Let people do less

The tendency to specialization made implementers to stop thinking about what they do at work and why. A programmer should code, an analyst should analyze, a tester should test. And someone else will think for them.

Thinking is easier than doing, and there are always more ideas than opportunities.

Even if you carefully select tasks from the backlog every time, sometimes something completely unimportant slips through.

This "unimportant" can turn out to be difficult, taking a long time to complete. Such tasks quickly spend an acceptable amount of investment. No, of course, nobody calculates the acceptable investment size beforehand. It's just a feeling "It seems to me, it is taking too long". Status checks, meetings, and motivation begin.

To compensate for the dissatisfaction, an implementer starts working on multiple tasks simultaneously. The average time to complete a task increases, the team slows down.

Because everything has become slower, customer dissatisfaction is growing. New parallel work is added. Parallelism affects the focus of implementers and the quality of the product. The speed becomes even lower. The waiting time for starting new task becomes longer.

To improve quality, tasks begin to be filled with more detailed requirements for the result. The quality may improve slightly. Well, or stay the same. In any case, the higher the quality, the lower the speed.

There will be no happy ending with this dynamic.

How to fix it

There is no way to check every hypothesis through actual technical implementation. There are always more ideas than time to test them. You can easily throw away 90% of hypotheses, leaving only the worthwhile ones, just by trying to estimate their importance in money.

Of course, not all ideas can directly impact profits. It is difficult to understand how much money the automation of collecting a financial report would bring. But you can easily imagine how much it would cost to hire people to collect this report manually or the potential fine from regulators for delaying submission.

In the end, we don't need to accurately calculate the value of each task. We just need to identify the complete crap and throw it away. And find the obvious gems and focus on them.

It might not be clear, but when someone tries to do two non-standard tasks simultaneously, they won't get twice as many results in the same timeframe. At best, it will be exactly as much as it would be if the tasks were done one after the other. The person will still think about just one task at a time, constantly switching attention.

We can discuss narrowing down the flow and WIP-limits in more detail another time.

How to filter

Wrong:

We need to create an automatic 2XG report for the regulator that can be downloaded from the admin panel. Deadline: 2 weeks.

Right:

The 2XG report is currently prepared manually by an analyst, taking around 10 days to complete. Sometimes the regulator requires the report to be submitted on Fridays, with a deadline of one week. The fine for the delay is $ 1000. We have already been late 2 times in the last six months and lost $ 2000. The next fine will cost us $ 5000. We don't want to pay fines and spend so much of our analyst's time. The next regulatory check is expected in June, so the deadline is May 31. If you unable to meet the deadline, we will pay a fine and do everything as before.

Wrong:

This will bring us millions in a year

or

If we don't do anything, we will go bankrupt

Right:

Each fine costs us $ 100. We get fines every month. We will lose $ 1200 a year

or

I have no idea how much we can earn on this, but I bet that sales for the first quarter will grow by 10%. If that is true, now we are losing $ 10,000 a day without this product. If my prediction does not come true, I will be much more careful with my forecasts in the future.

Wrong:

We can't sacrifice any functionality. Everything is very important for the release.

Right:

For now, let's skip that feature of sending a referral link through Telegram. Also we must not do such a complecated fraud prevention system. First of all, we should do a launch. Then if we see a significant response then we create new tasks to increase conversion or to decrease risk

Throw away everything that will definitely not pay off, take a few of the most important ideas and put them at the top of the list. Leave the rest untouched and don't waste your time on them for now.

Conclusion

Our goal is to minimize the scope as much as possible and do the bare minimum in small parts. Then we begin doing more different tasks because the waiting time to start a new project is reduced. We can make more experiments. The cost of mistakes will be small.

Of course, using such a brutal filtering we can occasionally throw away some pearls. I mean those projects that didn't look perspective, and we flushed them down the toilet, but in a parallel universe, they could have been successful. But there is no way to predict that.

I believe it's better to make a predictable average product built on weighted decisions (with rare brilliant experimental ideas) rather than spending all the company's money trying to do everything that seems interesting in the hope of finding a pearl at the first dive.