A vision for Digital Services 3
I blogged recently about governance principles for good procurement. But how might we put those into practice in Digital Services 3?
The procurement of creative work in the public sector is an unsolved problem. Most procurements are attempts to reduce subjectivity and variance, to make things comparable so that purchasing is based on objective assessments of value. All of which is fair enough, and for some things, the process works. But there’s no one approach that’s fit for every problem. And this approach, for digital, leads to perverse outcomes.
The work of government in the wake of the Better for Less report has done much to address these issues. But with the exception of G-Cloud, we’ve continued trying to fit procurement for IT and digital services into the same model we use to buy everything else. It’s time that we accepted that for creative services this model is irreparably broken, because our attempts to eliminate subjectivity and variance inevitably stifle exactly the creativity that projects need to succeed.
Fixing these problems structurally will require setting aside some of our assumptions about how these things have to work. Ultimately, it’ll require lobbying and change internationally. But for the time being, as long as government is willing to apply the regulations open-mindedly, I think we can go a long way.
We’re already half way there
The G-Cloud framework is the biggest shakeup IT procurement has had, ever. And it’s a runaway success, by every measure. It’s easy to get on, easy to sell and easy to compete. And plenty of brilliant creative work has been sold through it.
G-Cloud measures success and limits abuse by publishing spending data – not to mention service descriptions and prices. The selection process is minimal. The process, in general, is minimal. Digital Services needs to apply these lessons, and there’s no reason it can’t.
One of G-Cloud’s biggest, boldest decisions – and hopefully its lasting legacy – was to prove that you can push selection to the point of purchase, instead of having a big, up-front process to assess suppliers’ quality.
Selection questionnaires are fiction
Selection needs to happen at the point of purchase, and it needs to be based on reality. Procurement, as it stands, suffers from a major case of big design up front.
Selection questionnaires suffer from all the same problems as detailed specifications do in software development. They create the illusion of certainty, because they have lots of detail, and require lots of effort. Their comprehensiveness makes them feel rigorous, and coming up with the questions makes buyers feel reassured that they’re doing their due diligence.
But ultimately, none of it is real. It’s all just words. Until you actually get stuck into solving real problems, you don’t know what you’ll need. You won’t know what works. You won’t really know what your users needs are until you’ve failed to meet them at least once. And so, no matter how good the questions and how detailed the answers, it’s all supposition. Multiply this problem by a thousand when the selection questionnaire is for a framework, and not even for a particular project.
Multiply it by a thousand again if you publish a scoring matrix for the answers: scores might feel objective, but in reality, they just distort the process even more. Scores mean you need scoring criteria. Criteria need to be published, otherwise sellers don’t know what they’re being scored on. But published criteria make game-playing inevitable, and game-playing means you’ll just hear what you want to hear.
Ultimately, this process we follow to increase objectivity and reduce risk does neither. Arguably, it reduces objectivity and increases risk, because it distorts our view of what’s really going on. This is why, in software development, we take an iterative approach: preferring working software over detailed plans, so that we can make smaller plans based on evidence instead of assumption.
The same principles can work in procurement, and the Service Manual is already set up to support working in exactly this way: it breaks projects into phases you can procure separately, so that teams never need to let one big contract. And it requires the use of open standards and technologies so that no one gets stuck with a proprietary solution provided by a supplier they can then never leave.
Even a good selection questionnaire is bad
Even if you set aside these problems, and imagine the best selection questionnaire it’s possible to have, you still don’t end up with something fit for purpose. Because selection questionnaires are incapable of providing the information assessors need to make good decisions.
They can’t effectively assess working culture, mindset or team fit, because they don’t involve meeting or talking to anyone. Projects that depend on close working relationships between people are frequently procured in ways that prohibit the people involved from meeting at all. Sometimes buyers try to capture these things in questions, but trying to communicate culture and team fit in writing is bound to fail. All you’ll end up with is a game of unsubstantiated buzzword bingo.
As a result, at best, you get a very narrow view of potential suppliers, and make decisions based on very incomplete data. At worst you get no view at all, and make decisions based on pure fiction.
Let’s give them the boot
Selection questionnaires in G-Cloud are selection questionnaires in name only. What they really are is a data collection mechanism to power the filters in the Digital Marketplace. Which is a good idea.
Let’s have a framework for digital services where pretty much anyone selling relevant services get on, and then push selection to the point of purchase. That way, delivery teams can select suppliers in ways that make sense for their projects.
Let’s recognise that a good process for picking someone to build a new intranet might not be quite the same as one for a major digital public service, even though the delivery skills involved are similar.
Let’s focus the framework on making suppliers discoverable and helping everyone to stay within the law.
This approach relies on delivery teams knowing what they need, which is not always the position they’re in. That’s a problem. But it’s the situation we’re in anyway, and not one that a procurement process can solve. A questionnaire can never make up for an uninformed buyer, no matter how well designed. Even if it did, the buyer would still be uninformed for the whole of the rest of the project.
For that reason, we need a parallel function in government to help people learn, help organisations recruit for the right skills and expand their capability. If we do it well, we can make process simpler. If we try to make procurement protect buyers against themselves, we’ll inevitably end up with bloated, slow process – worse still, it will be a crutch that will impede progress towards ensuring that more of the public sector has the skills it needs.
Procurement process must assume that buyers are well-informed, and act accordingly.
Sensible process, sensibly enforced
Once again, G-Cloud shows the way. Search, longlist, shortlist, clarify, purchase. This process works for G-Cloud, and there’s no reason it can’t work for Digital Services.
But we can improve on what G-Cloud does. The digital marketplace shouldn’t just be a catalogue. It should be the tool used for all procurement through G-Cloud or Digital Services Framework – and ultimately, all frameworks. In the age where the web is pervasive, most procurement still happens in spreadsheets and word documents. Why?
Searching, longlisting and shortlisting should all happen through the marketplace. When contracts are let, we should do that through the marketplace. When invoices are issued, the marketplace should manage that too. There should be no need to prepare and submit management information: it should just exist in the marketplace, to be queried as needed. We shouldn’t have to wait for spending data releases: it should just be on the performance platform, all the time.
And the marketplace should be built to enforce the process. Some rules will always be necessary, but why bother with a rule if you can just have code?
Marketing exists, and people talk to each other
One feature of this approach is that it will produce a large number of suppliers in the catalogue, as it has on G-Cloud. This is both good (more competition) and bad (harder to find good suppliers in the crowd). But it gets easier if we abandon the idea that all information relevant to a procurement should be found and consumed as part of the process. It just doesn’t work like that. Companies supplying the public sector have marketing. They go to events. They talk to people. They have relationships. They advertise and they run promotions.
Companies that do this well will be more visible, so they will be shortlisted more often than suppliers who fail to stand out from the crowd.
Instead of designing an approach that tries to suppress this reality, let’s embrace it to make things easier. Companies that do a better job of describing their value to the public sector will win more work from it. Companies active in other industries that join frameworks but make no proactive attempt to develop a client base in the public sector probably won’t get much benefit. Which is entirely to be expected. This is how it works in every other industry. Why not ours?
Frameworks should let everyone in, but getting on a framework is never going mean getting a free ride. Fairness and prudence with taxpayers’ money does not require government to knock on every supplier’s door for every project. Good engagement with buyers will help companies to stand out from the crowd, which lessens the cost of dealing with a large number of potential bidders.
Of course, we must work hard to prevent fraud, abuse, cabals and cartels. But we’re now at the point where the process is so constrained that it leads to failure and waste on an extraordinary scale. What’s the point in a fair process if it can’t deliver?
Be open: it makes things better
Of course, a framework as open as this could easily be abused. GDS have a reasonable interest in ensuring that projects are delivered per service manual principles, but not everyone’s on board.
Outwardly, there would be nothing to stop an old-school public body buying a waterfall project through Digital Services, given that the oversight is minimal. And nothing to stop buyers running selection processes in bad faith, where the winning supplier has really been chosen before the process has even begun. This does need to be addressed.
So the marketplace should list everything that’s sold. Every project, every invoice, in real time. The process should facilitate conversation: between buyers, between suppliers, between buyers and suppliers. People should be able to see what’s going on. Indeed, they should be required to know what’s going on before they buy anything: how else can we prevent duplication? How else can we ensure that delivery teams learn each others’ lessons? This isn’t possible if you do procurement on spreadsheets, but on the web, it’s a synch.
As a by-product of the process, we should gather the data that lets buyers to look at past performance and experience. We should be able to see a list of every project being delivered. Spending data should be published in real time as well. If someone keeps letting contracts to their mate from the pub, we’ll soon find out.
And government need to support the process. There needs to be active, hands-on help and advice for buyers. This should be a core function of the Government Digital Service, and is reason enough to create an similar body or function for local government. There also need to be mechanisms for peer support to help make this practical.
And there needs to be audit and review – if someone’s caught selling services that are outside the scope of the framework, they should get the boot. It wouldn’t take that happening very often for people to get the message.
Let’s get started
We’re genuinely not far away from making this much better. We’ve trialled new approaches. We’ve had setbacks. But we’ve had successes too. We know what we need to do next. Suppliers, buyers, policymakers and politicians are all supportive and aware that change is needed.
All that’s left is to get it done.