paying disproportionally for their hosting against traffic ratio especially when comparing to NHSUK which has large traffic and hosting expenditure, but each page view only costs £0.006, up to 86% less than DFT and businesslink.

This suggests inefficiency in DFT and businesslink’s hosting infrastructure and cannot be explained by claiming these sites as complex, as all samples in this report are considered static, only serving text and images and simple search scripts.The best site for value for money from our sample is number10.gov.uk, with one of the lowest hosting expenditures at £86,000 per annum and low cost at £0.002 per page view, making the site very cost effective. Other cost effective sites are dwp.gov.uk, mod.uk, and hmrc.gov.uk.

Something to note is that businesslink.gov.uk is in the same government department, HMRC, as hmrc.gov.uk yet is the least cost effective site in our sample, suggesting inconsistency across government websites, and an individual sites development impact on value for money.


Department spend and cost per request

As DFT and businesslink in the department HMRC are spending up to 60% more than similar government sites for each request, analysis across departments is required for an explanation or trend in spending.

Table of sample sites grouped by department and cost per 1000 page views

Government Department Annual hosting and infrastructure spend Annual page requests Cost per 1000 page views
HMRC £5,732,000 541,497,159 £10
CLG £389,000 29,767,248 £12
FSA £356,000 13,813,065 £14
MOD £127,000 29,111,900 £4
NHS £2,610,000 395,352,875 £6
DWP £1,957,000 72,820,907 £3
CO £194,000 34,108,349 £2
DFT £4,472,000 17,497,004 £29
FCO £645,000 53,650,286 £12

Bar graph comparing annual hosting expenditure of sample government web sites


We found the trend shown by the single sites, dft.gov.uk and businesslink.gov.uk, is continued across the department with DFT and HMRC easily spending the most on hosting compared to every other department. In fact HMRC’s average cost per page view across the departments is kept lower due to the value for money given by hmrc.gov.uk. Again this gives weight to the theory that the web site businesslink.gov.uk can be optimised through the development process more than the hosting infrastructure.

It seems that both these departments are over spending for hosting static sites and could make large savings by moving to php that can be hosted on an open-source low cost stack in the cloud, especially as each are hosting more than one site. The cloud infrastructure would allow departmental sites to scale linearly instead of adding dedicated nodes and seeing expenditure jump incrementally.


Cost per visitor

Using the gathered data for cost per page view we can estimate the number of page views per visitor and calculate a total cost of one visitor to each Government site.

Table of sample sites grouped by department and cost per 1000 site visitors

Government Department Average page views per visitor Cost per 1000 site visitors
HMRC 6.67 £66.74
CLG 6.02 £72.26
FSA 1.57 £22.03
MOD 3.72 £14.88
NHS 4.01 £24.06
DWP 0.51 £1.52
CO 1.61 £3.22
DFT 0.70 £20.27
FCO 4.60 £55.18

Naturally we could expect the cost of 1000 page views to reflect the cost of 1000 site visitors. However, the FCO paying a lot less than DFT for 1000 page views, pays much more for 1000 site visitors. This can be explained by the number of pages being viewed by visitors to the FCO site when compared to DFT.

It seems as though DWP is keeping its costs low by paying one of the smallest amount for 1000 page views as well as 1000 site visitors. This suggests that DWP have optimised their hosting infrastructure relatively well when concerned with expenditure.


Site availability

We have ranked site availability to see if higher hosting and infrastructure spend improves uptime for government sites giving value for money with service availability, and if hosting expenditure can be qualified by availability, explaining the proportionally high hosting expenditure for some sites.

Sites are grouped by department, as they host internally and will be using the same infrastructure for all department websites. Equally the total department expenditure for hosting covers all the sites in a department.

Table of department hosting spend and service availability

Government Department Annual hosting spend Average service availability
AG £5,000 100%
CLG £389,000 100%
DEFRA £0 100%
FCO £645,000 100%
GCHQ £0 100%
HMT £56,000 100%
UKTI £3,322,000 100%
COI £4,000 99.990%
NHS £2,610,000 99.980%
TNA £117,000 99.980%
CO £194,000 99.959%
FSA £356,000 99.925%
DE £670,000 99.920%
DH £597,000 99.910%
Government Department Annual hosting spend Average service availability
HO £1,212,000 99.900%
MOD £127,000 99.890%
DCMS £60,000 99.800%
DWP £1,957,000 99.785%
DFT £4,472,000 99.750%
HMRC £5,732,000 99.750%
OFGEM £40,000 99.700%
OFT £195,000 99.620%
MOJ £0 99.400%
PSC £9,000 99.315%
BIS £258,110 99.260%
DFID £161,000 99.185%
AC £0 98.800%
OFSTED £378,000 98.000%

Bar graph comparing service availability of sample government webs ites

(1) Missing service availability data


Hosting spend seems to have no impact on the service availability of a departments websites. The high cost per request for dft.gov.uk and businesslink.gov.uk is not explained by 100% uptime. Interestingly OFSTED with the worst service availability time of all government departments spends close to the average department spend of £673,320.91 on its hosting, suggesting that their infrastructure spend is not being invested in a solution that guarantees uptime. From this sample data there is no value in service availability gained from investing heavily in hosting, a bigger expenditure does not deliver greater uptime.

The lack of correlation between government spend and availability suggests room for investigating more stable hosting platforms that can provide service uptime in exchange for investment in the infrastructure, and could be explained by high site performance which will be investigated in the following article


Development

Value for money also comes through the development of a site. A site with high development expenditure in relation to hosting expenditure can be expected to be giving value for money over the course of the year following the development. The quality of a sites build can also affect up time as coding errors will cause a site to go down as readily as infrastructure failures.

Table of sample site build to hosting spend and ratio

Government Site Annual hosting and infrastructure spend Annual development/build spend Ratio (Hosting:Build)
businesslink.gov.uk £4,661,000 £4,388,000 1:1.06
communities.gov.uk £372,000 £70,000 1:5.31
eatwell.gov.uk £85,000 £160,000 1:0.53
hmrc.gov.uk £1,071,000 £132,000 1:8.11
mod.uk £127,000 £418,000 1:0.3
nhs.uk £2,610,000 £6,377,000 1:0.41
dwp.gov.uk £197,000 £7,000 1:28.14
number10.gov.uk £86,000 £11,000 1:7.82
dft.gov.uk £810,000 £198,000 1:4.09
fco.gov.uk £645,000 £756,000 1:0.85

Bar graph comparing annual design and build expenditure with hosting for sample government web sites

(2) Site redirect discrepancy


Clearly there is no consistent correlation between hosting spend and build spend. In the case of the NHS (National Health Service) and DWP (Department of Work and Pensions) design and build costs are more than three times greater than hosting costs, suggesting either a redevelopment of the sites in the last year, or continuous development. (3)


We continued to test development expenditure against uptime as there is no correlation between hosting expenditure and uptime, perhaps development will reveal a trend.


Table of sample site build spend and service availability

Government Site Annual development/build spend Service availability
businesslink.gov.uk £4,388,000 99.500%
communities.gov.uk £70,000 100%
eatwell.gov.uk £160,000 99.950%
hmrc.gov.uk £132,000 100%
mod.uk £418,000 99.890%
nhs.uk £6,377,000 99.980%
dwp.gov.uk £7,000 99.600%
number10.gov.uk £11,000 99.900%
dft.gov.uk £198,000 99.900%
fco.gov.uk £756,000 100%

Again, there is no correlation between development and uptime, suggesting that the government gains no value in terms of availability from development expenditure. The sites communities.gov.uk and hmrc.gov.uk with the lowest build expenditure both achieved 100% uptime, where as the sites busineslink.gov.uk and nhs.uk with the highest build expenditure both failed to achieve 100% service availability with businesslink on 99.500% and nhs.uk on 99.980%.


Solutions

Analysing government value for money and hosting costs reveals that large savings can be made. Using government figures with CatN pricing for static sites we have produced some conservative estimates for hosting on an open source php platform that are similar to other php hosting companies as well.

By the CatN estimates a LAMP (Linux, Apache, MySQL, php) stack would certainly reduce the expenditure for hosting these static government sites, and savings could also be made in development as these products are open source and require no license purchases. Lowering hosting spend will immediately see greater value for money from these government sites, just because the expenditure will be less.

Moving these sites to a concurrent cloud hosting environment and optimising the websites code would reduce the hosting expenditure and the sites storage requirements, reducing total hosting expenditure, and improved performance for the site users. A cloud infrastructure would reduce costs by allocating resources as the web sites require them, reducing the amount spent on unused capacity. This should be a priority during this period of economic instability, and savings should be made wherever possible.

The prices given for the government websites to host on the CatN platform have been made using the Platform Hour calculator available at catn.com/vcluster/pricing/platformhours. Estimating page hits based on the published figures for page impressions allows hits to be entered into the Platform Hour calculator, producing an estimate for Platform Hour concurrent usage of the site across our stack resulting in an estimate bill being produced.


Table of CatN estimates for government sites

Government Site Current annual hosting spend Annual hits estimate CatN estimate Cost reduction
businesslink.gov.uk £4,661,000 12,402,974,760 £115,251 97.53%
communities.gov.uk £372,000 4,264,585,044 £40,803 89.03%
eatwell.gov.uk £85,000 2,821,976,640 £26,273 69.09%
hmrc.gov.uk £1,071,000 26,287,549,164 £248,274 76.82%
mod.uk £127,000 3,098,476,656 £33,399 73.70%
nhs.uk £2,610,000 74,545,534,224 £807,037 69.08%
dwp.gov.uk £197,000 4,832,969,760 £57,915 70.60%
number10.gov.uk £86,000 6,162,964,164 £66,216 23.00%
dft.gov.uk £810,000 2,949,687,360 £27,408 96.62%
fco.gov.uk £645,000 5,740,488,720 £59,572 90.76%

In some cases savings of over 90% puts forward a strong argument for using a cloud based php platform. Hosting sites that consistently receive high traffic and requests could be simplified by taking advantage of the scalability and clustered nature of a cloud solution that can respond to unpredictable traffic changes that are likely to occur on informative government sites depending on events in the country.

Currently, excluding the investigation of performance value for money, their seems to be no value gained from government expenditure, and no clear links between uptime, cost per request across departments and hosting expenditure. We have shown that value can be gained from investing in migrating to an open source cloud solution, and that using their current development and hosting expenditure, the government receives no value across the board from this investment.


Additional Notes

(1) The following web sites have no service availability data:

(2) A slight discrepancy we discovered while testing these sites, dius.gov.uk and bis.gov.uk direct to the same site, yet have different development costs. Again I received no response from either department about this matter.

(3) I attempted to contact all the departments in question but received no response from their press offices. I wanted to ask if any development had recently been carried out on the nhs.uk and dwp.gov.uk sites as their development/build expenditure is so high. I was also looking for a response as some sites had no service availability data, but received none.


Image courtesy of Alan Cleaver and Creative Commons

Joe Gardiner Head of Product

Joe is the Head of Product for CatN. He oversees product development, customer engagement and commercial activities. You can find him on Google Plus.