[insert word] soar [out of control]

image Something or other is always soaring as far as the newspapers are concerned.  It’s possibly their – tabloid & broadsheet – favourite word.  Whether it’s losses, costs, assaults, shares, emissions, corruptions, deaths, prices, provisions, sales, spirits, revenues or bankruptcies, they’re all soaring. Even when the market is crashing, something else is always soaring, echoing Newton’s what goes down must go up somewhere else. In the very worst case, of course, soaring things are actually soaring out of control.    As a word it’s far easier to use in multiple circumstances than “crash” (too many aircraft/helicopter/car connotations), “shrink” (nearly always financial, occasionally psychological).  The only word that appears to be used in equal measure is “fall”, although, again following Newton, they’re often paired as in “Wie continues her long fall as Matthew soars”.

image So I feel perfectly justified, if very late, in saying “Costs of government websites soar” on the basis that “soar” applies when things go up a little, a lot or enormously and even when the baseline is unclear.  Just searching for “costs soar government” turns up some gems. The UK child database has seen its costs apparently soar although it’s entirely unclear from what to what. In Durham, the costs of a reorganisation have, it seems, soared from �32 million to �12 million and back to �30 million (you’re confused? Don’t worry, I was too).  It would be helpful and perhaps Dan could oblige, to have a grading scale for when soar is to be used – for instance, a 5% increase is hardly a soar, but a doubling could legitimately be called “soaring out of control” unless, of course, there was evidence of control (which would usually only be found long in retrospect when the various audit bodies come along to see how things went … “costs have soared, more than doubling from their original forecast, however the significant increase in requirements/lengthy delay in securing planning permission/complete obfuscation on part of the management team appear to justify fully the delta.”)

So back to government websites, “soar” seems appropriate because I’m not sure we know how much they used to cost or, really, how much they cost now.  Every year a parliamentary question is submitted to most departments (called a round robin) and each minister would reply with what they knew.  Some costs would straight away look out of whack – too high or too low – although the high numbers were generally taken as honest and the low ones were generally taken as missing out a few important numbers, not through malice but because the question was never phrased in a way that would ensure all of the true costs were considered.  I very much doubt that any country in the world has a proper handle on how much its government websites cost – and I suspect “soar” is an appropriate word to apply to them all over the last decade.

In the past, I had several goes at figuring out how much UK government was spending on its websites, using bottom up data, private sector comparators and top down estimation.  The 3 slides below come from presentations delivered, respectively, on 21/5/2003, 28/9/2004 and 13/5/2004.  In the first I estimated government websites to cost �140-250 million in total, based on counting how many pages existed and figuring out a metric for frequency of maintenance and then cost of maintenance; in the 2nd deck, the page count has moved from 2.6 million to 5 million (in a little over 15 months) and you could therefore imagine the cost being nearer the top of my range or perhaps even higher; in the last deck I estimate costs being in a much higher range, from �400-700 million- this was after analysing the various PQ responses and also taking some private sector comparisons where even lean and mean websites – operating entirely on open source, in-house managed software – were found to be costing �150,000 per year when all costs were included (such as content author time, servers and bandwidth) – you try multiplying 3,233 websites by �150,000 lowest possible realistic, fully loaded cost – pretty soon you’re talking real money.  I have no idea which number is correct and I’m pretty sure that no one else has either and mine are as likely to be right as anyone’s.  The recent NAO report, delivered by an old friend, Patrick Dunleavy, suggests that the current costs are �208 million a year.  He might be right, he might be wrong. He’s more likely to be wrong:

PWC event slides - 21.05.2003 - tadpole slide IDeA slides - 28.09.2004 - website curve Dan Jellinek Transform slides - 13.05.2004 - website curve

Three conclusions pulled from the exec summary of the NAO report help answer why the number might be wrong, with my notes in []:

  1. Government web sites tend to be text heavy and complex to understand and to navigate [text heavy and complex to navigate means, for me, complex to maintain and, therefore, probably not well maintained]
  2. Many agencies have little information about how much online provision of services costs [see my next paragraph below]
  3. Most departments lack sufficient information about who is using their sites and how they are being used [see my previous blog post about the difficulty of establishing a business case for a government website]

How Much Should A Website Cost Anyway?Too many websites v2 .. web cost curve

To the second bullet above, quite regularly the answer to the round robin PQ would be “unknown” either because the costs were all wrapped up in a bigger outsource deal or because the cost of finding out was greater than �500 or so, which is the trigger for a disproportionate cost.  I’ve had a bit of fun analysing these costs in the past when trying to figure out what the breakeven price for a centrally managed content management system would be (what became DotP which, of course, is no more).   I came up with this hopelessly approximated curve (on the right) – it’s at times like this that I wish for the insight and mathematical capability of the late and much-respected Chris Lightfoot.  I used data from PQs and from samples that I collected here and there.  There’s probably more data available now and I really should do it again.  My point was to show that there were certainly plenty of departments that could operate at the left of the curve (90% of government websites, in 2004, had less than 2,000 pages; 1% had more than 200,000), but that there was a point when site management costs would get much higher (and that’s not necessarily a function of page count – the government gateway has a handful of pages but is more expensive than most by an order of magnitude I would guess, but then it’s doing far more that isn’t strictly “website” related).

In the graphs higher up the page, you’ll see that I always expected government would, one day, start cutting back on its web presence – and that one of the benefits of this would be increasing volume of usage and, specifically, transactions (fewer sites would make it easier to find what you needed – and I don’t mean “google” what you need but genuinely “find” what you need and find it in a way that is designed around the citizen rather than leaving you to piece together 101 bits from a dozen different departments or government entities).  It followed, then, that the operational cost (software + hardware + bandwidth + system support + editorial effort) would then be eliminated for each site that was closed down; some of this money could be redeployed at the centre (or in several centres if there were several main sites, such as NHS, business, citizen – although there are strong arguments for (citizen+NHS) and business as the two main sites),some could be saved and some would be paid in exit costs.  I have a whole deck on how to rationalise government websites (from March 2004) that, flicking through now, I realised proposed having a competitive environment of a few providers of content engines (probably as managed services) that would compete for business.  I’ll post that another time.

Dealing With The Unknowns – Pricing Central Infrastructure

All of this work, and worries about the total ongoing cost of websites – staying up to date with refreshes, adding new capability (such as transactions) and so on was what got me into DotP in the first place.  But one of the big problems was how do you charge for it?

Plainly you can’t charge the first department that adopts a central solution 100% of the cost of build, and you probably can’t charge the first two departments 50% each – so you have to estimate that if you got, say, 10 departments, would they all be willing to pay 10% of the cost and could you get ten on board quickly enough to cover the overdraft from building it?  Commercial companies face these choices every day with every product; government almost never faces it – its transactional services, all long established, are charged (to the customer or to HM Treasury) at: cost to provide divided by number of transactions plus a fudge factor to provide for future investment (whether it be IT, infrastructure, people or pensions).  Beyond that, each department has to factor in whether it will actually get its desired funding from HMT and then allocate money as it needs within their settlement.  It’s not simple, but it’s easier than trying to price central infrastructure from scratch.  Had the government gateway, ukonline and direct.gov not received central funding from [an inspired] HMT, there is no question that they would never have got off the ground.

Pricing Shared Services

Getting to a point where you can deliver shared services in a government is actually pretty difficult without (a) central funding to build it and cover initial hump costs, (b) a known customer base usually with some kind of mandatory commitment to join issued from the top of the organisation and (c) a flexible, long-term service provision contract that doesn’t see costs jump disproportionately with volume, time or refresh.   Of course we, and other governments, price shared services all the time – hospitals for instance.  They’re built and operated at a known cost for decades – someone has to budget for all the infrastructure upgrades, staff, drug, administration and provisions costs; and managed any associated indexation (and, of course, deal with any central or local reduction in budgets as things change around them). So what’s the difference with IT shared services?   That, I think, is a topic for another time.  But comments welcome as always.

3 thoughts on “[insert word] soar [out of control]

  1. There\’s a really big argument to offer the services for free, and not to give the money out to allow them to use much else. This surely will encourage software reuse.

  2. The \”free\” approach is interesting: someone has to fund the development in the first place – and that funding must go towards making something reusable (what we used to call \”productising\” – awful word) and reasonably generic (meaning that not everything that\’s needed is in there because some functions are just too specialist).The counter argument to \”free\” is that people value what they pay for – and government departments are no different. They need a vaguely commercial relationship so that they can be sure there is a long life for the product, that there are service levels and that there is a \”throat to choke\”.Finally, the original funders want to see a return on their investment at some time; or they want to see that they won\’t have to continue funding live operations.Every product that has been launched free – gateway and knowledge network were the first two in the UK – eventually ends up charging for services.

Leave a Reply