Project pitfalls

This is a list of common pitfalls we have seen when running customer projects. It’s designed to help us learn from the past, so we don’t repeat it. By its nature, this is what goes wrong, not what you can do. See other pages for what you should actually do.

Things on this page should not be identifiable to projects. In fact, each item on the page should an invented mix of multiple projects, so that there is even less chance of any one point identifying something .

Who is waiting for who?

The customer thinks they are waiting for us to do something, but we think we are waiting for the customer to do something. This can easily happen when someone asks a question and it isn’t noticed, but can also happen at a larger scale (“I didn’t know we made the final decision that we wanted this.”)

We should make sure we know who is the “prime mover” at all times, to know who will drive things forward.

Doing what the customer asks

The customer comes with some specific request, and we think we should do that. It may a specific idea, or the customer may not know exact details but still have some specific goal in mind.

This is called “managing up” or “managing the customer”. We aren’t consultants who want to do whatever and take their money. We are collaborators who have expertise that is different from the customers, and we need to work together. The best engineer projects come from the customer needs and engineering both adapting to each other. We are the professionals in projects, so we need to take the lead in helping customers figure out what to do.

“AI”-generated design or starting example

Customers come with a project idea, but it is machine generated. Thus, it may not be fully thought out, thus discussions about the project go in circles. Similarly, they may have some initial code, but it’s mostly machine generated and thus isn’t a good starting point.

We should recognize and ask about these situations, and know we need to take step back before diving in. It may very well mean we need to spend a lot of time on initial understanding, before thinking about what they are actually asking us to do.

Not getting frequent enough feedback

We work on a project for too long without sharing the progress with the customers. When they do see, it wasn’t quite what they expected.

This is common for many types of projects. There needs to be proactive showing the customer what is happening, making demos, getting it into their hands. If customers don’t have time to follow our progress, this is a major red flag.

Doing what the customer asks

Sometimes, the customer needs X, thinks Y can do it, so asks for Y. In reality Y wasn’t the best way to get X done, and we waste lots of time (this is the definition of the XY problem. In research, this can be especially common, since the problem is often under development as much as the solution.

Spend an appropriate amount of time understanding the big picture of the customer’s ultimate need, not just what they ask for. If they don’t have time for this, consider if we should do the project. For the biggest projects, a few online meetings isn’t enough. Go visit them, join their group meetings, etc.

Focusing only on the technical task

Customers come and we focus on what they ask. However, this fits into a bigger research group, with different people with different skills. “Who can do what?”, “How do people feel about computers?”, “How do people relate to the work?”, “Who is able to use and maintain the result?”, and so on, are affect the technical outcome.

Like above, one needs to spend enough time getting to know your customers to make sure that the technological solution is appropriate to them.

Goal drift

A project starts off with one idea, which may or may not be accomplished, but the goal drifts over time. The project goes on in a reactive mode, possibly without a direction or defined end point. The new goals may not be possible, a good idea, or have enough staff to accomplish. We may be acting more like researchers, investigating random things, hoping something comes out.

We are researchers and if the project goal involves testing out various things and we are paid to do so, this isn’t necessarily bad. Still, it good to identify these places and make sure our actions are aligned with everyone’s expectations, and there are no “planning pitfalls” surfacing.

No specific end goal was defined

The customer came and it sounded like a good idea. We got started, but it went on, and on, and on. Eventually, we realized we could never end, since there was never an end point. For anything we do, there can be a follow up that is part of the project.

Like above, we are researchers and this is our operating environment. If we are paid to experiment like this, it’s OK that we do it. If not (especially for free projects), this is a bigger problem. I’d recommend being explicit: is this something with a goal, or is the goal some experimentation? (Yes, experimenting for some fixed time can be a goal). Structure projects in phases, for example “phase 1 is understanding your existing work (the current goal). Then we decide the next phase, with a goal decided then.”

Taking forever analyzing options

Often times, the best solution isn’t clear. We like analyzing things, and like getting things right. We think forever about things, essentially going in circles re-analyzing the options. Everything is a trade-off and some decision needs to be made.

I think it’s worth recognizing these cases, and there is no perfect answer. Consider: who should make the decision about the trade-offs? Then, can you summarize the trade-offs (in writing) clearly enough that everyone can understand, comment, and then support the decider in making the choice? This hopefully limits the time taken to clarify options, discussion/understand, then making the least-bad decision.

Inconsistent names of projects

Sometimes we and/or customers don’t invent a consistent, representative name of a project. Then, we get confused about what project we are talking about. Gitlab issue numbers help, but we don’t use them in daily conversation.

It’s worth asking the name early on and making sure it is representative and distinct. Note that the first three words are sometimes used as a short placeholder in reports, so put the most distinct parts at the front (for example “Digital map creation webserver”, not “Webserver to make a digital map”).