In Data We Trust: building a performance practice that owned its own measurement

The agency I wanted to hire and couldn’t find

By 2010 the digital agency market had sorted itself into two unsatisfying options. There were creative shops that made beautiful work and could not tell you what it did. And there were media shops that could tell you what happened and had no idea why.

Both were built around campaigns. A client would buy a campaign, the campaign would run, a report would arrive, and the fundamental questions — is this business growing, which of these initiatives deserves next year’s money, what would we do if we had to double the customer base — were nobody’s job.

Prove was built for those questions. The tagline was on the cover of every deck we ever sent: In Data We Trust.

What we actually did differently

We owned the measurement layer

The single most consequential technical decision was moving clients onto first-party ad serving while most of the industry was still renting third-party infrastructure. This was years before Safari’s ITP and the broader cookie collapse made it a boardroom topic.

The reasoning was unglamorous. Third-party cookies were being rejected and deleted at rates that made conversion data quietly, systematically wrong — and wrong in a direction that flattered nobody. Attribution windows were missing conversions. Retargeting pools were undercounted. Frequency capping didn’t cap. Every optimisation decision downstream inherited the error.

Serving from first-party context fixed the data at the source. It also meant that when the industry-wide reckoning arrived, our clients had already migrated.

We started with a financial model, not a media plan

Many engagements opened with a model of the client’s value chain rather than a channel recommendation. What is a customer worth, over what horizon, at what margin, and therefore what can you afford to pay for one? Only then: which channels can deliver at that price?

This sounds obvious. It was not common practice, and it caught things. On a 2018 engagement with a $135M regional retailer, building the model surfaced that the business had been planning against an in-store average purchase value of $1,796. The real figure was $1,360 — a 24% gap that had been quietly propagating through every forecast in the building. Correcting it re-scoped the entire media plan.

We built roadmaps instead of selling campaigns

Where other agencies are oriented around "campaigns" we are a complete strategic partner, developing roadmaps for our clients to help prioritize the many initiatives that will scale up their businesses, including campaigns.

— Prove capabilities deck, 2017

The framework we used for market share was deliberately layered, bottom to top: a Foundation Layer of analytics, keyword-level call tracking, SEO, paid search and retargeting; an Engagement Layer of landing page optimisation, gated and ungated content, email capture and drip programmes; an Expansion Layer of conquesting, digital media and social; a Leadership Layer of blog outreach and partnerships; and at the top, a matrix-based analytics strategy tying it together.

The point of the layering was sequencing. Clients almost always wanted to start at the Expansion Layer, because that is the layer that feels like marketing. Starting there without the Foundation Layer is how companies spend two million dollars and cannot tell you what it bought.

Little bets

Our execution philosophy had four words on the slide — Launch Fast. Test & Optimize. Learn Fast. Evolve. — and one phrase underneath: little bets, predictable payoff. Ramp up, maximise learnings, minimise risk, incorporate successes. We did not do big-bang launches, and we did not let clients do them either.

Who we worked with

Across Prove and its predecessor Placements Media, the client list ran through enterprise healthcare and medical devices — Medtronic, Abbott Diagnostics, MannKind, Cedars-Sinai, Providence Health & Services, Natera — into energy with Southern California Edison, hospitality and gaming with MGM Mirage, and a long tail of consumer and industrial brands including General Electric, Qualcomm, Siemens, Thomson Reuters, Fidelity, Hyundai, Peet’s Coffee, Proactiv, Arctic Cat, Monoprice, Rosland Capital, Sit ‘n Sleep and USC. Some of that work was direct; some came through agency partnerships, where we ran performance as a white-label practice for firms including Ignited and Wongdoody.

The Medtronic relationship is the one I point to: five years, followed by a four-year renewal signed for 2018 through 2022. Enterprise healthcare marketing organisations do not re-sign an agency for a decade because the decks are pretty.

The team

Prove ran with a bench of directors rather than account managers — engineering, analytics, creative, media, client services, marketing automation and digital production each led by someone who had done the work. The 2017 capabilities deck put it at more than eighty combined years of B2B and B2C digital customer experience, analytics, and acquisition and retention expertise.

We also held SOC 2 Type 1 certification from 2014, which mattered more than it sounds — it was what allowed the healthcare and financial services work to happen at all.

Why I wound it down

Agencies have a structural problem: the model rewards holding onto accounts, and the best thing you can do for many clients is make them need you less. By 2019 I was more interested in building the products than in marketing other people’s, which is what led to World Back to Work, Wanderhome and eventually Diagnostic.ly.

But the operating grammar of Prove is still how Shor Consulting works. Model the economics first. Own the measurement. Sequence the layers. Make little bets.