Bumper

Context

Bumper is the vehicle history brand in the portfolio. It had traffic and it had a backlog, and what it did not have was an argument about which order to do things in.

That is the ordinary condition of an organic programme without a roadmap: everything on the list is defensible, nothing on the list is sequenced, and the work that ships is whichever item somebody argued for most recently. The problem to solve was not finding opportunities. It was committing to an order and holding it long enough to see whether the order was right.

Owned

• The six-month roadmap: themes, epics, and tickets, written in Jira down to acceptance criteria.
• The sequencing argument, which is the part of a roadmap that actually carries the risk.
• The measurement design, so progress could be read monthly rather than argued.
• The B2B fleet lead generation pipeline for Wingmate as a secondary track, using sales intelligence (Apollo) to serve enriched data to the sales team.

Execution was the team's. I wrote the specification and they built it, which is the division that matters for reading the outcome below.

Architecture

The roadmap was three levels. Themes named the bets, epics broke each bet into shippable units, and tickets carried acceptance criteria specific enough that an engineer did not have to reconstruct the intent.

Sequencing logic ran on two rules. Template and indexation work came before content volume, because publishing into a broken template multiplies the defect instead of the traffic. And inside each theme, the work that could be measured soonest went first, so a bet that was wrong would say so early rather than at the end of the quarter.

Progress was measured as tracked keyword count on a monthly snapshot, chosen deliberately over sessions. Sessions move with seasonality and with paid activity elsewhere in the portfolio, and attributing them to a roadmap is an argument. A keyword count on a fixed monthly cadence is a duller instrument and a harder one to talk yourself into misreading.

Decisions

• Specify to acceptance criteria, not to intent. A roadmap that says what to build gets built. One that says what to achieve gets negotiated.
• Templates and indexation before volume. Slower to show a number, and the reason the volume work compounded when it arrived.
• Keyword count over sessions as the progress measure. The less flattering instrument, and the one that could not be explained away.
• Sequenced so wrong bets surface early. Ordering by time-to-signal rather than by expected size.
• B2B pipeline kept as a separate track. Different buyer, different funnel, and folding it into the organic roadmap would have muddied both.

Outcome

Tracked keywords grew 147% (11,710 to 28,989), Jan to Jun 2026^{Source note 1}.

The roadmap executed on schedule and held after I left.

Monthly organic keywords tracked, source note 1

Month	Keywords tracked
Jan 2026	11,710
Feb 2026	13,315
Mar 2026	14,301
Apr 2026	20,322
May 2026	28,935
Jun 2026	28,989
Jul 2026	27,759
Aug 2026	28,228

Two things in that series are worth more than the headline. The step from April to May is where the sequencing argument either was right or was not, and it was: the volume work landed on templates that had already been fixed, which is why it moved that far in one month rather than grinding. And July and August sit level with June, months after I had gone, which is the only real test of whether a roadmap was a plan or a personality.

I wrote the specification. The team executed it. The keywords grew over that window and stayed grown once neither of us was pushing.

Revision

I would have set the monthly snapshot up before the roadmap started rather than alongside it. The January figure is the first snapshot rather than a settled baseline, so the earliest part of the series carries more uncertainty than the rest, and a quarter of prior history would have made the starting point arguable by someone other than me. I would also have written an explicit kill criterion for each theme, because sequencing to surface wrong bets early only pays if there is a number agreed in advance that ends one.