UX Refactoring. Up from the top. Start changes from the last stage.

Change the last stage — it is the most isolated, there are fewer risks and the most targeted audience. It is easier to smooth out the negative changes and improve the view of the process.

Toward the end — fewer risk, the impact of errors and easier to correct

Processes are a series of related stages. Whether it is the process of buying goods on the site, using a mobile application by a client, or using a calendar with reminders.

Russian version available here

At each stage, you can lose the user, he can get confused in the interface, not find the product or action he needs, or leave.

Only some get the final goals. If the conversion reached 3–4%, then consider yourself lucky.

Of course, there are many methods and techniques on the market on how to try to increase the conversion of visitors into buyers. One thing unites them all — if you start from any stage except the last one, then these changes are similar to attempts to carry water with a sieve, pouring more water. Although, first of all, you need to make fewer holes. It would be even better — initially to take a bucket or a bowl.

It is known from the theory of probability that reaching an event at a given funnel stage is a result of multiplication of events probabilities at all previous stages.

Conversion in this context — is the probability of passing each stage, isolated from the effects of others. For example, a wife found goods in an online store, collected them in a basket and sent it to you for ordering and further payment. Here are how many such men will make an order registration — is the conversion of this stage.

If you change the process at the beginning, the changes affect the entire audience that works with the process. Redesign of the main page will be seen by everyone. Changes in the layout grid for goods on the site — only those who went to use the search. A new payment page will be appreciated by rare users who already want to buy something.

And what will happen if an error crept into the change? If the change is incorrect, it affects all events after it. If the error is blocking, you will cut off all further actions on the funnel and all users in this stage will face this.

But if the change was at the last stage, then there is nothing else to influence. This minimizes the impact of errors on the system. And there are only hundredths of the original users. Your shame will not be so visible (minimizing reputational costs).

Two aspects are important — the effect of the error on the system in the subsequent stages (especially if there is no isolation, see below), plus the effect of the error on the user. Offset changes to the last stage minimize both of these moments.

Connectedness or Isolation? Nash equilibrium or Pareto efficiency?

In game theory, there is such a thing called Nash equilibrium — a state (or set of strategies) in which no participant can increase the gain by changing his strategy if other participants do not change their strategies. If we consider the individual modules/stages of work in the system — we just run into it.

To better imagine this concept, look at the picture. Like alpinists in the mountains are interconnected by a rope so that no one gets off the path, so the elements and stages of each system are interconnected logically, by requests and actions of users. We can give one of the alpinists more comfortable and warm clothes, change the stuff of the backpack, but we cannot give a jetpack or take him to the top by a helicopter, as he is connected with the other participants and goes as a part of the entire expedition.

And how to make them more efficient so that they are more likely to reach the top? Practice group an extra month. Or hire a Sherpa porter who will take part of the load. Or, suppose fantastic things, equip each member of the expedition with an exoskeleton. These changes apply to the whole group. You can apply them one by one, or you can apply it all together. Separately, for any one of our five — there will be no sense. Our climbers are strongly interconnected. Both figuratively and physically (with a rope).

The concept of Connectedness is widely used in some industries. For example, in linguistics, programming, or math. Please do not confuse with the concept of Uniform Connectedness in the design, which came from Gestalt psychology.

Connectedness is when in order to change one interface element, a stage on a site or a stage of a business process, it is necessary to change other parts associated with it.

To move to the next equilibrium state, you will need to simultaneously change several parts of a system, site, or application. For example, you want to sell bundles (sets of goods):

• How to calculate the price and show in the search?
• How to display these products in the cart?
• How to return one item from the set and how to take this into account on the site side?

Yes, all of this can be thought out and implemented, but the bottom line is that each subsequent change in strongly related process stages is an additional effort to change the links of an even greater number of parts of our system.

If we consider the usual redesign as a re-arrangement of the related elements of the system, then performing it, we always break the connectedness.

Let’s now take the two strongest, most trained climbers from our five and tie them together with one rope, and the other three with the other. We broke the general connection and we got a system of two isolated modules (3 and 2 people). Now, the two leaders can go ahead, drive in the hooks, pave the path, make an extra trip back and forth and raise a couple of backpacks from the weakest comrades to the next stop. Please note that all five expedition people continue to go to the top together. But more efficiently.

The concept of Isolation is known in programming and when working with databases. With the same success, it can be used in the analysis of systems to separate the variable parts.

Isolation is the reverse side of connectedness.

In the example with sets of goods, enter isolation through a number of assumptions and restrictions: two products in the set have one description — a photo, a single price and could be sold/returned only together. Then there will be a change only at the stage of the formation of this item, while in other stages the set will be displayed and follow the logic as an independent product. A set of dishes, a set of furniture, a set of tools — you come across these examples every day.

By the way, isolating the elements of the system from each other will help a lot with induced errors. When in one place we repair and breaks it in another. The influence of errors on the system (both before the stage where there was a change and after) is minimized by isolation. As a rule, when developing a new functional or some other changes, a technical task or something else is written, taking into the design the input data (previous stages). But the next stages, mistakes, exceptions in logic and undesirable events are considered in full only by few.

If we consider the system, but only in those aspects that are isolated from other parts, we can proceed to Pareto efficiency (this is from economics, not from game theory). Applying the rule that:

“Any change that does not cause damage to anyone, but benefits some people (by their own assessment), is an improvement.” (c) V. Pareto

.. we can improve parts of our system or funnel stages without affecting others until we reach “Pareto optimality” — when we can’t improve anything without deteriorating other elements or stages. It is isolation that allows UX refactoring to be carried out — to eliminate defects and disadvantages of the interface, to simplify it, without changing the rest of the logic.

Yes, it may well be a situation that you will not be able to achieve the same results as with the complex change of several adjacent systems at once. But the costs of improving one isolated element will be much less (by the way, in a connected system, maintaining actual connections with changes also requires costs). And the ratio of the benefits to the cost of changing the isolated part is much better than with the classical change of a strongly connected system. Therefore, given the above:

Isolate the last stages of the process and change them without touching others.

Changes from the last stage is a cost path from the least to the most. Even if you do not pass it completely, but easily cope with the initial stages.

Where are the funnel borders?

Have you ever wondered where the funnel of user interaction really begins and ends? Entering the main page is not always the first contact — this can be the result of switching from contextual advertising or some other marketing activity. Therefore, often the “first stage” in fact is not the first but is in the middle. And changing it, we must pay attention to the stages before and after.

Don Norman define UX like this: “User experience covers all aspects of the end-user interaction with the company, its products, and services.”

And the last stage is what? Checkout, payment or delivery? Or an outgoing support call with the question “Did you like everything?”. Or maybe it is, on the contrary, the incoming call of the user with a complaint about the purchased goods or a second appeal to the site.

It is difficult to find the first funnel stage — who knows to what marketing activity the user reacted. Therefore, it is easier to determine the most important, goal action where the contact with the user ends.

The last stage, when performing the goal action of the user, will be the most isolated. There is nothing behind him.

Suppose it is a payment card on the site. Proper internet-acquiring is an unaffordable luxury for most online stores. Therefore, they use the services of third-party companies (or banks with acquiring).

But for them, a form of payment on your site is the first stage in their funnel.

Payment platform makes pages for your site and gives it back to you. Two funnels, one connected to another. But your most important action, where the user makes the payment, is actually not in your funnel.

Many acquiring services make it possible to make a custom frame in a client’s design — this is much better. When paying with a card, going to an external site (even with a green search bar and an SSL certificate) is still a risk in the eyes of users. And even more so if the style, design, logic of elements (input fields and drop-down lists) differ. This stage for the user turns out to cut off according to the logic of interaction from the previous ones. He will receive cognitive dissonance: in his head, there is an expectation that the order and payment are on one site, but in reality, he has just been on a page with one kind of design and has already found himself in a completely unfamiliar design. And cognitive dissonance is bad, you can scare and lose a client.

Strong connectedness is wrong, but weak is also wrong. A balance between connectedness and isolation is needed.

The Gordian knot of isolation and connectedness must be untied, not cut.

Subsequent business units in the chain often break connectedness. Those who support it — has achieved staggering success. A striking example is Stripe.com, which gives more flexibility and is ready to change to the needs of the client in a wide range.

Everyone’s favorite Uber is also an example of the wonderful binding of passenger and taxi driver funnels. Where the passenger’s funnel ends — the beginning of the funnel of the taxi driver is organically interwoven.

By outsourcing a business stage, you risk a breakdown of connectedness and incorrect isolation. That is why many large companies are building vertical holdings — businesses are independent, but there are common signs (level of service, quality, history of interaction with the client) that are inherited between the stages.

How to properly isolate and remove excessive connectedness is also a big topic. In short, the properties between the stages, each of which can be carried out by an independent business unit, may be different, but the principles of tectonics, faktura, and construction must be maintained and signs must be correctly passed on.

Sequential isolated improvements can be perfectly spread over iterations in full accordance with the Agile-approach. Risks are reduced. And yes, if you have achieved mastery in isolation, and the changes pass without error, then from a mathematically logical point of view, you can safely make changes from any end of the process. But there are still psychological aspects …

Fear of novelty versus Endowment effect

Each of us has a so-called reptilian brain. He is responsible for basic instincts: food, survival, reproduction. The Fear of novelty is precisely the fear of a potential threat to survival. And the sooner the user sees the changes when working with the site (if the user has already used your service), the faster this trigger will work in his head. That is, you voluntarily, in the first stages, risk pushing the client away from the path to his goal. Therefore, users do not like updates on their favorite sites.

And if the change is at the end? Yes, there is this risk too. But the user has already gone almost all the way and it remains to take this last stage (for example, payment for goods). He is already highly motivated because the Endowment effect has entered the game. We are more afraid of losing what we already have. And if you have already found the right product in the search, filled out the necessary forms with the address for delivery, then in your head you already mentally imagined that it is yours, how will you use it, what will relatives and friends say, etc.

By the way, sellers use this very well when they claim that there is not much left, the last stocks in stock and similar advertising tricks. Here they are involved in the principle of Scarcity based on the same imaginary possession:

“ The joy is not in experiencing a scarce commodity but in possessing it” (Robert Cialdini, “Psychology of Influence”, p.200).

A more explicit application of the Endowment effect is seen in street vendors. Often a thing is forcibly thrust into your hand so that you hold it or try it on, and then bargaining begins.

Endowment effect force the user to overcome the fear of change. Or buy a product that he held or tried on.

Peak-end rule

Except for the two effects described above, there is another cognitive bias that actively influences the user’s perception of work experience.

Peak-end rule concentrates our memories on the most intense moments of the event (at the peak) and on when they end. The last day of vacation is beautifully remembered.

There are two moments:

• changes focused on the last stage allow you to quickly improve the user’s last emotional experience. It is easier defined than the first peak, which is possible both with a positive color (on this site I found the right product!) and with a negative one (well, why did they made this update?!)

• if the change being introduced is the most negative experience when working with the process (first peak) and is at the end, then there is an opportunity to compensate it with positive emotion from the successful completion of the whole process (second peak). Yes, an update is made on the site and the user sees it immediately, but it helps him to more quickly and simply complete the process and achieve the goal. The coincidence of the two peaks in the last stage, one extinguishes the second. For example, here shown the queue in the first McDonald’s in Moscow in 1990. People stood for hours (a huge negative first peak) to buy the coveted hamburger at the end (second peak).

Shoot the Bull’s Eye

The distribution of users to achieve the goal in the process (for example, buying a product on the site) can be described by a bell-shaped curve on which part does not buy exactly (accidentally hit the site), a part will definitely buy and a part between them, which will be determined in the process.

Working with changes in the last stage is working with the target audience itself.

Change the last stage — it is the most isolated, there are fewer risks and the most targeted audience. It is easier to smooth out the negative changes and improve the view of the process.

Facebook: facebook.com/lessplusmore
Twitter: twitter.com/lessplusmore

And clap the article, if you like it!

Special thanks to Wova Roodnyy for ideas and consultations.