Get best of the week

Product and prioritization: how to evaluate project objectives?

Sometimes in a team there are disagreements on a product: what tasks are now in priority. To prevent a wall of misunderstanding between developers and managers, the product must prioritize. And how to do it right - read the article.

Experienced products know that you can’t just poke a finger into the sky and among all burning tasks you need to single out the most significant ones. Misha Karpov, ex-Product Director Skyeng, through one of the studies, found that Russian and foreign companies break down prioritization into two stages: quick and slow assessment.

image

First up is a quick assessment that cuts off irrelevant tasks. After that, the product conducts a slow detailed assessment.

Quick assessment


image

For a quick evaluation, the products are armed with an Agile approach that takes into account the benefits of the task and the labor costs of the team:

  • First, the product manager and technical experts discuss how useful the new feature is, for example, the use of filters in the Metro application. The team votes: show 1-3 fingers - and the product records the average value. This happens with all features.
  • The following discusses how difficult it is to implement the update. As well as useful, the voting results are recorded in the column “Average Cost Estimation”.
  • We correlate the benefits and costs. The table in the figure above shows that Feature 2 and Feature 3 are very leaders - it means that these two updates should be launched in the near future. In the second column we see that they are easy to make, and in the first, that these features will be useful to users.

Labor costs are discussed with the technical team.

, , — , .



: , Intercom:


image

  • ;
  • (Reach) — , ;
  • (Impact) — , ;
  • (Confidence) — , «»;
  • (Effort) — . RICE score: Reach, Impact Confidence Effort.

Impact in these cases is determined by voting, or by user feedback.

Another quick assessment option is the “Metrics Hierarchy” using the VKontakte example:

image

  • According to the top-level metric, they consider how often users interact with features.
  • The next step is to use the service metric - the time of viewing the video. The product is discussing with the team what affects the video viewing time: duration and quantity. These indicators relate to lvl 1.
  • At the level of lvl 2, we analyze what affects each indicator in particular. The duration is affected by the percentage of searches and the duration of the video itself. Percentage of viewing is affected by the quality of the content and the speed of the video.
  • We decompose each item into levels: we get a layer cake from the main metric and the levels that affect it. At this point, the product thinks it could improve performance (for example, add video in HD). But before you implement the changes, you need to understand what metric this will affect. In this case, the HD format will affect the quality of the video. The product examines its tree and finds the level at which it posted Video Quality.

The closer to the main metric you need to pump, the more useful it will be. If the feature is at a distant level, then there is almost no chance to affect the global metric.

To measure how important the quality of the content is, the product builds a tree that analysts check. After that, developers have tasks for six months to a year in advance. They are gradually modifying the product.


In the picture below, “Hierarchy of metrics” in excel format for monthly or quarterly analysis:

image

  • the first column shows the metrics that need to be raised using new features;
  • in the second column - the expected digitized indicators;
  • in the third column features that are needed;
  • in the fourth column, the weight of features (from one to three).

What does a product face when working with wood?


  • It happens that there are 15 features in the table, and the remaining 180 are waiting in line;
  • When a tree is compiled, ideas for features are taken not only from the table, but also from the backlog, as team members can advise something important;
  • "Weight" is determined by the analysis of past projects and features;
  • If one project affects two metrics at once, it is defined in that metric, on which it will have more influence.

Slow rating


For example, in Skyeng there is a quest for students, where they are invited to complete various tasks in several classes. The quest begins a couple of classes before the second payment for training, and in order to complete it, the user will have to pay the next month of training on the platform.

You need to find out how much money the quest will bring and in what priority this task should be.

We compile a calculator with answers to the main questions (those items in which the product doubts are highlighted in yellow in the table):

image

  • How many users in 12 months will be at this decision point? For example, 100,000 people.
  • How many users will try the quest? The team suggests that this could be 44%.
  • How many users will complete the whole quest? Perhaps 81%.
  • How many more users will make a second payment after implementing quest lessons? Presumably, re-payments will be 4% more.
  • What profit does the second payment bring? The team knows for sure that 5000 rubles.
  • What additional profit will the company receive in 12 months from the launch date? We count the number of participants and multiply by the re-payment.

Why are the numbers highlighted in yellow like this: 40%, not 60% or 15%?


To answer, we consider probabilities for unknown coefficients (point 2 in the picture).

  1. Pessimistic, realistic and optimistic scenarios for those who try WOW lessons are considered. We substitute a figure for each scenario that meets our expectations.
  2. In the “Scenario Probability” column, we determine what is the probability that the scenarios will occur and consider the average value. The same with the rest of the unknown.
  3. Estimate the cost of development.
  4. We consider the ratio of the development cost and the money received (for the development cost, the product clarifies with the team how many hours they spent and multiplies by their bid).

After the feature is released, you need to look at the forecast for large releases and the ratio of the come true forecasts to the profit for the month.

For third-party people, it may seem that the product is sprayed and, when prioritizing, is wasting time.

We need to calculate how much the company will cost the product error. Amount can be in the millions.


Which calculator uses Skyeng to calculate data?


This is the Google Doc to which data is connected (current value, current LTV, etc.). Information falls into the cell of the document, and then all calculators refer to a specific cell. An advanced calculator is used when you need to clarify more details.

Brief Prioritization Algorithm


  1. Find the TOP 3 key metrics for a particular service.
  2. Collect hypotheses for pumping these metrics: from the backlog or outside it.
  3. If the market is new - use quality methods: ask potential users what they are currently using.
  4. Perform a quick assessment and discard the "weak" features.
  5. Perform a detailed assessment of the remaining features.

Skyeng use one fast and one slow methods and pre-determine the timing of prioritization. For example, for developers on a weekly planning meeting, the team discusses how much time each person will need to solve various problems. They provide the product with an estimate of the terms, after which he decides whether to take on the feature. In prioritizing the latter, they use the ROI metric to show the ratio of the money that the company will receive to the money that will be spent on development. It happens that the ROI can be more than 1000-3000%, and the development of features will take very little time.

Competent prioritization!

More articles:


All Articles