Architecture

The Importance of Flexible Application Design

I don’t normally blow my own trumpet, but this week the team I work with won a prestigious ‘Innovation’ award at Capita Plc for an iPad/ASP.NET/WebAPI solution we have built to address a specific business need.

The application essentially allows building Surveyors to record information about properties using an iPad in offline or online mode, sync that data up to a backend SQL database via a ASP.NET WebAPI service, and then expose the data through an ASP.NET Web Portal.  There’s also the usual Reports and dashboards that managers and such like generally swoon over.

The product itself is a great time saver, it allows surveys to be taken and reported on in a fraction of the time compared to pen and paper, or even an excel spreadsheet (by hooking costs into items that are picked by the surveyor).

As good as the solution is, from a business perspective, what really impressed the judges was how easily it it could adapted to running any kind of survey you want without have to re-code anything.

Let me explain a bit more.

Version 1 of the solution was built for a specific type of survey, and as such the original developer involved built the application with certain settings hard coded.  So for example, in this particular survey they wanted to segregate any one building into Blocks, Levels and Items, therefore to accommodate these business requirements the developer created Block, Level and Item entities.

The backend database used these entities, the front end iPad app used these entities.  And it worked fine.

Specific Entities

But then as these things go other areas of the business saw what had been done and naturally wanted the same thing for them.

Business versus Technical Design

The problem was that was different area of the business wanted something slightly different.  some wanted 4 or 5 levels, some wanted only 1 block, some didn’t really want to record costs but rather just snippets of information such as diary events during the course of a building project.

The original plan was to use the v1 as a core but then re-develop both the backend, and the iPad app for each client.

Now from a technical design point of view this is great.  We get to independently give each client exactly what they want.

However from a business perspective this really wasn’t very good.  You see there are 3 major issue with this way forward.

  • Each new client would take 3-4 months to re-develop
  • Multiple codebases – both front and backend for each client
  • no central location for management

Ultimately these 3 issues really come down to one common business problem – cost.

You see, many times the business would want it NOW.  Or at the very least within a couple of weeks, obviously being told 3-4 months PLUS testing is no good.

Secondly, although some commissions were large value, some were only for a few thousand (as it was only a handful of small properties/locations).  Again at 3-4 months the cost just becomes prohibitive.

Third, with multiple sets of code bases and no central management location, looking after all these different implementations would requirement far more support overhead – and therefore costs.

Start from scratch or accept the negatives?

It is about this time that I got involved with the project.

Immediately I saw the issue but more importantly the solution.

To some this may be obvious, but when you’re a true techie, especially someone who has already invented the wheel once, it’s quite hard to see more flexible alternatives.  After all, what does a techie care about costs?

Now I think I’m quite lazy!  But believe me, this has often been a useful trait.

You see I’m lazy in that I hate boring repetitive tasks,  I also hate having to re-do the same thing again and again.  Once done I want it finished so I can move onto another challenge.

So to me the solution was to have a generic ‘level’ that could recurse itself.  Each level then has a type associated with it, and a parentId.

In this way a ‘level’ can be anything you want.  A building, a floor, an area of ground, a room, a swing – whatever you wan and link it in a parent child hierarchy accordingly.  We can then define WHAT it actually is within a template that is also stored as a template entity.

The iPad app and the WebUI simple interrogated the template to work out what kind of control to use, and thus as the template changed the UI just flowed around it accordingly.

Flexible Entities

So what we can now do is build HOW the app will work within the app itself, without having to rebuild either of the UI’s or the backend.  We also get to keep everything in a central location and report across ALL different survey types.

This is nothing new.  There are LOTS of apps that do this, but its surprising how often this design pattern can be used.

The Business MUST come first

Now, this of course caused quite a heated ‘discussion’ between myself and the original developer.  He (quite rightly) pointed out that such a method would complicate the code more, making reporting a bit more difficult and increase the amount of required processing power.

However, although this is all true, the fact is, as developers we shouldn’t care about such things, it is our job (or at least should be) to bring the best possible value to the business – and if that means we have a harder time of it then so be it.

Now don’t get me wrong – I’m not suggesting we should embrace creating horribly complex code – quite the opposite – if you also apply SOLID patterns to your solutions then you can easily split up this complexity into nice manageable chunks.

And as for the processing power, well not really.  Even our 2nd generation iPads quite happily reconstructed the models according to a template on the fly – and we had some BIG templates with LOTS of data.

Reporting caused a few problems initially, but again Microsoft provide all the tools in the form of Analysis and Tabular services to help out when things got too hairy.

But let’s get back on track.  Eventually I won the discussion 🙂

The resultant solution now means new clients can be accommodate within DAYS, not months.  This has a huge impact on costs and allows smaller project to make use of our system.  It also gives the business confidence that it can bid for new work knowing that if they win (and by being able to do the job faster and therefore cheaper, they often win), we can turn around the new system for them within a few days and let them get on with the job as efficiently as possible.

Because lets not forget – software is about adding more value to the business, either with cost savings or a better proposition.