Goals and expectations

This is the central move of the whole model, and it is one distinction:

A goal is what someone is reaching for. An expectation is the measurable version of it.

Goals are never measured directly. Expectations are the only directly measurable thing here.

Everything else on this tier is arranged to make that separation usable.

Goal — the latent, emotional deliverable

A goal belongs to a persona, and it is what they are ultimately reaching for. Deliberately emotional:

to be secure · to be respected · to be free of a worry · to be proud of the work

Not "to complete onboarding in under five minutes". That is a proxy, and a proxy is the other concept.

Goals are latent: real, load-bearing, and not directly observable. You cannot instrument to feel in control of my finances. You can only find things that stand in for it — and the honesty of the model comes from admitting that rather than quietly redefining the goal as the thing you happened to be able to measure.

A goal is born held by exactly one persona. If two personas seem to hold the same goal, look again: usually they are reaching for subtly different things, and collapsing them loses the difference that mattered.

Expectation — the measurable proxy

An expectation belongs to a stakeholder — a persona situated in a context — and it proxies a goal:

to not be asked for documents I don't have to hand to see exactly where the money went to get an answer the same day

Concrete. Situated. Something you could check.

The chain reads: a persona holds a goal; that persona, situated in a context, is a stakeholder; that stakeholder holds an expectation; the expectation proxies the goal.

flowchart LR
  P["Persona<br/>small-business owner"] -->|holds| G["Goal<br/>to feel in control"]
  P --> S["Stakeholder<br/>owner, in onboarding"]
  OC["Operating context<br/>onboarding"] --> S
  S -->|holds| E["Expectation<br/>not asked for documents<br/>I don't have to hand"]
  E -.proxies.-> G

Why the indirection earns its keep

It would be simpler to record only the measurable things. Three reasons not to.

A proxy can be satisfied while the goal is missed. Onboarding completion goes up; the owner still does not feel in control, because you got them through the form by hiding the thing they were anxious about. With only the metric, that is invisible and looks like success. With both, the gap is a question somebody can ask.

One goal takes several proxies, and they change. To feel in control is served in onboarding, in billing, and in support, by three different promises. Replace a proxy — a better measure, a redesigned flow — and the goal is untouched. Fuse them and every measurement change looks like a change in what the customer wants.

It makes an honest failure mode visible. A goal with no expectation is one you have named and are not yet serving anywhere. That is worth knowing, and it is only expressible because the two are separate nodes.

The test for a good expectation

Could somebody disagree about whether it was met?

If yes, it is not yet an expectation — keep going. If the disagreement is only about degree, you are there.

The left column is a goal wearing an expectation's clothes. The right column can be checked, argued about, and measured.

When a goal has no measurable version

Sometimes you genuinely cannot find a proxy. That is a finding, not a failure — and it is one of the more valuable things this model surfaces.

It means one of three things, each worth knowing:

1. the goal is real and you are not serving it anywhere yet;
2. you are serving it, and have no idea whether it is working;
3. it is not actually a goal — it is a value, and values do not get proxies.

Recording the goal without an expectation keeps the question open where somebody will see it. Inventing a measurable-sounding expectation to fill the slot closes it with a false answer, which is strictly worse than the gap.

Before you build, ask whether it can be judged

An expectation that nothing measures cannot be validated, however well the work goes. You can ask that in advance rather than discovering it after delivery:

What measures this expectation? Can this ever be judged?

Asking before the work is much cheaper than learning afterwards that nothing was ever going to tell you whether it worked. See Metrics.

Working practices

• Write goals in the customer's emotional language, expectations in checkable ones. If your goal sounds like a KPI, it is an expectation and the real goal is still missing.
• One goal, several expectations, across contexts. That fan-out is the model working, not duplication.
• Do not promote a metric to a goal. "Increase activation by 10%" is your target, not what your customer is reaching for. The customer's goal is underneath it, and it is the thing that survives a strategy change.
• Let goals be few. A persona reaching for fifteen things is a persona nobody has thought hard about.

Read next: Strategy and initiatives.