In November Seth Godin published one of his very short blog posts under the title “One swell foop”, a play on the phrase “in one fell swoop”. He means the temptation to tackle something once and hard, to pay the money, grit your teeth, launch with verve and then move on, and his verdict on it fits into four words: “Foops aren’t swell. Foops fail.” What works instead, according to Godin, is a process, “an iterative, evolving, persistent path toward better”.

Almost everyone in almost any company would agree with that, and still it is worth opening the project status report, because it contains a curious sameness. The onboarding that is now running in its sixth version and the values campaign that nobody has touched since its launch sit in the project overview as identical rows: phase completed, status green, progress 100 percent.

Metrics are what I call the signal layer of action infrastructure, the place where a company tells its people every day which behavior was right. In September I described here, using the defensive dashboard, why these signals almost always reward defense. The row with the 100 percent shows a property that was still missing there, and it explains why organizations love the one big push even though nobody would defend it.

Where the dashboard stops looking

An ordinary project portfolio dashboard shows the number of projects in the portfolio, the completion rate with the cancellation rate next to it, schedule adherence at go-live, budget variance and go-lives per month against plan. Below that sits a project overview with phase, traffic light and progress bar. Each of these numbers makes sense while an initiative is running, and each ends at the same point, the day the project is closed and handed over to the line organization. Whether the new onboarding works after that, whether anyone improves it or whether it quietly decays, no longer appears on any screen, because to the metric it is no longer a project.

I call this property event logic, and it is the sixth alongside the five described in the defensive dashboard: metrics count events, meaning starts, go-lives and completions, but not cycles on the same object. How large the gap behind it is can be seen in PMI’s Pulse of the Profession series. In 2015 only one in five organizations reported high maturity in benefits realization, meaning following up after completion on whether a project actually delivers the promised benefit, and in 2017 the figure was 31 percent.

Why the big push looks so reasonable

Reading Godin, you might think organizations simply have not yet recognized the big push as a mistake, yet for the person responsible for it, it is a rather reasonable choice. A launch is visible, it has a date and it can be attributed to a person, while the fifth revision stays invisible and belongs to nobody. Budgets are approved as projects with a kick-off and an end date, and for the maintenance afterwards there is usually no cost center. The result is a pattern many organizations will recognize: the failed relaunch of a loyalty program is followed by the next relaunch, and the training day is followed by a Monday that nobody is responsible for anymore.

Godin’s sentence has a limit, by the way, which he does not mention himself, and Sweden demonstrated it in 1967. On 3 September the country switched overnight from left-hand to right-hand traffic: anyone on the road had to stop at 4:50 a.m., change sides and was allowed to drive on at 5:00 a.m., and on the following Monday there were 125 reported accidents, against 130 to 198 on previous Mondays. A gradual switch would have been deadly here, because every intermediate state is more dangerous than both ends. That gives a usable rule: iteration pays wherever intermediate states are viable, and only what has no viable intermediate state belongs in one stroke. With an onboarding, a loyalty program or a leadership culture every intermediate version is viable, and that is precisely where most organizations, unfortunately, decide not by this rule but by what their dashboard counts.

A dashboard that stops counting at go-live records the launch of the ship and forgets the voyage.

Progress bars that permit slacking off

Even where a dashboard shows progress, it often works against carrying on, and behavioral research knows three well-documented mechanisms for this. Ayelet Fishbach and Ravi Dhar showed in 2005 that visible progress invites slacking off as soon as people read it as “already done”, and only keeps them going when they read it as evidence of how much the goal matters to them. Ran Kivetz, Oleg Urminsky and Yuhuang Zheng found in 2006, using a café’s stamp cards, that customers came more often the closer the free coffee got, and after redeeming it their pace dropped back. Hengchen Dai, Katherine Milkman and Jason Riis described the fresh start effect in 2014: around temporal landmarks such as the start of a week, the first of a month or New Year, people take on something new remarkably often.

A project dashboard showing goal attainment in percent serves all three mechanisms at once, because it invites the final sprint before go-live, then reports “achieved”, permits the pause and offers the next fresh start as a new bar at zero. That is the big push in installments, and in the report it looks like diligence.

Five signals that count coming back

Anyone who wants to change this starts not with the screen but with the unit of counting, and the right unit is the completed cycle on the same object, not the start, not the hour and not the distance to the goal. The simplest signal for this is the version number, because “Onboarding v6” tells a different story from “six onboarding initiatives”, even though the same work may lie behind both, and a fresh start then appears as what it is, a v1. The second signal shows the difference from the last version instead of a level, most powerfully as a time-lapse with the version from 90 days ago next to today’s, and because there is no finish line, there is also no point at which the final sprint tips into the pause.

The third signal gives weight to every open beginning, so that a portfolio becomes visibly heavier with every initiative that was started and neither ended nor developed further. This needs a second exit next to “completed”, namely “deliberately ended”, which is essentially the kill rate from the defensive dashboard read from the individual’s point of view. An initiative that someone ends with reasons is also a completion, quietly leaving it lying is not, and it is precisely the silence that makes scattering one’s efforts so cheap. The fourth signal is a survival curve of one’s own initiatives, showing how many initiatives of a quarter are still alive after six months and in which version. Almost every company knows the churn curve of its customers, yet hardly any knows this curve for its own projects.

The fifth signal reports the rhythm instead of a deadline, meaning that the fixed working session on a matter took place, not that a result was achieved. In the Handover Rule this means creating a receiver, a fixed frame with a fixed trigger to which coming back is handed over. A streak is fragile too, though, because Janet Polivy and Peter Herman described in dieting research how a single slip tips into “what the hell”, and so a rhythm signal needs a return rule that counts coming back after a gap more highly than the unbroken streak.

Five signals that count iteration

1 Version number: What counts is the version of the same object, a fresh start is v1.

2 Difference from the last version: The comparison is with one’s own earlier version, not with a target.

3 Weight of open beginnings: Every initiative without a decision weighs, “deliberately ended” counts as a completion.

4 Survival curve: How many initiatives of a quarter are still alive after six months, and in which version?

5 Rhythm with a return rule: What is reported is the working session held, and coming back after a gap counts more than the streak.

How differently the same data look when shown with these signals becomes clear in the following example, a fictional portfolio of 14 initiatives, once in the usual layout and once rebuilt.

Project portfolio Q3
Status report · as of 30 Sep
Before
Projects in portfolio
14
4 new this quarter
Project completion rate
71 %10 of 14
target 70 %
Cancellation rate
14 %2 of 14
target below 10 %
Schedule adherence at go-live
80 %
on plan
Project overview
ProjectPhaseStatusProgress
Onboarding for new hiresCompletedgreen100 %
Values campaignCompletedgreen100 %
Mentoring programCompletedgreen100 %
Self-service portal pilotCancelledred40 %
Highlight from the status report
Launch of the quarter: values campaign, went live on schedule
Initiatives by version
As of 30 Sep · reviewed quarterly
After
Survival curve after six months
4of 10
completed initiatives are being developed further
Open ends
6
no decision yet: continue or deliberately end
Deliberately ended
2
with a documented insight, count as completions
Rhythm
11of 12
working sessions held, 2 returns after a gap
Version ladder
InitiativeVersionEffect since the last version
Onboardingv6first own customer request after 11 instead of 19 days
Mentoring programv43× no effect decision due
Self-service portal pilotv2deliberately ended: existing customers do not use self-service
Values campaignv1unchanged for seven months, open end
Applause this quarter
Version of the quarter: Onboarding v6, with the time-lapse of what has changed since v3

Both views show the same fictional portfolio with the same underlying data. All figures are example values.

What the before dashboard trains

The completion rate reports ten completed projects, and nothing in it shows that six of them are lying idle today. The only red rows are the two initiatives someone ended with reasons, and the one being celebrated is the campaign nobody has touched since its launch.

What the after dashboard trains

The same ten completions split into four living initiatives and six open ends that demand a decision. The stopped pilot counts as a completion with an insight, and the applause goes to the sixth version of the onboarding instead of the launch.

The disease of process signals

Process signals have a disease of their own, and from the outside it looks exactly like persistence. A version number can be counted up without anything improving, and as early as 1976 Barry Staw showed, under the title “Knee-Deep in the Big Muddy”, that people tend to put more money rather than less into a course they chose themselves after it produces bad results. Every iteration signal therefore needs an effect question per cycle that cannot be answered with diligence: what has changed since the last version for the people the initiative was built for? If the answer comes back empty three times in a row, “deliberately ended” is the right exit, and in the example this is exactly what happens to a mentoring program that keeps every working session and still receives the note “3× no effect”.

Applause for the sixth version

The strongest signal appears on no screen, because it is the stage on which an organization decides whom it celebrates. When the fourth version of something gets the applause at the all-hands meeting instead of its launch, everybody is watching, and no dashboard teaches as quickly what counts as achievement in this organization. Often the object itself carries the signal too, a document that visibly grows thicker, or a customer who notices the change and names it, because someone who knows that the next version will be laid next to the last one works differently from someone who only has a deadline.

Whoever is responsible for a portfolio keeps, in this picture, the logbook of a ship, and a logbook records every voyage, while the launch of the ship fills no more than the first page. Most project dashboards are still registers of launches, and so what remains at the end is a question that every portfolio could answer within a week and that hardly any has ever asked: how many of the initiatives that jumped to 100 percent last year are still alive today, and in which version?