Michael A. West IIISame audience. Different message.
After rebranding to Enterprise Mobility, the company stood up an internal startup to grow its car sales side of the business. Nobody yet knew which channels were actually reaching buyers. Over nine weeks I tracked my own queue, pitched a fix to senior leadership, and then scored my own forecast against what happened.
A new audience, and no read on what reached them
I was one of ten interns at an internal startup testing whether a central team could turn the lowest-intent audience we had into interested buyers before handing them to dealership consultants.
The assigned job was outreach. The open questions were which channel reached this audience, when they responded, and whether the message landed. None of that was being recorded, so I started recording it.
I tracked every touch myself
On my own initiative I built a tracker that recorded every contact at the moment its outcome was known: which channel it came through, where in the sequence it landed, and what the customer did. Replies were time-stamped to the minute and the whole thing was reconciled against team data each week.
Every lead recorded the moment its outcome was known: channel, touch stage, result.
Inbound responses captured to the minute, so reply timing could be read against when leads arrive.
Entries verified against team data and one-on-one recaps, so the numbers held up in a room.
Where the effort went, and what it returned
Most manual hours went to the channel that converted worst, and the outreach copy was being misread.
Live conversation carried the overwhelming share of real engagements: calls were 23% of touches and 65% of outcomes. Email ran the other way, taking 21% of touches for 3% of outcomes, with 1,176 sends producing 20 replies and no appointment-level engagements. Replies clustered at the start and end of the day, 40% before 11am and 38% after 2pm, while the queue loaded heaviest in the morning. The late-stage touches in the cadence returned little for the volume they consumed.
The message was the other half of it. These are people who rented a car, not people shopping for one, and our outreach asked about their “potential vehicle needs,” wording they read as a rental follow-up. 65% of calls went unanswered, and the ones reached late had usually bought elsewhere. Taken together, the numbers pointed at targeting and copy rather than effort.
Move the effort to calls, and fix the copy
Four changes, each requiring no new headcount and no new tools.
Automate the email step
Workflow automation, no new headcount, freeing the manual email effort for live calls and text.
Standardize the site-request reply
One templated email so a customer asking for our site gets a fast, consistent answer.
Tighten the call talk path
A short script refinement that confirms market intent quickly and routes real buyers to the dealership.
Reword the outreach copy
Clearer purchase language in the email and text templates, cutting the confusion with the rental relationship.
Two of the four are copy changes. All four were written to be adopted without new budget or headcount.
I wrote down what would happen, then checked
The recommendation shipped on July 10. That made the prediction falsifiable, so I scored it against eleven weeks of actuals.
The forecast held where the recommendation was adopted and missed where it wasn't. Filling the midday trough depended on moving the first call from four hours to two, which didn't ship, and the trough stayed flat at 21%. The miss is included here because the forecast was written down in June and could be checked against the actuals in August.
What a qualified buyer actually costs
Nine weeks, 7,590 logged tasks across 2,400+ leads, priced out.
Cutting two low-yield touches is worth about twelve seats at full national coverage, roughly $9,100 a week, because the constraint is reachable leads rather than hours in the seat. Every input is labeled actual or estimate on the page it appears. The loaded-rate multiplier and my own task timings are mine, not company figures.
What I put in front of leadership
Two reports, seven weeks apart. The midpoint report made the case for the change. The final one checked it against the actuals.
Shared with permission. Independently tracked by me during the season; Enterprise Mobility's own reporting may differ.
Where this could be wrong
Included in the deck as presented.
Not a randomised test
The two cadences ran in sequence, so lead mix, volume and seasonality travel with the change. The largest caveat on the deck.
Thin tail sample
The two touches I recommended cutting total 224 attempts. Zero appointments is directional, not decisive; the steady decay across all six call touches is the stronger evidence.
One seat, two months
Every operational figure is my own queue across 40 logged days. Team numbers appear only in the economics, where they are labeled.
These caveats were included in the deck as presented, rather than raised in the room afterward.
Marketing that has to answer for itself
Channel mix, message testing, timing, and the measurement to prove any of it, done from inside the funnel rather than from a deck. The changes I pitched in June shipped on July 10. The August report closed out every one of them, priced the work at $208 of labor per car sold, and showed that cutting two low-yield touches is worth about twelve seats at national coverage.