Service Intelligence Labs logoService Intelligence Labs
03Results

What the loop changed.

Below are case studies from our recent engagements. We partner with founder-led businesses to help them understand what their data is actually telling them, and make sure their expertise shows up where customers are now looking for it.   We believe businesses already have most of the answers, they're just scattered across systems that don't talk to each other. Our job is to pull that together, make sense of it, and turn it into a plan you can act on.

Figures from case 01

15,000+

Leads re-scored in the pipeline audit. Eighteen months of history that nobody had read end to end.

4th → 1st

Rank in brand mentions across 5 major AI platforms versus competitors across the timeline of engagement.

46%

Share of leads with no recorded lead source at all. First-touch capture now runs at intake and belongs to the client.

Data Science | AI Visibility | Measurement

01Legal services · Multi-state · 50+ employees

A founder-led multi-state law firm

Situation

Most businesses built their reputation the old way: word of mouth, good reviews, a strong website. That still matters. But more and more, people aren't asking Google anymore, they're asking AI. And AI doesn't work like a search engine. It picks a handful of answers and presents them as fact. If your business is not one of those answers, it's like being invisible to a growing share of your future customers.   We helped a national law firm address and fix this exact issue by first finding out where their money was actually coming from, and where it wasn't. Their team could tell you which cases they'd won, but couldn't tell you which leads were worth chasing because almost half of every lead that came in had no record of where it originated. At the same time, when we checked what AI platforms were telling people who asked "who's the best law firm", this firm wasn't the answer. A competitor was.

What we installed

  • COLLECT

    Two instruments where there had been none: a first-touch attribution field in the CRM, asked at intake after a lead qualifies, and a weekly sweep of 351 real buyer questions across four AI answer engines.

  • OWN

    We re-scored 15,914 deals. Close rate by lead source ran from 0.4% to 74%. The internal ranking the firm would have used to judge its own people reversed once we controlled for the mix of leads each one was handed.

  • ALIGN

    One metric system, one definition, every percentage published with its sample size and no blended averages. Brand-name search terms came out of the scorecard, because they measure whether people who already know the firm can find it.

  • DESIGN

    A scored lead queue the client's COO works daily, with the threshold tuned against 36 hand-labelled examples. Items below the threshold are still logged, so the threshold itself stays testable.

  • EXECUTE

    First signed client from a channel that had produced none. Attribution capture rolled out by the client's own COO, to his own intake team.

Lead source attribution

Empty on 46% of 15,914 deals → First-touch attribution is now assessed on every qualified call

Answer-engine share of voice

4.0% → 6.6% across 60 buyer-question prompts

New channel introduced

No attributable clients → First signed retainer from custom-built monitored queue

"This has already shown value in the few days we've been using it."

Client COO

What they keep

The attribution field lives in the client's CRM. The lead queue runs on a schedule and their COO works it. The written language set, the directory records, and the page briefs are theirs. The monitoring platform is ours and it leaves when we do, which is exactly why nothing that matters was built on top of it.

Omni-channel marketing

02Hair salon · Single location · Owner-operator

A founder-led service business, building the team

Situation

The client list existed only as a list. Nobody had counted how many of those clients had stopped coming, there was no way to reach them at scale, and platform decisions were being made on brand impression rather than on the fee math. There was no leadership team to disagree about the numbers, because there was no team yet. Everything had to be built before anything could be measured.

What we installed

  • COLLECT

    One client list where there had been a spreadsheet and a memory. 155 contacts consolidated and loaded, with the reachability of each one recorded: 19 have no email address at all and can be reached only by text. The list carries no birthdate and no address data, which rules out birthday and location campaigns before anyone plans one.

  • ALIGN

    Definitions written down once and applied everywhere. A client counts as slipping away when there is no visit on record in the current year, so the segments mean the same thing every time the list is pulled. The owner's eligibility facts were written down once as well, so every funding program gets scored against the same inputs instead of judged on reputation. The ranking is not the running order: one shortlisted item scores second but carries a processing window long enough that starting late costs a season, so it starts first.

  • DESIGN

    An eight-tab workbook that runs the funding search: live programs, an application pipeline, a watchlist, a change log, and the written procedure for working it. Anything surfaced by automated search lands in a quarantined candidates tab and reaches the working list only when a person promotes it. Where a funder's site blocks automated checks, that is recorded in the sheet and routed to the manual pass instead of failing silently. The recurring check is built to run inside the file the business owns rather than on our machines, which is the only reason it will still be running a year from now.

  • EXECUTE

    Two decisions closed on evidence instead of impression. On payments: both candidate platforms charge identical processing rates, and the storefront option adds a monthly fee plus a further surcharge on payments taken through an outside gateway, which is the gateway the booking system already runs on. The recommendation was to stay on hosted payment links until there is real product inventory, with a written trigger for reopening the question. On funding: the single most recommended program in the field was checked on the funder's own site and confirmed to have no track the owner could apply under this year. It stays on the sheet marked closed, with the reason and the date it was verified, so nobody researches it again next quarter.

  • OWN

    Not yet demonstrated, and we are not going to pretend otherwise. OWN is the phase where a movement gets a named cause and a named owner. Nothing has moved yet, so there is no cause to name. The instrumentation had to exist first. This is what the next quarter is for.

Client list

a spreadsheet nobody had counted → 155 contacts loaded by the owner, each one labelled, reachability recorded

Consent

outreach with no documented permission → an approved opt in message and a recorded opt in video, staged and not yet sent

Funding search

chased on reputation → a scored shortlist in a workbook that rescans monthly and quarantines whatever the machine finds

The owner loaded the list, not us. That was the point.

From the August status review

Next measurement

Three numbers exist to be taken and none of them exist yet: the opt in rate once the approved message actually sends, the return rate of the slipping away group measured against the labels now sitting in the list, and the first automated funding scan, scheduled for the first of the month. They get published here when they exist, and not before.

What they keep

The client list, the labels, and the consent assets live in the owner's own account under the owner's own login. The funding workbook is a file the business owns, and the recurring check is built to run inside that file rather than on our machines. The standing weekly advisory call is ours, and it ends when the engagement ends. Nothing that matters depends on it.

Next

Want numbers
like these?

Thirty minutes on your business and where the numbers stop being trustworthy. You leave with a clearer picture; we follow up in writing.

Book a discovery callTake the 90-second diagnostic