Thursday, November 19, 2009

When Agile Doesn’t Work

imageI am a huge fan of Agile software development and since becoming Elcom’s Technical Director I have made it my business to push us to become more Agile at every step. But a few months ago I realised I was pushing a square block into a round hole.

It started with a realisation that most proponents of Agile software development had a common single clause that modified their promises of great results.

Customers must get involved for Agile to work.

In Scrum it’s the Product Owner, and it must be the single wringable neck – the person accountable to the business for the project’s delivery of real value. It is the person who not only understands, but can make decisions about functionality, business needs and the user interface.

Oops.

While this sounds great, and you would think that people paying you tens (if not hundreds) of thousands of dollars would want to get involved, we had found all too often that big promises of ongoing involvement disappeared as soon as the project kicked off. Clients wanted to tell us what they wanted, approve our quote, pay us the money, and then pretend we didn’t exist until we had finished their project. Or they had a committee of thousands that wanted to give us their pent up IT requirements, regardless of whether it was in scope or not, and often in conflict with each other.

To be fair to our clients, we often are simply implementing our software and the amount of customer involvement is quite low – however that didn’t fit the Agile methodology we wanted to work within.

Things became clearer for me when I attended a ThoughtWorks Quarterly Technology Briefing in Sydney, and got the chance to hear Martin Fowler speak about how to fail in Agile adoption. During the Q&A I asked him how he suggested we handle customers who were reluctant to get involved.

Martin pointed out that there are in general two sorts of projects:

  1. Strategic: Key to business success, and as such important.
  2. Infrastructure: Necessary to do, but not seen as particularly vital for business success.

Immediately I saw that for many of our clients their projects were internally seen as necessary (new intranet, re-skinned website) but were not particularly risky or impactful on the organisation’s success. I mulled that answer over for a while and wondered what it might mean for our use of Agile.

MaisterProfessionalServicesFirm A few weekends later I was re-reading David Maister’s classic book Managing the Professional Service Firm and came across his definition of the three basic types of professional service delivery. He writes for any type of professional service and uses examples from disciplines as far apart as architecture, accounting, engineering and management consulting to make his point.

Brains/Expertise

  • Client problem is at forefront of professional or technical knowledge
  • “Hire us because we’re smart”
  • One-off projects
  • Highly skilled and highly paid professionals

Grey Hair/Experience

  • Clients require highly customised output, more familiar problems
  • “Hire us because we’ve done it before”
  • Specialised knowledge in project area
  • We sell our knowledge, experience and judgement

Procedure/Efficiency

  • Client problems are well-known and familiar, solutions also well-known
  • “Hire us because we can do it for you”
  • Repeat projects, well understood, standard solutions
  • Well trained and supervised junior staff

image

Maister points out that you can see a continuum from Brains to Grey Hair to Procedure for a range of different service features.

image

Using this idea, I could see that most of the “infrastructure” projects we had involved procedure or efficiency services. A simple deployment of CommunityManager.NET with implementation of a specified basic website design and some custom styling of forms and other modules. Agile simply slowed these projects down and was generally ignored by the team doing most of this work.

On the other hand some of our custom development work fell into the realm of grey hair or experience services – enhancing or extending CommunityManager.NET for a client. These usually did not need to be Agile projects, but were being forced into our product sprint cycle, even when they were a day or two of custom work not dependent upon enhancements to our core product.

Our product work deserves a mention here. Whilst not for any specific client, this is in fact the work that provides the greatest return on investment for the business. However a need for cash flow meant there was a limit to how much could be done at any given time. We had a settled sprint duration of 6 weeks (5 weeks dev + 1 week regression testing and release preparation), wit 1 week in between sprints, that the team had a good rhythm with. Agile was a great fit for this work.

After this analysis there was a small, but important part of our business. Small in terms of number of projects, but important in terms of revenue. This was the projects that involved significant custom development work and had little dependency upon CommunityManager.NET. Sometimes these jobs could have been delivered without our product at all, but using it gave a convenient head start to the project. These projects were strategic to the client, but as a result often had loose, or non-existent requirements because it was acknowledged that the requirements would change during the course of the project. This type of work suits Agile methodologies, but the business needed to recognise that this work shouldn’t be sold the same way our other services were.

Clearly Agile suited some parts of our business, but was a poor fit for others. Our sales material made a big deal of our Agile methodology, but it was not always relevant or appropriate for our clients. Worse, working in an Agile manner was causing many projects to have a greater chance of failure, and to delay client solutions.

Agile Wasn’t Working

I now knew why Agile wasn’t a great fit for much of our business, but it still had value in some areas. Maister also made the point that when for a professional service firms to mix Brains, Grey Hair and Procedure work successfully it is necessary to create divisions within the firm to allow each to have a management style and culture that fit their style of work.

I really like the way that Jurgen Appelo organises his development teams, with a project manager assigned 3-5 other staff as their permanent team, with them needing to complete each of their projects using that team. By pushing resource constraints to the lowest level (that of the project team) they ensure that other projects’ resource variances do not adversely affect them. However I know that Elcom is much smaller than Jurgen’s company, and such a profound shift would create more friction than it would generate in gains.

I discussed these ideas with my boss and some of my colleagues, but felt that we quickly needed to make a change as a large custom development project was about to kickoff its second stage.

The team structure we are now heading towards is this:

Delivery Team

This is a team made up of web designers, front-end web developers, and .NET analyst/programmers. Their role will be to implement new customers’ versions of CommunityManager.NET as painlessly and efficiently as possible. They will handle custom development work that is tightly integrated to, or built around, CommunityManager.NET. They work in traditional BUFD (big up-front design) projects with well-scoped requirements. Project duration would usually be anything from 1-8 weeks.

Solutions Team

This is a team made up of user experience (UX) consultants, designers, and web developers. It is the renaissance team that is responsible for finding new ways to deliver innovative solutions for strategic client problems using design thinking and Agile software development methodologies. They work in a more flexible manner, preferably with a pseudo time and materials approach, with the client purchasing entire sprint’s worth of work. Project duration would usually be more than 12 weeks, with usually two week sprints.

Product Team

This is a team made with a web designer, front-end web developer and .NET developers. It is responsible for enhancements to, and strategic development of our products, particularly CommunityManager.NET. They follow an Agile software development methodology and are very much the traditional product-centred Agile team. Sprint duration would depend upon the ability to regression test in a timely manner – but for the moment this would be around the same 6 week duration we had used previously.

Right Now

When we first made this change, there wasn’t enough strategic product work for us to justify having a dedicated Product Team. Instead members of the Delivery Team are made available as and when required to work on the product. We still have 6 week product sprints.

The Delivery Team has started transitioning away from Agile, and their time is now more easily shifted from one piece of client work to another (they previously were committed  up to 6 weeks in advance for a particular sprint). Our clients have started to notice that we have become more flexible and responsive to sudden requests. Our project managers have been learning a new resource planning tool which is slightly clunky (built in Excel, requiring double data entry) but that we soon hope to replace with a more effective system.

The Solutions Team has had some success working in a more Agile manner, but they need to learn to do it without the moral, emotional or practical support of the rest of the technical teams. By removing many of the restrictions that product work placed on them, they are now free to explore new technologies and test them in practice before we roll them into our products. For instance they have used Domain-Driven Design (DDD) and LINQ2SQL.

The teams are differentiated by more than just name. They have different KPIs, different management styles and different things to get good at.

The Delivery Team needs to become more process-driven. Too many things are left to individual choice when the best solution is already known and has been tested in practice. Also documenting requirements up front is a skill set that needs to be re-discovered and applied. After all if you are going to do BUFD projects then you need to be good at it.

The Solutions Team needs to become more Agile, and look at ideas like TDD/BDD and more comprehensive automated testing so that short sprints do not lead to a reduction in quality. Selling a job as an Agile development project is proving harder than we would have thought. It seems lip service to Agile is easy to sell, it sounds good and requires few changes in the way clients work. Real Agile development requires an altogether different mindset, and this is much harder to sell up front. We may well need to lead these projects with an initial fixed price project that uses BDUF, and then transition to a more Agile approach once we have some credits in the trust account with our client.

The Future

The Solutions Team will be undertaking a major strategic product initiative for Elcom over the Christmas period and will bring their Agile approach and technical expertise to bear in order to give us a major new element to our product for CommunityManager.NET version 7.0. After this we hope to have another major project lined up, but if not then they may become our Product Team.

If another Solutions Team project does come up, then the next step will be to create a full-time Product Team. We believe that version 7.0 will create enough demand that a full-time team will be justified. An important part of this strategy is to find partners that are comfortable enough implementing CommunityManager.NET that our Delivery Team is not a bottleneck on our sales. One of the keys to version 7’s success will be the way that it makes it far easier for digital agencies and systems integrators to use our product for their own web projects.

After that fame, fortune and fun might consume our days – or more likely we will find ourselves looking beyond enterprise content management to create a new market niche in the manner of good Blue Ocean strategists. Whatever we do, Agile will remain a part of it, but only when it deserves to be.

6 comments:

  1. Thanks for this great post Angus, sharing your experience with implementing Agile. It's good to see that it was recognized that Agile methods might be a great fit for some but not all areas of the company, but that Agile principles (transparency, flexibility, client involvement) can still work and be applied.

    ReplyDelete
  2. Thanks for the encouragement Robert.

    I checked your website out and Scrum'd looks interesting!

    ReplyDelete
  3. Thanks for the mention. If you use it we'd love your feedback on how we can make it better.

    ReplyDelete
  4. Hi Angus, great post. Nice to see the changes so clearly articulated. The change has been positive and surely indicative of an organising reaching a deeper level of technical maturity. Well done!

    ReplyDelete
  5. Really Nice and Interesting Blog Post! we can share the information about the management software. Anyway, Thanks for the sharing of such information, revenue forecasting software .

    ReplyDelete
  6. Great Article Angus.
    Once understood it sounds simple, but getting to it is always the difficulty ;)

    Thanks,
    Bora

    ReplyDelete