Measurable strategy in e-commerce: why goals without numbers aren't goals

Almost every shop we look into has numbers. Shopware reports, analytics, ad accounts, a spreadsheet for the monthly close, often a BI tool that was connected at some point. What almost none of them have is an answer to the question which goal those numbers are supposed to prove.

That is not a tooling problem. It is a strategy problem disguised as a tooling problem. A strategy nobody translates into metrics is not a basis for decisions — it is an opinion with a slide layout. And a dashboard without a goal behind it is not an instrument of control, it is a rear-view mirror.

This article is the entry point to a series. It deliberately stays general: what makes a strategy in e-commerce clear and measurable, and where it falls apart in practice. The details — KPI definitions, attribution, visibility in AI answers, personas, personalization, assortment steering — follow one at a time in the next articles.

Where measurable strategies break down

Measurable strategies rarely fail for lack of will. They fail on five recurring patterns: goals without a number, inconsistent metric definitions, measures with no link to a goal, a missing planned trajectory, and impact that is claimed rather than measured. They show up regardless of company size or industry — and they reinforce each other.

Goals are adjectives instead of numbers. "More revenue," "better visibility," "become more profitable," "a stronger brand." Those are directions, not goals. They can neither be reached nor missed — and therefore cannot be steered. A goal needs a number, a date, and a name that owns it.

Every metric has three definitions. Ask marketing, sales, and finance for last month's revenue and you get three answers: gross, net, after returns, with or without shipping, by order date or by invoice date. The same applies to conversion rate, CAC, and contribution margin. As long as a KPI is not a stored formula over named data sources, every meeting negotiates the number first and the decision second.

Measures have no link to a goal. The backlog is full, the team ships, the roadmap gets worked through. Yet for a large share of the activity, nobody can say which goal it feeds. What is not assigned cannot be prioritized — it gets built because someone asked loudly enough.

There are actuals, but no planned trajectory. The year-end target is set, the current value is known — but nobody knows what should have been reached by today. So the gap only becomes visible in Q4, when no measure has time left to take effect.

Impact is claimed, not measured. A measure gets an expectation ("+8% checkout conversion"), goes live, and after that nobody follows up. Without a before/after and without a control group, every impact claim stays a story. In the next planning cycle it gets passed on as experience.

How far expectation and impact can drift apart is illustrated by an example Ron Kohavi and Stefan Thomke documented in Harvard Business Review, from Microsoft's Bing team: a single change to the way ad headlines were displayed raised revenue by 12% in testing — described in the article as the best revenue-generating idea in Bing's history. Without measurement it would have stayed one opinion among many.

What makes a strategy measurable

A strategy becomes measurable when four layers interlock cleanly. In our F.O.C.U.S. framework this is the Objectives phase — clarity before speed, impact instead of features.

1. A guiding statement that rules decisions out

Strategy is mostly about what you give up. A useful guiding statement doesn't only say where you are going; it makes visible what you will not do any more: which audience you won't serve, which channel you won't run, which part of the assortment you won't expand. If a guiding statement doesn't eliminate a single idea, it's a marketing slogan.

It also needs versioning. A statement that quietly changed three times in two years says more about a company than any report — but only if those changes are traceable, with dates and reasons.

2. Goals with a number, a date, and an owner

Every goal needs a target value, a date, an owner, and a baseline. Without a baseline, progress cannot be calculated; without an owner, nobody follows up. Whether you use OKRs or a strategy map in the tradition of the Balanced Scorecard matters less than this: the structure has to work in both directions — top down ("which metrics carry this goal?") and bottom up ("which goal does this metric feed?").

3. KPIs as formulas over defined sources

A KPI is not a term, it is a calculation rule: which fields from which system, over which period, with which boundaries. It needs a target value and, where available, a benchmark — because a conversion rate of 2.1% is neither good nor bad without a reference point.

Once the formula exists, the uncomfortable question can be answered too: is the data even there? Many metrics that show up in strategy documents cannot be calculated from the systems that are actually connected. Knowing that early is worth more than a nice target number.

4. Measures with expected and measured impact

Every measure carries a hypothesis, a measurement window, and an owner — and after go-live, a measured value next to the expected one. That produces the only statistic that genuinely helps in planning: how well does your team hit its own expectations? A team that lands at 40% of expected impact plans differently from one that lands at 110%.

What that looks like operationally is described on our page on growth and scaling: few metrics, clear guardrails, short cycles with before/after.

The planned trajectory matters more than the target

The most common steering mistake is knowing only the end value. "€1.2M revenue by year end" works as a goal but is useless as a steering signal — until October, everything looks like interpretation.

It only becomes useful with three lines: plan (what should be reached at any point in time), actual (what is reached), and forecast (where the current trajectory ends if nothing changes). Only the distance between plan and actual as of today turns a goal into a decision: add resources, re-prioritize, or deliberately correct the goal.

Seasonality belongs in there as well. A linear plan is wrong in almost every assortment — anyone who leaves the Christmas quarter, a season start, or B2B budget cycles out of the trajectory creates panic in spring and false confidence in autumn.

From extrapolation to probability

A forecast that merely extends the curve so far answers the wrong question. What matters is not where the line ends, but how likely it is that the goal will be reached by its date — as a number, per goal, with a named model and a chosen horizon behind it.

The difference is practical, not academic. A trend line invites debate ("December will make up for it"). A probability sorts: which goals need a decision this month, and which will get there on their own? That requires the model and the horizon to stay visible and selectable — a forecast whose assumptions you cannot inspect is just another opinion, this time with a decimal place.

And it is only as good as the data beneath it. With thin history, after an assortment change, or in a season with no prior-year comparison, a forecast is an indication, not a commitment. Which is exactly why the probability belongs next to the planned trajectory rather than in its place: the plan says what should be, the actual says what is — and the forecast says how urgent the gap between them has become.

Measurable means one data model, not five reports

This is where the strategy question becomes a technical one. As long as orders sit in the shop, sessions in the analytics tool, cost in the ad account, and visibility in Search Console, nobody can calculate across them. And the interesting metrics live exactly at those boundaries:

• Revenue per organic click instead of clicks and revenue side by side
• Contribution margin per channel instead of revenue-based ROAS
• Return rate per campaign instead of conversion per campaign
• Repeat purchase rate per entry source instead of new customers per channel

That requires orders, sessions, clicks, and cost to be joined on shared keys — customer, product, campaign. And it requires a decision about the attribution model that is made deliberately and documented, instead of being inherited from a tool's default. Which model you choose matters less than everyone using the same one.

The same applies to new visibility surfaces. How often your shop appears as a source in AI answers is now a metric with a revenue connection — and it shows up in no standard report. We cover that topic under Generative Engine Optimization for Shopware and in our article on AI sales channels.

Goal conflicts belong on the table

Any serious goal structure produces conflicts, and it should. Two goals that are both correct pull in opposite directions — and without a decision about which metric wins in case of doubt, the outcome gets decided by chance. The four most common conflicts in e-commerce:

• Revenue growth against contribution margin
• Conversion optimization against return rate
• Assortment breadth against data quality and inventory cost
• New customer growth against customer acquisition cost

If these conflicts are not named, they still get decided — by whichever team happens to ship first. A measurable strategy makes them explicit and defines which metric wins in case of doubt. That is exactly the difference between guardrails and good intentions.

Alignment is a metric in itself

Probably the most uncomfortable number of all: what share of your ongoing activity is assigned to a strategic goal? Alignment does not measure how well individual measures perform; it measures how much work pays into the stated strategy at all. It is the only metric that puts a number on the distance between the strategy document and the day-to-day.

We have asked this question in many projects, and the answer rarely sits above two thirds. That this is not a local problem is borne out by research from Donald Sull, Rebecca Homkes and Charles Sull in Harvard Business Review: two-thirds to three-quarters of large organizations struggle with executing their strategy — not with formulating it. The rest isn't inherently wrong — mandatory work, compliance, operations, and maintenance have to happen, as we described in Shopware maintenance and ongoing development. But a third of unassigned work is something you should know rather than suspect.

The same logic applies to budget: if you cannot say how much money flows into which goal and what contribution margin it returns, you are steering marketing and development budgets by feel.

What we're building: nuonic Lens

We have assembled this structure in projects over and over — in spreadsheets, Notion boards, dashboards, and meeting notes — and every time it started to decay after about three months. Not because the logic was wrong, but because holding goals, metrics, data sources, and measures together was manual work.

So we are turning it into a product: nuonic Lens, a commerce BI that brings strategy and shop data into one system. Not as another dashboard, but with the goal structure at the center: a guiding statement and goals with a planned trajectory, KPIs as traceable formulas over connected sources, measures with expected and measured impact — and underneath, a shared data model built from shop, analytics, ads, Search Console, and AI visibility.

Every goal there carries a third line next to plan and actual: the forecast, along with the probability of reaching the goal by its date. Model and horizon are selectable, the probability also appears in the goal overview next to status and owner — and the suggestions that follow from it each come with an expected impact they have to be measured against later.

The next articles in this series go through the building blocks one at a time:

• KPI definitions nobody negotiates any more
• Forecasting that produces probabilities instead of extended trend lines
• Channels and attribution, including contribution margin instead of plain ROAS
• Visibility in classic search and in AI answers
• Personas as a measurable assumption rather than a workshop output
• Personalization you can switch off and recalculate
• Assortment steering along margin, returns, and data quality

nuonic Lens is still in development — for now there is no public sign-up and no pricing. If you want to follow the series or talk to us about it early, just get in touch.

The next step

You don't need a new tool to apply this article. Three questions are enough to start — and if one of them stays unanswered, you have found your first task:

1. Name your three most important goals with number, date, and owner.
2. Name the metric you measure each goal with — including formula and data source.
3. Name the expected and the measured impact of your last three measures.

Wherever that gets stuck, it is almost always the data situation, not the willingness. That is where we start: in consulting we build the goal structure and the KPI definitions, in growth we run it as a cycle. If you want to talk it through, get in touch — a conversation with the people who also build it.

Sources and further information

• Harvard Business Review: Why Strategy Execution Unravels — and What to Do About It
• Harvard Business Review: The Surprising Power of Online Experiments
• Harvard Business Review: The Balanced Scorecard — Measures that Drive Performance
• What Matters (John Doerr): OKR — definition and examples
• Google Ads Help: About attribution models
• Google Analytics Help: About key events
• Search Console Help: What are impressions, position, and clicks?

Further reading

• F.O.C.U.S. — the methodology behind our projects
• Growth: scaling is a system
• Consulting: decisions on a data basis
• Generative Engine Optimization (GEO) for Shopware
• Shopware maintenance: retainer and ongoing development