XX.III. Funnel

The 3rd process architecture: a process with rapidly diminishing number of items conveyed from step to step
Dr. Albert Künstler

funnel is basically a chain, a multi-step process, with high loss rates on each step. The funnel usually receives a large number of items from different sources. They enter the funnel and are being handled, but their number significantly diminishes from step to step. The majority of them drop out in the midst of the process and only a tiny fraction makes it to the end.

Examples: sales funnel, recruiting process, attendance of boring lectures over the course, farming, and the gathering and delivery of perishable fruits.

Research landscape of the Funnel

There are three main research spaces to be explored: Steps, Segments and Channels.


The funnel may be observed from two different viewpoints:

  • Server: The point of view of the process owner, the server, who collects the incoming items and drives them through the funnel.
  • Visitor: The point of view of the item that joins the funnel and is served or processed by a server.

From the Server point of view, the steps are usually quite easy to identify and disaggregate. In a quest, the sponsor most likely owns the process so your crew can collect information from the business customer about the steps. They should have the statistics and know the main vulnerabilities of the process. Some steps may have a very high loss rate, and others not. For simplicity, it may be reasonable to merge such a low-churn step with the previous or the next step in order to reduce the number of steps in total.

Example: The sales funnel from the point of view of the sales manager is making calls, pitching the product, sending the invoice, receiving the payment, shipping the product, and making sure the customer is satisfied with their purchase.

The Visitor point of view is harder to research but it is more valuable. It gives the understanding of the item lifecycle and better explains why they leave the funnel. It is worth creating a map and exploring the funnel from the item’s point of view. Some steps, the touchpoints of the funnel, will be quite the same as the steps of the Server’s process. However, some other hidden steps or states may be discovered in their journey. If these points are decisive for the successful completion through the funnel, they must be explored and must give clues for the reengineering of the current process.

Example: The sales funnel from the point of view of the customer (customer journey) is understanding they need a product, talking with friends who already bought the product, collecting information about the products and features, talking with potential sellers, comparing the offers, making the decision to buy, confirming the purchase with the sales manager, making payment, receiving the delivery, unpacking and trying the product, and deciding whether they like it or should they return the product and claim the money-back.


Another terra incognita of the research landscape is the segmentation of items. Usually, items are not exactly the same. Some are fragile, others resilient, and some are highly relevant while others accidentally popped in. If you understand those differences and are able to classify items accordingly, you can change the process so that the relevant items are better served and handled while the irrelevant items are sorted out earlier in the process.

Example: On citrus farms, harvested lemons should be carefully sorted after picking. The fruit may belong to different segments: damaged, infected, fully ripe or immature. It is important to identify and sort out damaged and infected lemons. If you let them into the storage, they will spoil other fruits around and increase your losses. Ripe and unripe fruits should also be processed differently. The first of them should be sent for the fastest sale in nearby markets, and the others can be stored for longer or sent abroad.


Finally, it is also recommended to create a map of channels that feed the funnel with the new items. Usually, these sources are not the same. They bring items with various properties and representation of relevant segments. By mapping and exploring these channels, you may suggest ideas for improving the overall throughput of the funnel.

Example: In marketing, it is not necessarily the channel that generates the most leads, generates the most sales. To optimize costs, the goal is to invest in channels that bring in relevant leads at a reasonable cost, even if there are less expensive but irrelevant sources.

Problem analysis

The main three problems in the funnel usually are:

  • Conversion/loss rate at each step of the funnel and retention of the relevant items as they go through the steps
  • The cost of the new items and optimization of the sources that feed the funnel
  • Effectiveness and cost of the processing on each step as in any other chain

To improve the conversion rate, it is usually recommended to start reengineering the end of the funnel — the last steps of the process. The reason is simple: in the last steps you lose the most relevant items. They have come a long way from the very first steps and proven their quality in several stages of sorting, and finally, you have made the highest individual investment in these items, since each touchpoint has cost you something.

On the contrary, a low conversion rate in the first steps may be acceptable. The first steps are usually sorting/qualification steps. What if we sort out even more irrelevant items? Here is the paradox — you lose more items at the initial steps, but down the chain you have a small number of relevant items. By means of rigorous sorting, you reduce the operating cost. You spend less effort on pushing the irrelevant items down the funnel and can pay more attention to the relevant ones.

For better sorting, your research may be aimed at understanding the lifecycle of items, their journey from the Visitor point of view. Apply item segmentation to identify differences between items and the stages of their lifecycle. Identify the distinctive attributes and behavioral patterns that are signs of a relevant item that will not drop out of the funnel until the end.

Low conversion at the initial steps may also designate the wrong sourcesthat feed the funnel. Each of these channels have certain capacity, relevance, and cost per item. These properties are not static. Over time, some channels can be developed, the others inevitably dry up, reducing their capacity and generating less items. The cost of items may also go up. By doing your research over channels, your crew may identify new undiscovered channelswith high potential. You may find the ways to maintain and broaden existing sources, predict inevitable decline of channels that were plentiful in the past, and to distribute resources and budgets across channels.

Common flaws and illnesses of a funnel:

  • Low conversion/retention rate at the step
  • Expensive sources of new items
  • High percentage of irrelevant items
  • Expensive or slow processing at the step
  • Too much unfinished work in progress

Leveling guide — Funnel.

Leveling guide — Funnel

If you’re designing and building a process from scratch:

Create Placeholders of all steps of the funnel and one of the most standard and capable channels.

Upgrade all steps of the funnel to the Quick & Dirty level (Low Fidelity, Medium Quality). Conversion rates may be far from ideal yet, but it is important to make sure some items successfully reach the end of the funnel. If some steps are blocking the path and all the items drop out, it is important both to improve this step and/or increase the number of items that enter the funnel. So one or two channels must get funded and leveled up to the Quick & Dirty level. Don’t try to optimize the funnel and the channels yet. Just make sure the funnel operates, and some items successfully travel from the first to the last step.

If you’re improving an existing process, skip phases 1 and 2 and go to the next phase.

Target discovery: start from the end of the funnel and identify the main leaks — the steps with maximum loss/waste/churn/abandonment.

Explore the potential improvements of the identified leaks at the end of the funnel: create several experimental Quick & Dirty alternatives and upgrade leaking steps to a Good Enough level.

Segment the items and understand the difference between relevantand irrelevant ones. For the relevant items review the funnel from the Visitor point of view. Add hidden steps in the funnel if necessary.

Explore whether the first qualification step of the funnel can be upgraded to Good Enough level so that more of the irrelevant items are sorted out, and the relevant ones are processed with attention and care.

Review the existing channels, their current and achievable capacity, relevance, and cost. Rebalance the resources spent to maintain and develop these channels, drain the worst and upgrade the best. Try to discover new channels with high potential.

Continue identifying leaks in the funnel as well as experimenting with the channels. Upgrade the steps to Good Enough and the most important ones to State of the Art level.

Always evaluate not just improvement of the individual nodes, but also the performance of the funnel as a whole.