Every Metric Has a Hostage

Wells Fargo once ran its retail bank on a growth target simple enough to print on a coffee mug: eight products per customer. Leadership called it "Eight is Great," and the bank chased it hard enough that the target hardened into a quota, the quota into daily pressure, and the pressure into as many as 3.5 million accounts customers never asked for.

The tactics read like satire of a growth team gone feral. Employees enrolled elderly customers in online banking when those customers had no internet access. Some put their own contact details on applications so the real customer never received a single notice. Every one of those accounts counted toward the target, and for years the dashboard glowed.

The bill arrived in February 2020: $3 billion to settle criminal and civil investigations, plus a $17.5 million fine and a lifetime banking ban for the former CEO. The SEC's piece of the settlement included a line that works as a thesis for this whole post. It said the bank's cross-sell metric was inflated by accounts that were "unused, unneeded or unauthorized."

Wells Fargo counted accounts and yet, nobody counted whether anyone used them.

Swap "accounts" for "signups" and you get a slide most growth leaders have presented at least once. The stakes run smaller in software, but the mechanism stays identical: one metric gets all the attention, the metrics around it quietly pay the cost, and the bill shows up after the bonus. Sustainable growth comes from keeping every metric in balance, and design holds the lever with the widest reach across all of them.

Dashboards report in silos, businesses fail in systems

Most companies hand each stage of growth to a team and each team a number. Marketing owns signups. Product owns activation. Lifecycle owns retention. Finance owns revenue. Each team chases its number with real rigor, every team reports green at the quarterly review, and the company still misses plan.

Economists named this pattern decades ago. Goodhart's law, in its popular phrasing, says that once a measure becomes a target, it stops working as a measure. The number keeps climbing. It just stops telling you the truth about the business underneath it.

Wells Fargo had the data to see its problem early. Nobody owned the question of what the cross-sell number cost, so the question never came up in the meetings where the number got celebrated.

Metrics move each other, often from a distance

Three forces connect the numbers on your dashboard, and single-metric focus ignores all of them.

Correlation. Some metrics rise and fall together, which makes it easy to credit the wrong one. Signups and revenue both climb in the same quarter, so the acquisition team claims the revenue, even when a pricing change did the work.

Transitivity. Acquisition never touches retention directly. It touches activation, and activation touches retention. Pour low-intent users into the top of the funnel and activation dips first. Retention cohorts sour a step later, on a dashboard the acquisition team never opens. CAC payback stretches, LTV shrinks, and nobody connects cause to effect because two other teams sit between them.

Time lag. The pushed metric reacts in days or weeks. The damage reports months later. By then the team has banked the win, planning has locked the next cycle, and the lagging number reads like bad luck. Wells Fargo's lag ran for over a decade: the DOJ found that the head of the community bank knew about fraudulent sales tactics as early as 2005, and the scandal broke publicly in September 2016.

Ruthless focus on one metric works like a payday loan. The money arrives now, and you learn the interest rate later.

Striking a Balanced Approach

Balance needs a structure, or it dissolves into a feeling someone has in a planning meeting. I use a two-by-two. The top row holds goals and the bottom row holds the measures that track them. The left column pushes a quantitative goal forward. The right column protects what the business refuses to give up while it pushes.

A few rules keep the grid honest:

• Every Push goal gets a Protect goal. An empty Protect cell tells you the team hasn't priced the cost of winning. The prompt that fills it: what would make this win embarrassing?

• The Protect measure can stay qualitative. Interview themes, support ticket language, and review sentiment all count. So does a quantitative proxy like refund rate, unsubscribes, or detractor share. Either works, as long as the team sets the tripwire before the work starts, while nobody has a stake in the result.

• The grid rotates when it needs to. The Protect column plays defense by default, but a brand team might push perception while protecting conversion. The structure holds either way: one goal pulls, the other anchors.

• Stack the grids. One team's Protect metric often shows up as another team's Push metric. Line the grids up across teams and you get a readable map of the transitive relationships above, small enough for an executive team to work through in a single meeting.

Our example, tested

Run Wells Fargo through the grid and the failure jumps out: a fully specified Push column beside an empty Protect column.

Every signal in that Protect column existed. Unused accounts sat in the bank's own records. Customers complained. The Los Angeles Times reported on the bank's pressure-cooker sales culture in 2013, three years before the scandal broke. Nobody owned those signals in the same room as the cross-sell number, so they never got a chance to compete with it.

A fair reader will object here: Wells Fargo committed fraud, and your company would never forge a customer's signature. Probably true. Nobody at Wells Fargo set out to commit fraud either. Leadership set a Push goal with no Protect goal, tied careers to it, and let thousands of people under pressure fill the gap on their own. The mechanism scales down to any team with a quota and no guardrail. Your version will look tamer: a cancel flow buried four screens deep, a free trial that quietly converts, a referral program that pays out on self-referrals.

Here's the same grid for a software acquisition goal:

Fill in only the left column and you've rebuilt Wells Fargo in miniature: one person, forty accounts, zero logins, and a signup chart pointing up and to the right.

Design, the lever with the widest reach

Growth conversations default to spend levers: ad budget, sales headcount, discounts. Each one moves a single stage of the funnel, and each one costs more the harder you pull it.

Design shapes every surface where a growth metric gets produced. Signup, onboarding, the upgrade prompt, the renewal email, the cancel flow: a design team touches all of them, often in the same sprint. No other function sits at every handoff in the funnel, which puts design in the one seat that can watch a Push metric in one stage borrow from a Protect metric in another.

Design also runs the instruments the Protect column depends on. Research, usability testing, journey mapping, and the habit of reading customer verbatims already live with design teams. Yet most growth reviews invite design in to execute experiments, and few hand design ownership of the guardrails. That arrangement parks the most system-aware function in the building on the ticket queue.

Go back to Wells Fargo one more time. Employees routed account notices to their own contact details so customers never saw them. A single design guardrail, a confirmation each customer approves from their own device before any new account goes live, would have mitigated the entire fiasco. That call belongs to product design, and it likely costs a fraction of one quarter's ad spend.

How I'd implement it

A framework gets most value in the planning cycle, so here's how I roll this out with a leadership team.

1. Audit the dashboards. List every team's current Push metric on one page. Most companies find KPI targets that nobody ever reconciled against each other.

2. Map the chain. Run one cross-functional session with growth, product, marketing, customer success, and design to draw the links between those metrics. Design facilitates, because design already maps journeys for a living.

3. Fill the Protect column. Every Push goal gets a Protect goal and a measure. An empty cell blocks that goal from entering the plan.

4. Set tripwires up front. Agree on the Protect movement that pauses or reverses an experiment, and agree before launch, while nobody has a stake in the result.

5. Instrument the qualitative side. Give it a cadence: [a fixed number of customer interviews per cycle], support tickets tagged by theme, review sentiment tracked month over month.

6. Review them together. Push and Protect share one slide in one meeting, every time. Separate meetings recreate the silos.

7. Assign ownership. The design leader reports on the Protect column.

8. Run a retro after one quarter. Ask which tripwires fired and what they saved. That number becomes the business case for keeping the system.

For design leaders: earning the seat

If you lead a design team and want into these conversations, the grid hands you an opening. Use it well.

• Bring the map. Show up with step two already drafted. A finished artifact turns a request for a seat into an offer of value.

• Speak Push first. Growth leaders start listening once you show how a Protect signal predicts a Push number they already care about. Lead with their metric, then show yours moving ahead of it.

• Pilot small. Pick one team and one cycle, and let the retro carry the argument for you.

• Volunteer for the tripwires. Owning a number earns you a voice in setting it. Owning the alarm earns you a voice in every launch.

The Monday move

Growth leaders: pull up your current growth OKR, fill in the Protect column, and invite your design leader to own it.

Design leaders: draft the metric map for one team and bring it to their next planning session.

Wells Fargo hit its number for years. The U.S. Attorney's summary of what that cost: the bank "traded its hard-earned reputation for short-term profits." I'd bet that every growth team makes that trade eventually unless someone in the room owns the other side of the ledger.