Puget Sound Ener­gy (PSE) went live on its new SAP Sales and Ser­vice Cloud Plat­form on June 13. The project team worked tire­less­ly for months prepar­ing to make the tran­si­tion as smooth as possible.

Get­ting that day to pass unno­ticed took near­ly three years of design, led on the busi­ness side by Amy Carp­er, a Senior Busi­ness Tech­nol­o­gy Ana­lyst – Billing and Pay­ment Oper­a­tions, and on the tech­ni­cal side by Jay­want Sub­ra­ma­ni­am, Advi­sor and Appli­ca­tion Con­fig­u­ra­tion Ana­lyst. The project moved PSE off SAP CRM Web IC, the on-premise sys­tem it had run for thir­teen years, and onto SAP Ser­vice Cloud Pri­vate Edition.

In the days after, Carp­er kept hear­ing the same thing from her teams. ​“A lot of peo­ple were say­ing, ​‘I was tense for weeks, and now we’re kind of sit­ting back and excit­ed, because this has been smooth sail­ing,’” she said. The relief traced back to 2013, when PSE first moved to Web IC and Carp­er worked the call cen­ter floor. ​“I remem­ber a trau­mat­ic expe­ri­ence going from that one sys­tem to the next,” she said.

A broad­er mod­ern­iza­tion toward S/4HANA gave PSE room to move, start­ing with Ser­vice Cloud. ​“The sys­tem was get­ting clunky, and we want­ed to adopt some­thing new,” said Sub­ra­ma­ni­am. Carper’s pri­or­i­ty came straight from that pre­vi­ous expe­ri­ence over a decade ago. ​“Our pri­ma­ry thing was to have this be the least impact­ful to the busi­ness as pos­si­ble,” she said.

A Delib­er­ate Detour Around the Big Bang

PSE did not start with a fin­ished list of every­thing that could go wrong. Its safe­guards grew from insti­tu­tion­al mem­o­ry, from col­leagues who had worked on the orig­i­nal imple­men­ta­tion and knew where it hurt. Those pain points emerged in ear­ly work­shops and inter­views. ​“I don’t think we ever sat down and decid­ed that these are the things we have to do,” said Sub­ra­ma­ni­am. As each bot­tle­neck sur­faced, the team sort­ed it into a process prob­lem or a tech­nol­o­gy prob­lem and mit­i­gat­ed it.

The biggest deci­sion from that process con­cerned deploy­ment itself. When they can­vassed peers, the team found none that had gone phased, and even SAP con­firmed that big bang as the norm. When asked what came next for those who had, ​“Every­body said the same thing. We had a lot of trou­ble for the first few months sta­bi­liz­ing the sys­tem,” said Sub­ra­ma­ni­am. That was all PSE need­ed to hear to decide to take a dif­fer­ent route — and Sub­ra­ma­ni­am would urge any util­i­ty to fol­low suit. ​“I’m not sure why we are the out­lier,” he said.

PSE’s exist­ing on-premise CRM and the new Sales and Ser­vice Cloud sys­tem co-exist­ed in par­al­lel for six months, an approach Sub­ra­ma­ni­am calls blue-green. Only Carp­er and a hand­ful of val­ida­tors touched Ser­vice Cloud dur­ing those months; every­one else stayed on Web IC. The new sys­tem ran live in pro­duc­tion the whole time, not in a sand­box, and Carp­er test­ed it like a skep­tic. A cus­tomer her­self, she had a gas-emer­gency order entered on her per­son­al account and called dis­patch to con­firm it arrived, with a warn­ing not to roll a truck, then reprint­ed an invoice to her manager’s home to check that every con­nec­tion between the sys­tems held. This con­trolled envi­ron­ment helped PSE iden­ti­fy and resolve rough­ly 15 pri­or­i­ty-one issues, by her esti­mate, avoid­ing direct cus­tomer impact.

Run­ning the sys­tem live also brought in SAP’s own experts. Through its SAP Max­At­ten­tion pro­gram, SAP ran vol­ume and per­for­mance tests, found the bot­tle­necks slow­ing PSE’s data-heavy screens, and tuned the back­end, adding app servers to bring load times down. ​“If the pro­duc­tion instance was not live, we would nev­er expe­ri­ence those issues ahead of time,” said Subramaniam.

The split between process and tech­nol­o­gy fix­es showed most clear­ly in the secu­ri­ty mod­el, a rebuild two years in the mak­ing. It began with a walk through all 61 depart­ments to record, screen by screen, what each team need­ed to see and do. Dur­ing the par­al­lel run, the PSE team spent two and a half months test­ing them line by line, sur­fac­ing anoth­er 60 to 70 small prob­lems before go-live.

The exer­cise answered a wound from 2013, when a request for one role could qui­et­ly break anoth­er and lock agents out. PSE’s tech­nol­o­gy fix was to auto­mate pro­vi­sion­ing so a request­ed role arrives clean; its process fix was man­u­al test­ing that con­firmed each role worked for its hold­er. That lev­el of care ran through the whole project and suit­ed a shop like PSE, always on the lat­est enhance­ment packs.

A clean green­field build might rea­son­ably choose a big bang instead, but the phased approach paid off at cutover for PSE: with most of the data already migrat­ed by May, only the last month remained for the final week­end, and the switch only took six hours. ​“I don’t say we are an excep­tion­al bunch of peo­ple, but it worked for us,” he said.

Built for Storm Country

A big Pacif­ic North­west storm can knock out pow­er to hun­dreds of thou­sands of PSE cus­tomers for days, the worst time for the call cen­ter to lose its sys­tem. PSE built two answers to that risk. The first was the pri­vate, sin­gle-ten­ant edi­tion of SAP Ser­vice Cloud that it runs; the sec­ond was a fall­back it designed by hand. Ser­vice Cloud earns its keep part­ly on tim­ing, let­ting PSE set SAP’s main­te­nance sched­ule rather than accept it.

The sce­nario Carp­er keeps in mind is a main­te­nance win­dow land­ing on a Sat­ur­day in the mid­dle of a storm, exact­ly the kind of change PSE can now decline. It did so days after go-live, judg­ing a hot­fix SAP had set for June 18 too risky so soon after cutover and get­ting it deferred. Sin­gle ten­an­cy also buys secu­ri­ty and speed, with no oth­er ten­ant to bring out­side inter­fer­ence or slow the sys­tem with a heavy job. ​“If you are the only cus­tomer in the restau­rant, the entire resources are ded­i­cat­ed to you,” not­ed Subramaniam.

The sec­ond answer assumes the first will some­times fail. Even a pri­vate ten­ant has a ceil­ing. SAP’s avail­abil­i­ty sits near 99.7%, leav­ing a win­dow for out­ages, and SAP’s instinct is to build every process inside the plat­form any­way. ​“It was all eggs in one bas­ket,” Sub­ra­ma­ni­am said. So PSE kept some of the crit­i­cal process­es out of Ser­vice Cloud and built its own Fiori-based appli­ca­tions, host­ed in-house.

“We have lit­er­al­ly three lay­ers of redun­dan­cy to min­i­mize oper­a­tional impact,” said Sub­ra­ma­ni­am: a back­up appli­ca­tion, then an access data­base, then paper. Under Web IC, an out­age left the call cen­ter able to take only gas and elec­tric emer­gen­cies; the in-house ver­sion adds billing, pay­ments, and cus­tomer pref­er­ences, with only move-ins, move-outs, and new tick­et cre­ation still stop­ping. Because they run on Fiori, the apps will car­ry for­ward to S/4HANA with­out a rebuild and sync back auto­mat­i­cal­ly when Ser­vice Cloud returns. PSE has not had to use them yet, since Ser­vice Cloud has not gone down.

Bring­ing 61 Depart­ments Along

Plen­ty of peo­ple resist­ed the new sys­tem before ever set­ting foot in a train­ing room. ​“They don’t like change, and that’s com­plete­ly accept­able,” said Carp­er. PSE defused that ear­ly, nam­ing a change leader in each depart­ment to car­ry con­cerns back and hold­ing month­ly progress meet­ings. Train­ing ran for two and a half months ahead of go-live. PSE’s train­ing team doc­u­ment­ed every process with click-by-click screen­shot demos for agents to rehearse on their own time, a first for the company.

Resis­tance fad­ed once peo­ple were inside the sys­tem, which kept most of Web IC’s ter­mi­nol­o­gy and felt famil­iar. The inter­face struck peo­ple as intu­itive, and one reac­tion came back more than any oth­er. ​“Search­ing is so much eas­i­er,” agents told Carp­er repeat­ed­ly. Ser­vice lev­els held far bet­ter than in 2013, and even man­agers who had braced for a cri­sis sought Carp­er out after­ward to relay their relief that it went well.

The les­son Sub­ra­ma­ni­am would press hard­est on the next util­i­ty is one PSE learned the expen­sive way. ​“The data qual­i­ty is of para­mount impor­tance,” he said. Ser­vice Cloud repli­cates back­end records into the front end, and PSE assumed the migra­tion jobs would sim­ply car­ry the data across. ​“That is not true,” said Sub­ra­ma­ni­am, who described the man­u­al cor­rec­tion, pro­gram by pro­gram, that stretched the timeline.

Since SAP gives lit­tle guid­ance on what it will and will not accept, he argues the cleanup should be its own ​“pre-project.” ​“We wast­ed a lot of time cor­rect­ing data,” he said. Carper’s take­away is prac­ti­cal and just as hard-won. Trained in two days on a brand-new sys­tem in 2013, she would tell any com­pa­ny to build a train­ing library like PSE’s.

PSE’s imme­di­ate focus is sta­bi­liza­tion. Hyper­care runs for a few more weeks while it works to get ser­vice lev­els back up and lets agents set­tle in. Beyond that, PSE has stood up a process to triage enhance­ment requests across all 61 depart­ments and will pub­lish a release cal­en­dar show­ing each one and what is com­ing over the next two and six months.

A Sales Cloud use case is already under­way, the platform’s first step beyond cus­tomer ser­vice. Fur­ther out, Sub­ra­ma­ni­am wants Ser­vice Cloud to grow into an enter­prise sys­tem, a sin­gle view of the cus­tomer any team can pull up, so any­one at PSE can find what they need, what­ev­er their role. That vision reach­es beyond any sin­gle go-live.