Complexity is a source of income in open source ecosystems

Derek Jones from The Shape of Code

I am someone who regularly uses R, and my interest in programming languages means that on a semi-regular basis spend time reading blog posts about the language. Over the last year, or so, I had noticed several patterns of behavior, and after reading a recent blog post things started to make sense (the blog post gets a lot of things wrong, but more of that later).

What are the patterns that have caught my attention?

Some background: Hadley Wickham is the guy behind some very useful R packages. Hadley was an academic, and is now the chief scientist at RStudio, the company behind the R language specific IDE of the same name. As Hadley’s thinking about how to manipulate data has evolved, he has created new packages, and has been very prolific. The term Hadley-verse was coined to describe an approach to data manipulation and program structuring, based around use of packages written by the man.

For the last nine-months I have noticed that the term Tidyverse is being used more regularly to describe what had been the Hadley-verse. And???

Another thing that has become very noticeable, over the last six-months, is the extent to which a wide range of packages now have dependencies on packages in the HadleyTidyverse. And???

A recent post by Norman Matloff complains about the Tidyverse’s complexity (and about the consistency between its packages; which I had always thought was a good design principle), and how RStudio’s promotion of the Tidyverse could result in it becoming the dominant R world view. Matloff has an academic world view and misses what is going on.

RStudio, the company, need to sell their services (their IDE is clunky and will be wiped out if a top of the range product, such as Jetbrains, adds support for R). If R were simple to use, companies would have less need to hire external experts. A widely used complicated library of packages is a god-send for a company looking to sell R services.

I don’t think Hadley Wickam intentionally made things complicated, any more than the creators of the Microsoft server protocols added interdependencies to make life difficult for competitors.

A complex package ecosystem was probably not part of RStudio’s product vision, at least for many years. But sooner or later, RStudio management will have realised that simplicity and ease of use is not in their interest.

Once a collection of complicated packages exist, it is in RStudio’s interest to get as many other packages using them, as quickly as possible. Infect the host quickly, before anybody notices; all the while telling people how much the company is investing in the community that it cares about (making lots of money from).

Having this package ecosystem known as the Hadley-verse gives too much influence to one person, and makes it difficult to fire him later. Rebranding as the Tidyverse solves these problems.

Matloff accuses RStudio of monopoly behavior, I would have said they are fighting for survival (i.e., creating an environment capable of generating the kind of income a VC funded company is expected to make). Having worked in language environments where multiple, and incompatible, package ecosystems existed, I can see advantages in there being a monopoly. Matloff is also upset about a commercial company swooping in to steal their precious, a common academic complaint (academics swooping in to steal ideas from commercially developed software is, of course, perfectly respectable). Matloff also makes claims about teachability of programming that are not derived from any experimental evidence, but then everybody makes claims about programming languages without there being any experimental evidence.

RStudio management rode in on the data science wave, raising money from VCs. The wave is subsiding and they now need to appear to have a viable business (so they can be sold to a bigger fish), which means there has to be a visible market they can sell into. One way to sell in an open source environment is for things to be so complicated, that large companies will pay somebody to handle the complexity.

The Product Owner Delta

Allan Kelly from Allan Kelly Associates

ValueAddPO-2019-07-1-08-19.jpg

As regular readers might know I’m working on a book called The Art of Product Ownership to be published by Apress later this year. One of the chapters is entitled “Why have a Product Owner” and a few days ago a bunch of ideas crystallised into this…

The aim of the Product Owner is to increase, even maximise, the business value delivered by the team as a whole. The Product Owner does not so much create value themselves as increase the value created by others.

Think of it like this: if the team randomly selected work to do and delivered it to customers then some value would be created. (For the moment I’ll ignore the scenario where that work detracts from the existing value.) The aim of the PO is to ensure the work done creates more value than a simple random selection. The greater the difference, or delta to use a mathematical term, between random selection and an informed selection the better.

The general hypothesis is that intelligent selection of work by a skilled Product Owner will result in both more value being delivered and an increasing delta between intelligent PO selected work and randomly selected work.

This difference the value added by a Product Owner. I like to call this difference the Product Owner Delta.

Now in real life work is seldom randomly so Product Owners are not competing against random selection. In some cases the alternative to a designated Product Owners is someone else: a senior developer, an architect, a manager or someone else. In such cases this person is taking on the Product Owner role. They may not have the title, the aptitude, the skills or official position but when work is selected by one person they are de facto the Product Owner.

In other cases the alternative to the PO might be selection by consensus on the team, or a sub-set of the team. Now it is entirely possible that such a group could outperform a single Product Owner in selecting work – especially is they have market and customer knowledge, some analysis skills, time to do the background research and so on. In some cases this works, for example think of a small start-up staffed by software developers creating software development tools.

However, in some cases selection by committee might be inferior to a random selection. Imagine a team which has never met a customer, argue about what to do, duck key decisions and never say No to any request. Its easy to image a dysfunctional selection committee.

There is more to increasing the Product Owner Delta than simply selecting the highest value items. Timely selection can help too. If decisions are not being made, or committees are spending a long time making decisions then having one person simply make those decisions in an efficient, timely, manner can increase the delta.

Time has another role. Because of cost-of-delay simply selecting the highest value items at any one point in time does not maximise the value delivered. Time Value Profiles (see Little Book of User Stories or my presentations on value “How much? When?”) expose this and need to be another tool in the Product Owners repertoire.

And of course, the Product Owner Delta is not the only reason to have a Product Owner in the team, but it is probably the main reason.


Like this post? – Like to receive these posts by e-mail?

Subscribe to my newsletter & receive a free eBook “Xanpan: Team Centric Agile Software Development”

Check out my latest books – Continuous Digital and Project Myopia – and the Project Myopia audio edition

The post The Product Owner Delta appeared first on Allan Kelly Associates.

Story Generators

Allan Kelly from Allan Kelly Associates

iStock-913773630small-2019-03-22-17-35.jpg

Recently I’ve been looking again at Jobs to be Done and OKRs (Objectives and Key Results). I increasingly see them as story generators and a potential solution to the tyranny of the backlog I described last time.

When I first looked at Jobs to be Done (and OKRs actually) I wondered if they constituted a fourth, top, level on top of Epics, Stories and Tasks. I’ve long argued against having more than three levels of things to do (or requirements as we used to call them.) There are big meaningful things to do (stories), really big things which we don’t as yet understand but look really valuable (epics) and the immediate small things to do right now (tasks).

Actually, I’d rather think most things can be dealt with by two levels and one level is the even better. So adding a fourth “even bigger” thing on top of Epics just felt wrong. Technologists (like myself) have a tendency to map everything into hierarchies; inverted trees with fractal like branches. But not everything is, or should be, a hierarchy, mapping the world into a tree like structure can add complications.

Unlike stories (and epics and tasks) Jobs to be Done don’t really lend themselves to the transactional “Done”. While you could put a Job all the way to Done on your Kanban board and track it from “To do” to “Done” in reality the customer job still exists. Sure you’ve improved it but you can improve it again – another example of Stable Intermediate Forms. This seems to be the great potential of Jobs to be Done, they keep on giving: as much as you improve your product to help with the job you can still improve it some more.

So each time you analyse the Job to be Done you should be able to find more stories to deliver to improve it. Hence the Job to be Done is not a “story” to do, it is a Story Generator. Every time you look at the job to be done you find more stories, every time you examine the result of the latest improvement you find more stories. The job will never be done. Some might see that as a bad thing but that also means the job presents a stable focus for ongoing work.

The same might be true of OKRs but in a slightly different way. Because the objective is reviewed periodically – every quarter or so – it lacks the continuity of Jobs to be Done but perhaps allows the team to switch targets, maybe it is stable enough.

The key results may well be stories in their own right, or they may be things which lead to stories. Either way one can expect some key results to be achieved and marked as done regularly. As they fall they are either replaced by new key results building towards the objective (which themselves lead to stories) or new key results are added for new objectives.

I’m sure there are other story generators out there but the key thing for me is not the mechanism but the existence of the generator. Once you have a story generator you do not need a big backlog of things to do. The generator will replenish the backlog whenever you need more stories – either because you have done them or the value has fallen.

Using a generator removes the need to have a big backlog which removes the tyranny of the backlog. The team are now free(r) to concentrate on delivering value towards their objective.

Finally, I wonder if anyone has used both OKRs and Jobs to be Done together? Right now they feel like alternative generators to me, having both seems like a bit like overkill. Although I accept that maybe OKRs are more corporate and Jobs to be Done are more product focused. Anyone got any experience using them together?


Like this post? – Like to receive these posts by e-mail?

Subscribe to my newsletter & receive a free eBook “Xanpan: Team Centric Agile Software Development”

Check out my latest books – Continuous Digital and Project Myopia – and now Project Myopia audio edition

The post Story Generators appeared first on Allan Kelly Associates.

Agile won the war but lost the peace

Allan Kelly from Allan Kelly Associates

iStock-856693018Medium-2018-11-8-16-53.jpg

“I’ve spoken of the shining city all my political life, … in my mind it was a tall, proud city built on rocks stronger than oceans, wind-swept, God-blessed, and teeming with people of all kinds living in harmony and peace; a city with free ports that hummed with commerce and creativity. And if there had to be city walls, the walls had doors and the doors were open to anyone with the will and the heart to get here. That’s how I saw it, and see it still” President Ronald Reagan, Farewell to the Nation, January 11, 1989

Back in 2001 when the word agile appeared it was a manifesto – a set of ideas, the term “agile” also served to group a bunch of tools and techniques which could make software development “better.” More importantly to my mind, it painted a picture of a shining city on a hill we all wanted to live in.

Agile was a place you wanted to go, it was a journey you wanted to make, it offered hope. More important as the tools – sprints, stand-ups, etc. – and approaches – just in time, last responsible moment, test first – were the stories agile people – including myself – told. These were stories of a better world, of that shining city on the hill.

And not unimportantly, in a world of search engines “agile” gave you something to search for. Before agile you could search “make my software development team better” or “software development process improvement” but what you got was a very mixed offering. AltaVista (and the young Google) would suggest links for CMMI, or ISO-9000, or vendor tools to “fix it”, or proper design, or… there was no coherent message. Most of these ideas resolved around senior people making big decisions and then imposing them.

Then along came agile: it offered to involve everyone, everyone made decisions, everyone was happy and we could all go to that shining city on a hill – more than that, we all had an important part to play in building that city.

Today everyone is agile. Nobody is promoting traditional (“waterfall”) working, CMMI, PMI and everyone else has incorporated agile (to some degree). Not being agile is about as popular as leprosy.

But very few of us have reached the shining city on the hill.

Along the way agile has been watered down, in becoming compatible with everything else it is less different, it is less attractive, fewer workers are motivated to take the journey. And as “the powers that be” have found ways to bring control-and-command back to teams (maybe in the name of scaling) fewer people are invited to help build the city.

Ironically, as we (the agile community) has made agile management friendly we have made it less worker friendly. Today senior managers “get agile” and want their organisations to be agile. But those at the code face seem to have less and less motivation. And those in the middle… sometimes they seem to want to change just enough to declare success but no so much that things really change.

For some people agile has become completely discredited – I wrote Why do Dev’s hate agile? last year and I’m presenting it in London next week. Agile isn’t a shining city on a hill, agile is trench warfare.

And Googling “agile” presents a long long list of links with less and less coherence.

Agile won the war. Agile is respectable and everyone is agile now. Big business rush to be agile, Governments want to be agile, blue-chip consultancies will sell you agile.

But agile lost the peace.

While many say they are agile few software developers live in a shiny city. The place they live in might be better than the place they came from but it doesn’t live up to the dream many of us shared 15 years ago. Agile has become an excuse for failure and a thing to be imposed.

The thing that passes for “agile” today is too often a watered down version of the original dream. Worse still, we don’t have a word to describe that shining city we all want to get to. Russians have an expression for this:

“We wanted the best, it turned out like always.” Viktor Stepanovich Chernomyrdin, Prime Minister Russia, 1998-1999

Me? – I still dream of that shining city on the hill, I still believe agile is the right way to get their, I still wave the flag for agile but more and more I feel the need to explain myself and tell people that the agile I dream of is not the agile they may experience.

The post Agile won the war but lost the peace appeared first on Allan Kelly Associates.

Continuous Digital published – done?

Allan Kelly from Allan Kelly Associates

CDpile2cut-2018-10-9-14-43.jpg

Continuous Digital is done.

Probably. Maybe. Definitely maybe.

Continuous Digital is the second of my two #NoProjects books. Many people ask: “why two?” “What is the difference between them?” “Do I need to read both?”

Short answer: Project Myopia explains why the project model is bad for software development. Continuous Digital describes what to do instead.

Long answer: as the #NoProjects hypothesis grew, as I thought about it more, as I talked to others about the ideas – specifically Steve Smith, Joshua Arnold and Evan Leybourn – the ideas grew. My thinking both on “what to do instead of project management” and “why do something different” grew.

Specifically I saw that the combination of Continual Delivery and Digital Business meant there was a stand alone case for moving beyond the project model. Whether you agree with the problems I discuss in Project Myopia or not there is a case for changing the way businesses are managed.

That is why I split the too books. Project Myopia is a companion book, it is not a prequel, a sequel, a book one or a book two. It is a book some people will read in its own right.

Continuous Digital argues that since business are increasingly digital, and as businesses strive to survive and grow then technology development is not a separate “project” it is inherent to the business. Technology and innovation are business as usual.

Stopping, even pausing, work – as in the project model – surrenders competitive advantage and introduces extra costs (time, money, risk). What is needed is a new model. A continuous model.

Continuous Digital is now published on Amazon in digital form and will soon be there – and in other booksellers – in physical form. (If you can’t wait for a print copy you can buy one from Lulu where they are slightly cheaper too.)

So I’d like to say Continuous Digital is done. But…

Even before I saw the final print version I had requests for an audio version of both Project Myopia and Continuous Digital. I’m debating whether to do these, if you would buy an audio version please let me know, if enough people want it I’ll do it.

Second, once I saw and held the final, done, version in print new ideas came to me. I don’t want to revisit the text – although I might fix a couple of typos – but Continuous Digital is a big book, 350 pages. And I know many people will be put off by the size.

So I’m thinking of turning it into four smaller books, each around 100 pages in length and each corresponding to one part of Continuous Digital. Maybe.

It is never done. It is continual.

The post Continuous Digital published – done? appeared first on Allan Kelly Associates.

Business school research in software engineering is some of the best

Derek Jones from The Shape of Code

There is a group of software engineering researchers that don’t feature as often as I would like in my evidence-based software engineering book; academics working in business schools.

Business school academics have written some of the best papers I have read on software engineering; the catch is that the data they use is confidential. For somebody writing a book that only discusses a topic if there is data publicly available, this is a problem.

These business school researchers show that it is possible for academics to obtain ‘interesting’ software engineering data from industry. My experience with talking to researchers in computing departments is that most are too involved in their own algorithmic bubble to want to talk to anybody else.

One big difference between the data analysis papers written by academics in computing departments and business schools, is statistical sophistication. Computing papers are still using stone-age pre-computer age techniques, the business papers use a wide range of sophisticated techniques (sometimes cutting edge).

There is one aspect of software engineering papers written by business school researchers that grates with me, many of the authors obviously don’t understand software engineering from a developer’s perspective; well, obviously, they are business oriented people.

The person who has done the largest amount of interesting software engineering research, whose work I don’t (yet; I will find a way) discuss, is Chris Kemerer; a researcher who has a long list of empirical papers going back to the early 1990s, and rarely gets cited by papers by people in computing departments (I am the only person I know, who limits themself to papers where the data is publicly available).

#NoProjects: Project Myopia is published

Allan Kelly from Allan Kelly Associates

ProjectMyopiaNew-2018-09-10-11-17-1.jpg

Project Myopia – the original case for #NoProjects – has been a long time in the works but it is now done. Published. For sale on Amazon.

Projects fail. Some say 40% of all IT projects fail, some say 70%. And it has been that way for years. Each project fails for its own reasons but they all share one thing in common: the Project Model. Could it be the project model itself which creates failure?

Projects end. Successful software continues. Twenty-first century digital businesses want to continue and grow.

Project Myopia is available to buy on Amazon today – the physical version should joined the eBook in a few days.

Project Myopia gives the case against projects – the hard core #NoProjects arguments. A second book, Continuous Digital will join Project Myopia in a few weeks on Amazon. Right now copyediting isn’t finished on Continuous Digital, plus the physical copy needs to be worked out. In the meantime late drafts of Continuous Digital are available on LeanPub.

The post #NoProjects: Project Myopia is published appeared first on Allan Kelly Associates.

#NoProjects: Project Myopia is published

Allan Kelly from Allan Kelly Associates

ProjectMyopiaNew-2018-09-10-11-17.jpg

Project Myopia – the original case for #NoProjects – has been a long time in the works but it is now done. Published. For sale on Amazon.

Projects fail. Some say 40% of all IT projects fail, some say 70%. And it has been that way for years. Each project fails for its own reasons but they all share one thing in common: the Project Model. Could it be the project model itself which creates failure?

Projects end. Successful software continues. Twenty-first century digital businesses want to continue and grow.

Project Myopia is available to buy on Amazon today – the physical version should joined the eBook in a few days.

Project Myopia gives the case against projects – the hard core #NoProjects arguments. A second book, Continuous Digital will join Project Myopia in a few weeks on Amazon. Right now copyediting isn’t finished on Continuous Digital, plus the physical copy needs to be worked out. In the meantime late drafts of Continuous Digital are available on LeanPub.

The post #NoProjects: Project Myopia is published appeared first on Allan Kelly Associates.

Release or be damned

Allan Kelly from Allan Kelly Associates

iStock-166161352small-2018-09-6-12-26.jpg

Back when I was still paid to code I had a simple question I posed to troubled development efforts:

“Why can’t we release tomorrow?”

This short simple question turns out to be amazingly powerful. I remember one effort I was involved with in California where a new CEO took over and started cutting jobs. I posed this question to the team and in a week or two we did a “beta release” – we did those sort of things back then. Asking this question was the key that allows us to question everything, to cut the feature list – or rather push work back, it stayed on the to-do list but we didn’t let it stop us from pushing to release.

We rethought what we were trying to achieve: we didn’t need the whole product, we just needed enough of the product to work to show to one specific target customer. Even if they signed there and then we had weeks before they used it in anger. But until we released something, until we had something “done” our team, our product, look like just another “maybe.” We had to draw a line under it so the new CEO wouldn’t draw a line under us.

Saying “only do the essential” is easy and come up again and again, whether it is Minimal Viable Product, Minimal Subset, Must haves in Moscow rules, but it is far easier said than done. One persons “essential” is so often another persons “optional extra.” In this context, when I say “essential” I mean “the parts needed to make the system work end to end” – I’m far closer to the old walking skeleton idea.

I was reminded of this question by a couple of endeavours that came to my attention during the summer. Well, I say came to my attention, I feel a bit responsible. Both endeavours are happening at clients; clients who I had fallen out of touch with. My style of working is to help clients who want help, I don’t like selling myself. These clients didn’t ask for more help so I didn’t jam my foot in the door, in retrospect maybe I should have.

In one case the team were doing very well. They were iterating, they were TDD/BDD’ing, they were demoing, they were working with the client, they were doing everything … except releasing. Then one day the client asked “when will it be done?”

Now think for a moment: What if you could release your product tomorrow?

The thing is, without actual products those around the team look for signs that the team can be trusted, that they team will deliver, that the team are thinking about what is to be done. People ask for proxy-products: plans, schedules, risk-logs, budget forecasts and so on. When stakeholders can’t see progress they look for things to assure them that there is (or will be) progress (soon).

Who needs plans and predictions about the future when the future is here tomorrow?

Actual releases are they key to reaching the new world, they change everything.

So I feel guilty: I should have inflicted myself on these teams, I should have been there again and again bugging them “Go to release”, “Remove that barrier”, “Force it through”.

Being able to ship an update of your product has a transformative effect.

It demonstrates the team have the ability to do the job in hand.
It demonstrates you have quality. It obliterates the need for a test-fix-test-fix aka stabilisation aka hardening phase.
It blows away sunk costs because something has been delivered.
It removes “maybe” and “ready but…”
It is probably the greatest risk mitigation strategy possible.
It creates trust and provides a platform for solid conversations.

Most of all, a released product is a far better statement of progress than any number of plans or forecasts.

This does not mean everything is done. Sure there are things left undone but there will be things left undone when I’m on my deathbed, that is the nature of life. As much as we (especially men) love to collect entire sets there are few prizes in life for completing everything on your bucket list.

Having a released product utterly changes the nature of the conversation. Conversations are no longer full of “ifs” “maybes” “shoulds” “how long will it take?” “what are the quick wins?”. Those questions can go away. In its place you can have serious conversations about prioritisation and “what do you want tomorrow?”

This is all part of the reason I love continuous delivery. Teams can focus on real priorities and stop wasting time on conjecture.

In my book if you don’t have a releasable product at least every two weeks – say every second Thursday – you are not Agile. And if you haven’t released a product to live in the last two weeks you are probably not Agile.

I don’t care how close you get to a releasable product: it isn’t a release if it isn’t released to a live environment – close but no cigar as they say. (OK, I’ll accept the live environment may not be publicly know, or might be called a beta, but it has to be the real thing.)

Nor should you rest on your laurels once you have regular releases (to live) every second week. That is but first base. You have opened the door, now go further. There are at least 13 opportunities to improve.

If you cannot do that now then ask yourself: Why can’t we release tomorrow?

And start working to remove those obstacles:

  • Reduce the number of work items you are aiming to put in the release.
  • Fix show-stopper defects now.
  • Running tests now.
  • Get those people who need to sign-off to sign-off.

Software development has diseconomies of scale: many small is cheaper than few large.

And once you have your release you can turn your attention to making sure these things don’t happen again:

  • Reduce the amount of work you accept into development at one time.
  • Fix every defects as soon as they are found.
  • Automate tests so they can run more often. (Automate anything that moves, and if it doesn’t move, automate it in case.)
  • Find a way to reduce the time it takes to get sign-offs: remove the sign-off, make sure the signer prioritises signing or delegate someone else to sign (or automate the signature.)

If there are essential processes, activities, third-parties (or anything else) that has limited bandwidth which need to be done before release but inject delay then re-orientate your process around that bottleneck. For example, if your code needs to pass a security audit before release (an audit you can’t automate that is) then, downsize all the other activities so that the audit process is 100% utilised. (OK, 100% is wrong, 76% might be better, but thats a long conversation about queuing theory.)

Again and again I seem condemned to learn the lesson: nothing counts but working software which is used.

As for my team, and my job in California, it didn’t save me. I regret not asking the question sooner.

The post Release or be damned appeared first on Allan Kelly Associates.

Agile is the process digital technology needs

Allan Kelly from Allan Kelly Associates

1200px-Workers_in_the_fuse_factory_Woolwich_Arsenal_Flickr_4615367952_d40a18ec24_o-2018-07-18-12-19.jpg

In my presentation at Agile on the Beach last week I continued my discussion of Agile and Digital. It is increasingly clear that digital and agile are intrinsically linked. Specifically, business need agile processes to get the most out of digital technology. My “Agile, Digital & the new management paradigms” presentation is online but let me give you the argument here.

There is a long standing model of technology change – so widespread I can’t find the original source – which says change comes in three steps:

  1. First new technology allows the same processes and activities to be done better, faster, cheaper, more efficiently. In this stage new technology is used to do the same things, the processes and practices change little.
  2. Next new technology allows process and practices to be reconsidered and changed to make the most of new technology. Work becomes even better – whether that be faster, cheaper, higher efficiency, superior products, whatever.
  3. Finally new innovations appear because of the technology and new processes. One can see opportunities for new businesses, new business models, the next round of technology innovation and more.

So the whole thing repeats.

Look at the photo above. According to WikiCommons this is a picture of a factory at Woolwich Arsenal sometime in the 1800s. Notice the belts stretching from the ceiling to the workstations. These carried power, or to be more precise motion. Above the workers is the line shaft which turns. The shaft is driven by a central power (motion) source somewhere, probably a water wheel or a steam engine.

This is before electricity. The line shaft and the belts carry the power the factory needs to work. And they break, the longer they are the more prone to breaking they are. Factory design is constrained by the need to have straight lines for the line shaft and short distances between the shaft and the workstation. And factory design dictates layout and processes.

Then came electricity.

Electricity allowed each workstation to have its own motion generator. At first factory owners used electricity to do the same things faster and more reliably. They could dispense with the steam engine and thus the stokers and coal it needed. But at first they didn’t seize all the advantages electricity brought.

It took time to understand how a factory could be laid out more efficiently and how processes could be changed. When they did factories got even more efficient and faster. Some might argue that it took the coming of Lean manufacturing to complete these process changes.

The same story has played out in industry after industry with technology after technology. Think of Word processors: first they helped secretaries do their job faster, then processes changed and everyone wrote themselves, goodbye secretaries. Containerisation in the shipping industry is another. First ships loaded and unloaded faster. Then the shipping companies innovated but more importantly world trade innovated. Some observers claim containerisation was a more significant factor in trade globalisation than free-trade agreements.

Digital technology is like electricity. It changes business, it creates new opportunities for doing things differently. To get the most from digital technology you need new processes. Right now most companies are stuck – even happy – doing things faster. Only when they change processes will they get the full benefits.

Agile processes are that change.

Agile ways of working help companies get more from digital technologies. Without Agile companies using digital technologies are just doing the same old thing faster.

Agile started in software development for two reasons. First software developers had a lot of problems, they had the need to change. Second, programmers had the first access to digital technologies.

Ray Tomlinson, a programmer, was the first person with e-mail. Tim Berners-Lee, a programmer, had the first web-browser. Ward Cunningham, a programmer, had the first Wiki. I could continue.

Software developers created Agile because they needed to and they could.

This is why Agile is taking off in marketing.

Outside of technology itself marketing has probably been more exposed to digital technology than any other part of business. First with digital publishing then with social media. At first digital helped marketing departments do the same work faster. Next it changed what you could do entirely. Marketing is adopting agile because those processes allow marketeers to do a better job when working with new digital technology.

So forget all those arguments about agile being a better way of working (it is but never mind).

Forget all those stories of agile like processes and practices before 1998 (yes they existed but that doesn’t change things).

Forget the debate about waterfall and upfront planning versus agile and just-in-time (that is history).

All you need to know is:

  1. Digital technology is helping you do things faster/better/cheaper.
  2. Agile ways of working allow you to get more from digital tools.
  3. More innovation is coming.

Agile is the process for digital businesses.

Sign-up to receive these posts by e-mail and free eBook of Xanpan

Image of Woolwich Arsenal factory taken from WikiCommons, no known copyright.

The post Agile is the process digital technology needs appeared first on Allan Kelly Associates.