Autodesk
Manual Add
Overview
BuildingConnected is the largest real-time construction network that connects owners and builders through an easy-to-use platform to streamline the bid and risk management process. Within BuildingConnected, my team works on Bid Board, which helps subcontractors manage their bid invites so they win the right jobs.
To make Bid Board the fastest and easiest system for tracking construction opportunities, my team shipped a V2 of Manual Add — a feature that allows subcontractors to add bids to Bid Board manually. The first release of V2 did not outperform V1 in our success metrics, so we iterated on V2 until it was successful.
My role
The designs for this project were passed to me from another designer. I was in charge of helping my team ship the designs and iterate on the release to ensure it was successful. I worked closely with my PM and epic anchor (engineer lead on the project) on the iterations.
Problem & Solution Overview
The V1 of Manual Add had several problems that made it cumbersome to add bids to Bid Board. For example, users needed to go through several screens before adding a new bid, and then they need to go back to Bid Board if they wanted to create another. V1 also had limited fields in the creation flow, so users needed to save the new manual bid before filling out other fields.
V1
V2 (final release)
Success Metrics
Before our first release of V2 (let’s call it V2.0), we defined some success metrics, including:
Primary Metric:
Creations per active office
Secondary Metrics:
Unique user conversion from viewing manual add UX to creating opportunity
Creations per creator
We also had several Health Metrics, but the most informative one ended up being:
Usage of fields in new form
As part of our Success Metrics definition, we said that all aforementioned metrics should increase.
V2.0 Results
Unfortunately, V2.0 did not outperform V1 on our primary and secondary metrics.
In our health metrics, I noticed that the client field metric was particularly alarming but informative.
User Session Recordings
After seeing these metrics, I looked at user session recordings for some qualitative data. I found several usage patterns, but in summary, out of 9 session recordings, only two users added a client successfully.
The quantitative and qualitative data informed me that we needed to make filling out the client field easier.
Releasing V2.1
For the V2.1 release, I streamlined the add client UX.
The session recordings revealed that users struggled to find their desired contact by searching, so V2.1 limited the client field to allow only email search; it did not allow search by name and phone. It also had other copy and UI improvements.
V2.0
V2.1
V2.1 Results
V2.1 improved over v2.0 in:
Unique conversion to create (+17%)
Creations per office (+3.5%)
However, V2.1 still underperformed V1.0, so we needed to improve the add client UX further.
Releasing V2.2
V2.2 required exact match on email to show a result — previous versions allowed fuzzy match and returned multiple results. This further streamlined the search UX.
V2.1
V2.2
V2.2 Results (Success!)
V2.2 outperformed V1.0 in our primary metric and one secondary metric. With these results, we rolled out V2.2 to 100% of users.
Final test
However, the other secondary metric of creations per creator was still lagging behind in V2.2. We hypothesized that that metric was still underperforming since users did not understand what happened to a manual bid after creation.
Thus, we released two more variants that tested post-creation UX made with Pendo (a tool that allows us to create in-app guides and doesn’t cost us engineering resources).
A modal that appears post-creation.
A flyout that appears post-creation. It draws attention to the toast.
Results
The Pendo post-creation UX was effective — it increased creation per creator! We rolled out that variant to all remaining users.
Learnings
I learned a lot through iterating on the Manual Add release. My main takeaway is that we changed too many things with the release of V2.0 (initiation, presentation, content, post-creation), which made it difficult to parse the impact. In the future, it would be beneficial to slowly release changes, which we did in our release of the Junk Invites solution.