This month’s T-SQL Tuesday invitation is to write about a project that went horribly wrong. And hoo-boy, do I have a project that comes to mind.
My biggest project was my biggest failure
The biggest project I ever worked on, that I ever led, was also my biggest failure. In many ways it has defined the early trajectory of my career and shaped what I value in my new career.
About 8 months into my previous job, my boss wanted us to move to a different ERP system. This would allow us to consolidate 5-6 different pieces of software into a single piece of software. The theoretical business benefits would be quite significant. No longer would we need to have this awkward patchwork of integrations between different line-of-business applications. And while we eventually did it, I consider it the biggest personal failure of my career. So what went wrong?
The software was built in-house
So the first issue was that the software was built in-house by another company in the same industry. Imagine, for example, if a large bakery had created an ERP system and another large bakery wanted to move to that system. Sounds great, right? Well, you run into two issues in that scenario.
First, a bakery is not an independent software vendor. Programming, by definition, is not their core competency. Which means that you may run into fragility or issues that you wouldn’t run into with a commercial piece of software. It also means that there isn’t going to be any documentation on migrating to the software or implementing it. Why would there be. If you built software for one company, why would you create scaffolding to move other companies onto it?
Second, not every business is the same. A lot of the fundamentals are the same, but you will run into many edge cases. We do invoices this way. They do workorders this way. We handle purchase orders this way. They handle inventory that way.
The way that I think about it is like a sea shell. It’s this intricate curve that’s grown over time, organically, to fit that creature. If you just try to fit a different snail or mollusk in that shell, it may not work out.
I was not qualified
This was my second real job ever and I had been working less than 2 years in the industry. I always tell people the jobs said .net / SQL developer, and it turned out to be a lot more SQL and a lot more DBA.
When I started that job, I didn’t know what a stored procedure was. I didn’t know what a view was. 8 months later, when this project started, I had never done any large scale integrations or migrations. At that point I had done some small integrations between pairs of systems, but nothing nearly at this scale.
Add on top of that the need to staff up and add someone to the team. So, about year in to my new job I was also now a manager.
The software stack was older and different
Because this software had started way back in the Apple II days and grown over time, much of the technology stack was quite old. The application was built on VB6, MySQL, Classic ASP, Crystal Reports 9 and Adobe Flex. This presented challenges in getting it up and running, as well migrating to it.
Migrating data from SQL server is a giant pain. MySQL workbench has an import wizard, which works decently well, but it can be a bit of tedious process. Later I was able to set up linked servers, but that involved looking up strange property settings and forcing char padding on MySQL.
The software made a lot of assumptions
Because it was homegrown, the software made a lot of assumptions about the data. Project numbers under 1000 were reserved for certain pay codes. Half of the columns would crash the software if they had null values. A lot of the tables had to be populated for the application to even run.
So, so, so much of the process was modify a view, migrate the data, see what blows up, repeat. This is a big part of what made the weekly data load so painful. So much of the migration work was sheer trial and error.
Migrating ERP systems requires understanding 2 businesses
Migrating from one ERP system to another in this case require understanding the business that created the software and the business intending to use the software. This is a lot of learning and a lot of weird edge cases. I think I deeply underestimated the sheer complexity of running a business and all the departments therein.
When we deployed, there was a whole chunk of functionality we hadn’t implemented because “certificate of work” meant one thing to use and something totally different to the other company. They whole system updated maintenance cycles based on these certificates, whereas we had assumed they we merely customer deliverables and thus optional.
What would I have done differently?
Ultimately the project took 13 months instead of the estimated 8, and even then that was only because I decided for us to have a hard cutoff at the calendar year. So, if I could travel back in time, what would I have told my younger self?
- Tighten the feedback loop. Migrating the data from one system to another was a largely manual process and quite painful, so we only did it maybe a week. Near the end I had started automating the process. In retrospect, I should have made it a nightly process.
- Learn SSIS and BIML. So much of the pain came from the slow turnaround cycle. I suspect that if I had learned SSIS, I could have made things into a daily or even hourly migration process and saved so much time.
- Read Rapid Development. I wish I had a better idea of what I needed to know from the outset. Rapid Development was a revolutionary book for me and had so much good advice in it.
- Identify a measurable list of use cases. When we made the cut-over, a number of basic things did not work because we hadn’t tested them. What would have been smart would be to have a checklist that we could run through before hand to test the use cases.
- Don’t take it personally. I was young and I treated things outside of my control as my responsibility. I was sore for a long time about how things went, and I thought too much on how that reflected on me and my abilities.
So now what?
Even to this day, I’ve got a certain amount of skittishness around the idea of large projects. There is a whole suite of soft-skills, of methodologies involved in making sure an IT project is successful, and I’ve learned the hard way what happens when you don’t have those skills.
For now, I’m excited to focus on course authoring and projects where I just need to create content.
Finally, I’ll leave you with a favorite quote of mine:
Good judgement comes from experience. Experience comes from bad judgement.
6 thoughts on “T-SQL Tuesday #107: Death March”
Rapid Development looks like a good book. We have a scheduling problem at my place of work, I’ll add this to my list.
I really agree with that quote! There are things I do differently now that I’ve failed and burned a few times. #1 is creating automated checks that verify my ETLs are operating as expected. Not process completed/failed, but that there’s one row in the destination for every row in the source, etc.
Comments are closed.